JP7674176B2 - Text data analysis system, text data analysis method, and computer program - Google Patents

Description

The present invention relates to a text data analysis system, a text data analysis method, and a computer program.

Conventionally, a document reading device that reads a document using an optical character reader is known (Patent Document 1). This technology has an error recognition database that stores misrecognized character strings that include misrecognized characters and correction character strings that correct the misrecognized characters in correspondence with each other, and searches the error recognition database for character strings in read data obtained by reading a document using the optical character reader, and in the case of misrecognized character strings, creates correction data that converts them into the corresponding correction character string. In addition, for characters that were not correctly corrected, the success rate of misrecognition is increased by adding them to the error recognition database.

Patent No. 3349699

However, when searching for a string that matches a given keyword in a master that stores strings representing medical procedures, medicines, and illnesses and injuries, if the string listed in the medical bill or other document is not identical to the string stored in the master, or if the string listed in the medical bill or other document cannot be accurately converted into text data, it is not possible to extract the string representing the medical procedure or other document.

In response to this, there is a conventional technique for increasing the success rate of searches when a character string that matches a keyword is not stored in the master by adding that keyword to the master. However, this technique has the problem that it is not possible to extract a character string before adding a new keyword to the master, and the data capacity of the master increases as new keywords are added to the master.

The present invention has been made in consideration of the above background, and aims to provide a text data analysis system, a text data analysis method, and a computer program that can extract character strings representing medical procedures, medicines, and names of injuries and illnesses from text data without adding more keywords than necessary to the master.

A text data analysis system for achieving the above-mentioned object includes a data acquisition means for acquiring text data, a character string extraction means for extracting a group of character strings representing one item from the text data acquired by the data acquisition means, a first search means for searching a medical practice/drug master that stores character strings representing medical practices or pharmaceuticals for a character string that matches the first keyword, using the character string extracted by the character string extraction means as a first keyword, and if a character string that matches the first keyword is found as a result of the search, a first search means for outputting information on the medical practice or pharmaceutical corresponding to the hit character string, and if a character string that matches the first keyword is not found as a result of the search by the first search means, a first character string generation means for generating a character string from the first keyword by removing the prefix, by referring to a prefix master that stores a prefix, which is a specific character string that is added to the beginning of a character string, and a first string generation means for generating a character string by removing the prefix from the first keyword, and a second search means for searching the medical practice/drug master for a string that matches the second keyword using the string generated by the generating means as a second keyword, and for outputting information on the medical practice or drug corresponding to the hit string when the search results in a string that matches the second keyword; a second string generating means for generating a string from the second keyword by removing at least parentheses and the string enclosed by the parentheses when the search results in no string that matches the second keyword; a similar string extracting means for extracting a string including the third keyword from the medical practice/drug master using the string generated by the second string generating means as a third keyword; and an output means for outputting information on at least one medical practice or drug corresponding to the string extracted by the similar string extracting means.

With such a system, it is possible to extract character strings representing medical procedures or medicines from text data without adding more keywords than necessary to the master.

The text data analysis system may further include a first similarity calculation means for calculating the similarity between each of the character strings extracted by the similar character string extraction means and the second keyword when there are multiple character strings extracted by the similar character string extraction means, and the output means may be configured to output information on medical procedures or medicines corresponding to character strings among the character strings extracted by the similar character string extraction means that have a similarity calculated by the first similarity calculation means that is equal to or greater than a predetermined value.

This allows you to narrow down and output information on medical procedures and medicines.

The medical practice/drug master stores character strings representing one or more medical practices or drugs in correspondence with index character strings that are character strings included in the character strings representing the one or more medical practices or drugs, and the text data analysis system further includes a second similarity calculation means for calculating the similarity between the third keyword and the index character string when the similar character string extraction means is unable to extract a character string including the third keyword from the medical practice/drug master, and the similar character string extraction means can be configured to extract a character string corresponding to the index character string from the medical practice/drug master when there is an index character string whose similarity calculated by the second similarity calculation means is equal to or greater than a predetermined value.

This makes it possible to more reliably extract character strings representing medical procedures or medicines from text data.

The text data analysis system may further include an element string extraction means for extracting at least one element string from the second keyword by referring to an element string master that stores element strings that are specific strings included in strings representing medical procedures or medicines when there is no index string with a similarity calculated by the second similarity calculation means that is equal to or greater than a predetermined value, and the similar string extraction means may be configured to extract a string including the fourth keyword from the medical procedure/medicine master by using the element string extracted by the element string extraction means as a fourth keyword.

This makes it possible to more reliably extract character strings representing medical procedures or medicines from text data.

The text data analysis system may further include a third similarity calculation means for calculating the similarity between the second keyword and the string extracted by the similar string extraction means, using an element string located close to the center of the second keyword as a fourth keyword, before other element strings, when there are multiple element strings extracted by the element string extraction means, and the output means may be configured to output information on a medical procedure or medicine corresponding to the string when a string with a similarity equal to or greater than a predetermined value is included among the strings for which the third similarity calculation means has calculated the similarity first.

This reduces the amount of processing required to output information about medical procedures or medicines, thereby increasing processing speed.

The second character string generating means may be configured to generate a character string by further removing at least one of the following character strings (1) to (5) from the second keyword:
(1) A space at the beginning or end of a word; (2) A space in the middle of a word and the characters following that space; (3) A dot; (4) A comma; (5) A number and the characters immediately following that number that represent the unit.

This makes it easier to narrow down information on medical procedures and medicines.

A text data analysis system for achieving the above-mentioned object includes a data acquisition means for acquiring text data, a character string extraction means for extracting a group of character strings representing one item from the text data acquired by the data acquisition means, a first search means for searching a disease name master storing character strings representing names of injuries or illnesses, using the character string extracted by the character string extraction means as a first keyword, to see if a character string matching the first keyword is found, and if a character string matching the first keyword is found as a result of the search, a first search means for outputting information on the disease name corresponding to the found character string, a first character string generation means for generating a character string from the first keyword by removing the suffix, by referring to a suffix master storing suffixes, which are specific character strings added to the end of a character string, if a character string matching the first keyword is not found as a result of the search by the first search means, and a second character string generation means for generating a character string by removing the suffix from the first keyword, and a second character string generation means for generating a character string by referring to a suffix master storing suffixes, which are specific character strings added to the end of a character string, and a second character string generation means for generating a character string by removing the suffix from the first keyword, if the character string generated by the first character string generation means is not found as a result of the search by the first search means. The present invention is characterized in that it includes a second search means for searching the injury/illness name master as a first keyword to see if a character string matching the second keyword is found, and outputting information on the injury/illness name corresponding to the found character string when a character string matching the second keyword is found as a result of the search by the second search means; a second character string generation means for generating a character string from the second keyword by removing the prefix by referring to a prefix master that stores a prefix, which is a specific character string attached to the beginning of a character string, when a character string matching the second keyword is not found as a result of the search by the second search means; and a third search means for searching the injury/illness name master as a third keyword to see if a character string matching the third keyword is found, and outputting information on the injury/illness name corresponding to the found character string when a character string matching the third keyword is found as a result of the search by the second search means.

With this type of system, it is possible to extract character strings representing the names of injuries and illnesses from text data without adding more keywords than necessary to the master.

The text data analysis system may further include an element string extraction means for extracting at least one element string from the third keyword by referring to an element string master that stores element strings that are specific strings contained in the injury or illness name when the search by the third search means does not hit a string matching the third keyword, a similar string extraction means for extracting a string including the fourth keyword from the injury or illness name master using the element string extracted by the element string extraction means as a fourth keyword, and an output means for outputting information on at least one injury or illness name corresponding to the string extracted by the similar string extraction means.

This makes it possible to more reliably extract character strings representing the names of injuries and illnesses from text data.

The text data analysis system may further include a similarity calculation means for calculating a similarity between the character string extracted by the similar character string extraction means and the third keyword, and the output means may be configured to output information on the names of injuries or illnesses corresponding to character strings among the character strings extracted by the similar character string extraction means that have a similarity calculated by the similarity calculation means that is equal to or greater than a predetermined value.

This allows you to narrow down and output information on injury or illness names.

The similarity calculation means may be configured to, when there are a plurality of element strings extracted by the element string extraction means, calculate the similarity between the element string located close to the center of the third keyword and the third keyword for the string extracted by the similar string extraction means as the fourth keyword before calculating the similarity between the element string and the third keyword for the string extracted by the similar string extraction means ..., when the output means is configured to output information on the name of the injury or illness corresponding to the string if the string includes a string with a similarity equal to or greater than a predetermined value among the strings whose similarity has been calculated earlier.

This reduces the amount of processing required to output the injury or illness information, thereby speeding up the processing speed.

In addition, a text data analysis method for achieving the above-mentioned object includes a computer having means including a data acquisition step for acquiring text data, a string extraction step for extracting a group of strings representing one item from the text data acquired in the data acquisition step, a first search step for searching a medical practice/drug master that stores strings representing medical practices or pharmaceuticals for a string that matches the first keyword using the string extracted in the string extraction step as a first keyword, and if a string that matches the first keyword is found as a result of the search, a first search step for outputting information on the medical practice or pharmaceutical corresponding to the hit string, and a first string generation step for generating a string from the first keyword by removing the prefix, by referring to a prefix master that stores a prefix, which is a specific string that is added to the beginning of a string, if a string that matches the first keyword is not found as a result of the search in the first search step, and The method includes a second search step in which the character string generated in the first character string generation step is used as a second keyword to search the medical practice/medicine master for a character string that matches the second keyword, and if a character string that matches the second keyword is found as a result of the search, a second search step in which information on the medical practice or medicine corresponding to the hit character string is output; a second character string generation step in which, if a character string that matches the second keyword is not found as a result of the search in the second search step, a character string is generated by removing at least parentheses and the character string enclosed by the parentheses from the second keyword; a similar character string extraction step in which, using the character string generated in the second character string generation step as a third keyword, a character string including the third keyword is extracted from the medical practice/medicine master; and an output step in which information on at least one medical practice or medicine corresponding to the character string extracted in the similar character string extraction step is output.

This method makes it possible to extract character strings representing medical procedures or medicines from text data without adding more keywords than necessary to the master.

In addition, a text data analysis method for achieving the above-mentioned object includes a computer having a means for performing a data acquisition step of acquiring text data, a string extraction step of extracting a group of strings representing one item from the text data acquired in the data acquisition step, a first search step of searching a disease name master storing strings representing names of injuries and illnesses for a string matching the first keyword using the string extracted in the string extraction step as a first keyword, and if a string matching the first keyword is found as a result of the search, a first search step of outputting information on the disease name corresponding to the hit string, a first string generation step of generating a string from the first keyword by removing the suffix, by referring to a suffix master storing a suffix that is a specific string added to the end of a string, if a string matching the first keyword is not found as a result of the search in the first search step, and The method includes a second search step in which a character string is used as a second keyword to search the injury/illness name master for a character string that matches the second keyword, and if a character string that matches the second keyword is found as a result of the search, outputting information on the injury/illness name corresponding to the matched character string; a second string generation step in which, if a character string that matches the second keyword is not found as a result of the search in the second search step, a character string is generated from the second keyword by removing the prefix by referring to a prefix master that stores a prefix, which is a specific character string that is added to the beginning of a character string; and a third search step in which, using the character string generated in the second string generation step as a third keyword, the injury/illness master is searched for a character string that matches the third keyword, and if a character string that matches the third keyword is found as a result of the search, outputting information on the injury/illness name corresponding to the matched character string.

This method makes it possible to extract character strings representing the names of illnesses and injuries from text data without adding more keywords than necessary to the master.

A computer program for achieving the above-mentioned object includes a computer including: a data acquisition means for acquiring text data; a character string extraction means for extracting a group of character strings representing one item from the text data acquired by the data acquisition means; a first search means for searching a medical practice/drug master that stores character strings representing medical practices or pharmaceuticals, using the character string extracted by the character string extraction means as a first keyword, to see if a character string matching the first keyword is found; and a first search means for outputting information on the medical practice or pharmaceutical corresponding to the hit character string when a character string matching the first keyword is found as a result of the search by the first search means; a first character string generation means for generating a character string from the first keyword by removing the prefix, by referring to a prefix master that stores a prefix, which is a specific character string added to the beginning of a character string; and a first search means for generating a character string by removing the prefix from the first keyword when a character string matching the first keyword is not found as a result of the search by the first search means. The device is characterized by having a second search means that uses the character string generated by the character string generating means as a second keyword to search the medical practice/medicine master for a character string that matches the second keyword, and when a character string that matches the second keyword is found as a result of the search, outputs information on the medical practice or medicine corresponding to the hit character string; a second character string generating means that generates a character string from the second keyword by removing at least parentheses and the character string surrounded by the parentheses when a character string that matches the second keyword is not found as a result of the search by the second search means; a similar character string extracting means that uses the character string generated by the second character string generating means as a third keyword to extract a character string including the third keyword from the medical practice/medicine master; and an output means that outputs information on at least one medical practice or medicine corresponding to the character string extracted by the similar character string extracting means.

With such a program, it is possible to extract character strings representing medical procedures or medicines from text data without adding more keywords than necessary to the master.

A computer program for achieving the above-mentioned object includes a computer including: a data acquisition means for acquiring text data; a character string extraction means for extracting a group of character strings representing an item from the text data acquired by the data acquisition means; a first search means for searching, using the character string extracted by the character string extraction means as a first keyword, a disease name master storing character strings representing names of injuries or illnesses to see if a character string matching the first keyword is found; and, if a character string matching the first keyword is found as a result of the search, a first search means for outputting information on the disease name corresponding to the found character string; a first character string generation means for, if a character string matching the first keyword is not found as a result of the search by the first search means, referring to a suffix master storing suffixes, which are specific character strings added to the end of a character string, generating a character string from the first keyword by removing the suffix; and a second search means for generating the character string generated by the first character string generation means. A second search means searches the injury/illness name master as a second keyword to see if a character string matching the second keyword is found, and if a character string matching the second keyword is found as a result of the search, outputs information on the injury/illness name corresponding to the found character string; a second string generation means, if a character string matching the second keyword is not found as a result of the search by the second search means, references a prefix master that stores a prefix, which is a specific character string that is added to the beginning of a character string, to generate a character string from the second keyword with the prefix removed; and a third search means, using the character string generated by the second string generation means as a third keyword, searches the injury/illness name master to see if a character string matching the third keyword is found, and if a character string matching the third keyword is found as a result of the search, functions as a third search means that outputs information on the injury/illness name corresponding to the found character string.

With such a program, it is possible to extract character strings representing the names of injuries and illnesses from text data without adding more keywords than necessary to the master.

According to the present invention, it is possible to extract character strings representing medical procedures, medicines, and names of injuries and illnesses from text data without adding more keywords than necessary to the master.

1 is a block diagram of a text data analysis system according to a first embodiment. FIG. 13 is a diagram illustrating the first medical practice/drug master. FIG. 13 is a diagram for explaining the second medical practice/drug master. FIG. 1A is a diagram for explaining a prefix master, and FIG. 1B is a diagram for explaining an element character string master. FIG. 2 is a diagram illustrating a first example of a process in the text data analysis system of the first embodiment. FIG. 11 is a diagram illustrating a second example of the processing in the text data analysis system of the first embodiment. FIG. 11 is a diagram illustrating a third example of a process in the text data analysis system of the first embodiment. FIG. 8 is a diagram continuing from FIG. 7 for explaining a third example of processing in the text data analysis system of the first embodiment. 4 is a flowchart illustrating the operation of the text data analysis system of the first embodiment. 10 is a flowchart continuing from FIG. 9 , illustrating the operation of the text data analysis system of the first embodiment. FIG. 11 is a block diagram of a text data analysis system according to a second embodiment. FIG. 13 is a diagram for explaining the injury/illness name master. 1A is a diagram for explaining a suffix master, FIG. 1B is a diagram for explaining a prefix master, and FIG. 1C is a diagram for explaining an element character string master. FIG. 11 is a diagram illustrating a first example of a process in the text data analysis system of the second embodiment. FIG. 11 is a diagram illustrating a second example of a process in the text data analysis system according to the second embodiment. FIG. 15 is a diagram illustrating a second example of processing in the text data analysis system of the second embodiment. 10 is a flowchart illustrating the operation of the text data analysis system according to the second embodiment.

Next, a first embodiment will be described.
As shown in Fig. 1, the text data analysis system 1 according to the first embodiment is a system that converts and extracts items related to medical procedures and items related to pharmaceuticals from text data created based on items described in, for example, a medical certificate, a medical treatment statement, a prescription statement, etc., into a format included in a basic master defined by the Ministry of Health, Labor and Welfare. The text data analysis system 1 includes a data acquisition means 11, a character string extraction means 12, a first search means 13, a first character string generation means 14, a second search means 15, a second character string generation means 16, a similar character string extraction means 17, a first similarity calculation means 18, an output means 19, a second similarity calculation means 20, an element character string extraction means 21, and a third similarity calculation means 22.

The text data analysis system 1 is made up of a computer including a CPU, RAM, ROM, etc. (not shown), and a storage device 90. The text data analysis system 1 realizes each means by loading a computer program stored in the ROM or storage device 90 into the RAM and executing it. In other words, the computer program causes the computer constituting the text data analysis system 1 to function as a data acquisition means 11, a string extraction means 12, a first search means 13, a first string generation means 14, a second search means 15, a second string generation means 16, a similar string extraction means 17, a first similarity calculation means 18, an output means 19, a second similarity calculation means 20, a component string extraction means 21, and a third similarity calculation means 22.

The storage device 90 stores a medical practice/drug master, a prefix master, and an element character string master.
The medical procedure/drug master stores a character string representing a medical procedure or a drug. Here, the medical procedure includes at least one of the medical procedures listed in the medical procedure master, the medical procedures listed in the dental procedure master, and the dispensing procedures listed in the dispensing procedure master. Furthermore, the drugs include at least one of the drugs listed in the drug master and the specific equipment listed in the specific equipment master. The medical procedure master, the dental procedure master, the dispensing procedure master, the drug master, and the specific equipment master are basic masters established by the Ministry of Health, Labor and Welfare.

In the first embodiment, examples of medical procedures include medical procedures listed in the medical procedure master, and examples of pharmaceuticals include pharmaceuticals listed in the pharmaceutical master.

The medical practice/drug master data includes a first medical practice/drug master data and a second medical practice/drug master data.

As shown in FIG. 2, the first medical procedure/drug master is configured as a table that stores character strings representing medical procedures (abbreviated kanji names and basic kanji names) listed in the medical procedure master or the medical fee schedule in association with fee schedule category numbers, and also stores character strings representing drugs (kanji names and basic kanji names) listed in the drug master in association with drug price standards codes.

As shown in FIG. 3, the second medical procedure/drug master is data prepared to make it easier for the similar string extraction means 17 to extract strings including a third keyword described below, and stores strings representing one or more medical procedures or drugs in correspondence with index strings that are strings included in the strings representing the one or more medical procedures or drugs. In more detail, the second medical procedure/drug master is configured as a table that stores index strings prepared to make it easier to extract strings including the third keyword, such as index strings such as "esophageal resection reconstruction," "endoscopic colonic polyp mucosal resection," and "lens reconstruction," in correspondence with at least one fee schedule classification number or drug price code that corresponds to a string representing a medical procedure or drug that includes the index string.

As shown in FIG. 4(a), the prefix master is configured as a table that stores prefixes, which are specific character strings that are added to the beginning of a character string. A prefix is a character string that may be added to the beginning of a medical procedure or medicine listed on a medical bill, etc., and indicates, for example, the name of a body part or a position such as "lower," "upper," "left," "left side," "right," or "right side." Note that in the present invention, a character string includes a single character. A prefix can be determined, for example, by referring to modifiers used for prefixes listed in the modifier master of the basic master defined by the Ministry of Health, Labor and Welfare.

As shown in FIG. 4(b), the element string master is configured as a table that stores element strings. An element string is a specific string contained in a string representing a medical procedure or medicine, such as "lens" or "insertion." As element strings, strings that do not result in too many candidates being hit when narrowing down strings representing medical procedures or medicines, for example, strings that can narrow down the candidates to around 500 or less, are used.

Returning to FIG. 1, the data acquisition means 11 acquires text data. The data acquisition means 11 may be configured to acquire text data that is created in advance and stored in the storage device 90 or a storage medium, for example, by reading the data, or may be configured to acquire text data that is input by an input device such as a keyboard connected to the text data analysis system 1. The text data created in advance may be, for example, data generated by reading a medical statement or the like with a scanner and using optical character recognition (OCR). The data acquisition means 11 may also be configured to acquire text data generated by reading a medical statement or the like with a scanner connected to the text data analysis system 1 and using OCR.

The character string extraction means 12 extracts a group of character strings representing an item from the text data acquired by the data acquisition means 11. When a group of character strings representing an item exists across multiple lines as shown in FIG. 5(a), the character string extraction means 12 extracts the character strings as a single line of character strings as shown in FIG. 5(b).

The first search means 13 uses the character string extracted by the character string extraction means 12 as a first keyword and searches the medical practice/drug master for a character string that matches the first keyword. In detail, the first search means 13 refers to the first medical practice/drug master and searches for a character string that matches the first keyword. If the search results in a character string that matches the first keyword, the first search means 13 outputs information about the medical practice or drug corresponding to the hit character string.

Here, in the first embodiment, the first search means 13, the second search means 15, and the output means 19 output at least one of the medical procedure name and fee schedule classification number, and the drug name and drug price standard code information as information on the medical procedure or drug. The medical procedure name is at least one of the abbreviated kanji name and basic kanji name listed in the medical procedure master, and the drug name is at least one of the kanji name and basic kanji name listed in the drug master.

The output method is arbitrary. For example, the information may be output by displaying it on a display connected to the text data analysis system 1, by printing it using a printer connected to the text data analysis system 1, or by transmitting the information to a user's terminal connected to the text data analysis system 1 via a network such as the Internet.

When the search by the first search means 13 does not find a string matching the first keyword, the first character string generating means 14 refers to the prefix master and generates a string from the first keyword by removing the prefix. Furthermore, when the first keyword does not have a prefix stored in the prefix master, the first character string generating means 14 regards the generated string as the first keyword.

The second search means 15 uses the character string generated by the first character string generation means 14 as a second keyword and searches the medical practice/drug master for a character string that matches the second keyword. In detail, the second search means 15 refers to the first medical practice/drug master and searches for a character string that matches the second keyword. If the search results in a character string that matches the second keyword, the second search means 15 outputs information about the medical practice or drug corresponding to the hit character string.

The second character string generating means 16 generates a character string by removing parentheses and the character string enclosed by the parentheses from the second keyword when the search by the second search means 15 does not find a character string matching the second keyword. The second character string generating means 16 also generates a character string by further removing the following character strings (1) to (5) from the second keyword.
(1) A space at the beginning or end of a word; (2) A space in the middle of a word and the characters following that space; (3) A dot; (4) A comma; (5) A number and the characters immediately following that number that represent the unit.

As an example, the second character string generating means 16 first performs a process of removing any spaces at the beginning or end of the second keyword, and then performs a process of removing any spaces in the middle and any characters following the spaces. Next, the second character string generating means 16 performs a process of removing parentheses such as parentheses "(", ")" and brackets "", """ and any characters enclosed by such parentheses, and then performs a process of removing a dot "." and a comma ",". Furthermore, the second character string generating means 16 performs a process of removing numbers and any characters representing units immediately following the numbers, such as "10 mg" or "2%".

In addition, when the first keyword and the second keyword are the same character string, the second search means 15 may not search for a character string that matches the second keyword, and the second character string generation means 16 may generate a character string by removing parentheses and the character string enclosed by the parentheses from the second keyword.

In addition, if the second keyword does not contain parentheses, a character string enclosed by the parentheses, or any of the above character strings (1) to (5), the second character string generating means 16 regards the generated character string as the second keyword.

The similar string extraction means 17 extracts a string including the third keyword from the medical practice/drug master using the string generated by the second string generation means 16 as the third keyword. In detail, the similar string extraction means 17 refers to the second medical practice/drug master and the first medical practice/drug master to extract a string including the third keyword (hereinafter also referred to as a "first similar string"). In more detail, the similar string extraction means 17 extracts a fee table category number or drug price code associated with the first similar string from the third keyword and the second medical practice/drug master, and extracts a first similar string associated with the extracted fee table category number or drug price code from the first medical practice/drug master.

When there are multiple first similar strings extracted by the similar string extraction means 17, the first similarity calculation means 18 calculates the similarity between each of the first similar strings extracted by the similar string extraction means 17 and the second keyword.

Here, in the first embodiment, the first similarity calculation means 18, the second similarity calculation means 20, and the third similarity calculation means 22 calculate the similarity between two character strings based on at least one of the Levenshtein distance and the Jaro-Winkler distance, as an example.

The output means 19 outputs information on at least one medical procedure or drug corresponding to the first similar string extracted by the similar string extraction means 17. In particular, when the similar string extraction means 17 extracts one first similar string, the output means 19 outputs information on the medical procedure or drug corresponding to the first similar string. Furthermore, when the similar string extraction means 17 extracts a plurality of first similar strings, the output means 19 outputs information on the medical procedure or drug corresponding to the first similar string among the plurality of first similar strings, the similarity calculated by the first similarity calculation means 18 being equal to or greater than a predetermined value.

In the first embodiment, the similarity is calculated as a value between 0 and 1, with a larger value indicating a higher similarity between the two character strings, and a value of 1 indicating that the two character strings match. The specified value of the similarity is, for example, 0.8. Note that the text data analysis system 1 may be configured to allow the user to arbitrarily set the specified value of the similarity.

If the similar string extraction means 17 is unable to extract a string (first similar string) containing the third keyword from the medical practice/drug master, the second similarity calculation means 20 calculates the similarity between the third keyword generated by the second string generation means 16 and the index string stored in the second medical practice/drug master.

If there is an index string whose similarity calculated by the second similarity calculation means 20 is equal to or greater than a predetermined level, the similar string extraction means 17 extracts a string corresponding to the index string from the medical procedure/drug master. In detail, if there is an index string whose similarity to the third keyword is equal to or greater than a predetermined level, the similar string extraction means 17 extracts the point table category number or drug price code corresponding to the similar string from the second medical procedure/drug master, and extracts a string (hereinafter also referred to as a "corresponding string") corresponding to the extracted point table category number or drug price code from the first medical procedure/drug master.

When there are multiple corresponding character strings extracted by the similar character string extraction means 17, the first similarity calculation means 18 calculates the similarity between each of the corresponding character strings extracted by the similar character string extraction means 17 and the second keyword, and the output means 19 outputs information on the medical procedure or medicine that corresponds to the corresponding character string whose similarity calculated by the first similarity calculation means 18 is equal to or greater than a predetermined value.

If there is no index string whose similarity calculated by the second similarity calculation means 20 is equal to or greater than a predetermined value, the element string extraction means 21 refers to the element string master and extracts at least one element string from the second keyword.

The similar string extraction means 17 extracts strings including the fourth keyword from the medical practice/medicine master, using the element string extracted by the element string extraction means 21 as the fourth keyword. In more detail, the similar string extraction means 17 extracts strings including the fourth keyword (hereinafter also referred to as "second similar strings") from the first medical practice/medicine master. In the first embodiment, when there are multiple element strings extracted by the element string extraction means 21, the similar string extraction means 17 first extracts the second similar string from the medical practice/medicine master, using the element string (hereinafter also referred to as "priority element string" in the first embodiment) located close to the center of the second keyword as the fourth keyword.

The third similarity calculation means 22 calculates the similarity between the second keyword and the second similar string extracted by the similar string extraction means 17. In detail, the third similarity calculation means 22 calculates the similarity between the second keyword and the second similar string extracted by the similar string extraction means 17 with the priority element string as the fourth keyword before calculating the similarity between the second keyword and the other element strings.

If there is a string whose similarity is equal to or greater than a predetermined level among the second similar strings whose similarity has been previously calculated by the third similarity calculation means 22, the output means 19 outputs information about the medical procedure or medicine corresponding to the second similar string. When the output means 19 outputs information about the medical procedure or medicine corresponding to the second similar string whose similarity has been previously calculated, thereafter, the similar string extraction means 17 does not extract second similar strings for other element strings, and the third similarity calculation means 22 does not calculate similarity.

If there is no string with a similarity equal to or greater than a predetermined level among the second similar strings for which similarity has been previously calculated, the similar string extraction means 17 next extracts a second similar string from the first medical procedure/drug master using an element string located close to the center of the second keyword as a fourth keyword, and the third similarity calculation means 22 calculates the similarity between the extracted second similar string and the second keyword. Then, if there is a string with a similarity equal to or greater than a predetermined level among the second similar strings for which similarity has been calculated, the output means 19 outputs information on the medical procedure or drug corresponding to the second similar string.

When the element string extraction means 21 extracts a plurality of element strings, and the positions of the second keyword from the center of the two element strings are the same, the similar string extraction means 17 extracts second similar strings from the medical practice/drug master with each element string as a fourth keyword, and calculates the similarity between the extracted second similar strings and the second keyword. Then, if there is a string with a predetermined similarity or higher among the second similar strings for which similarity has been calculated, the output means 19 outputs information on the medical practice or drug corresponding to the second similar string. Note that in this case, the similarity may be calculated for the one with fewer extracted second similar strings before the one with a larger number, and if there is a string with a predetermined similarity or higher among the second similar strings for which similarity has been calculated first, information on the medical practice or drug corresponding to the second similar string may be output and processing may end.

Here, the processing in the text data analysis system 1 of the first embodiment will be described with reference to a specific example.
As a first example, as shown in FIG. 5(c), the first search means 13 uses the character string “left lens reconstruction (intraocular lens insertion and others)” acquired by the data acquisition means 11 and extracted by the character string extraction means 12 as the first keyword, and searches the first medical procedure/drug master for a character string that matches the first keyword.

If the search by the first search means 13 does not yield a string matching the first keyword, the first string generation means 14 refers to the prefix master and generates the string "Lens reconstruction surgery (when inserting an intraocular lens or other procedures)" by removing the prefix "left" from the first keyword, as shown in FIG. 5(d).

The second search means 15 uses the character string generated by the first character string generation means 14, "Lens reconstruction surgery (intraocular lens insertion and others)" as a second keyword, and searches the first medical procedure/drug master for a character string that matches the second keyword.

If the search by the second search means 15 does not yield a string matching the second keyword, the second string generation means 16 generates the string "Lens reconstruction surgery" by removing the parentheses and the string enclosed by the parentheses "(when inserting an intraocular lens/other)" from the second keyword, as shown in FIG. 5(e).

The similar string extraction means 17 uses the string "lens reconstruction surgery" generated by the second string generation means 16 as the third keyword, and extracts from the second medical procedure/drug master the point table category numbers "K2821I", "K2821B", "K2822", and "K2823" that are associated with the string containing the third keyword (first similar string). The similar string extraction means 17 then extracts from the first medical procedure/drug master the four first similar strings shown in FIG. 5(f) that are associated with the extracted point table category numbers.

When there are multiple first similar strings extracted by the similar string extraction means 17, the first similarity calculation means 18 calculates the similarity between each of the first similar strings and the second keyword "lens reconstruction surgery (intraocular lens insertion and others)" as shown in FIG. 5(g).

The output means 19 outputs information on medical procedures and medicines corresponding to the first similar strings, among the first similar strings extracted by the similar string extraction means 17, whose similarity calculated by the first similarity calculation means 18 is equal to or greater than a predetermined value of 0.8. Specifically, the output means 19 outputs information on the medical procedure name "Lens reconstruction surgery (when an intraocular lens is inserted) (other)" and the point table category number "K2821ro", and the medical procedure name "Lens reconstruction surgery (when an intraocular lens is not inserted)" and the point table category number "K2822". The output means 19 may be configured to further output information on the similarity calculated by the first similarity calculation means 18.

Next, we will explain the second example. In the second example, the same character string as in the first example, "Left lens reconstruction surgery (in case of inserting an intraocular lens, etc.)", is mistakenly read as "water crystal reconstruction surgery" by OCR.

In the second example, as shown in FIG. 6(a), the first search means 13 searches the first medical procedure/drug master for a character string that matches the first keyword, using the character string "left hydrate resection (when inserting an intraocular lens/other)" as the first keyword. Since there is no match, the first character string generation means 14 refers to the prefix master and generates the character string "hydrate resection (when inserting an intraocular lens/other)" by removing the prefix "left" from the first keyword, as shown in FIG. 6(b).

The second search means 15 searches the first medical procedure/drug master for a character string that matches the second keyword, using the character string "Hydrangea resection (when inserting an intraocular lens/other)" as the second keyword, and since there is no hit, the second character string generation means 16 generates the character string "Left hydrangea resection" by removing the parentheses and the character string "(when inserting an intraocular lens/other)" enclosed by the parentheses from the second keyword, as shown in FIG. 6(c).

The similar string extraction means 17 attempts to extract a string containing the third keyword from the second medical practice/drug master using the string "Suichangtai Sairyunjutsu" as the third keyword, but in the second example, it is not possible to extract a string containing the third keyword. Therefore, as shown in FIG. 6(d), the second similarity calculation means 20 calculates the similarity between the third keyword "Suichangtai Sairyunjutsu" and each index string stored in the second medical practice/drug master.

When there is an index string whose similarity calculated by the second similarity calculation means 20 is equal to or greater than a predetermined value of 0.8, the similar string extraction means 17 extracts a string that corresponds to the index string (corresponding string) from the first medical practice/drug master. Specifically, since there is an index string "lens reconstruction surgery" whose similarity is equal to or greater than a predetermined value of 0.8, the similar string extraction means 17 extracts the four corresponding strings shown in FIG. 5(e) from the first medical practice/drug master.

When there are multiple corresponding character strings extracted by the similar character string extraction means 17, the first similarity calculation means 18 calculates the similarity between each corresponding character string and the second keyword “water crystal resection (in case of inserting an intraocular lens and others)” as shown in FIG. 6(f).
The output means 19 outputs information on medical procedures and medicines corresponding to corresponding strings extracted by the similar string extraction means 17, the similarity calculated by the first similarity calculation means 18 being a predetermined value of 0.8 or more.

Next, we will explain the third example. In the third example, the item "Short hand 3 (Lens reconstruction surgery, intraocular lens insertion, other items (one side))" written on the medical statement etc. is mistakenly read as "Fence 3 (Lens reconstruction finger, intraocular lens insertion, other items (one side))" by OCR.

In the case of the third example, as shown in FIG. 7(a), the first search means 13 searches the first medical practice/drug master for a character string that matches the first keyword, using the character string "Kakide 3 (Lens reconstruction finger/Intra-capsule lens insertion/Other (One example))" as the first keyword, and since there is no match, the first character string generation means 14 attempts to generate a character string from the first keyword with the prefix removed, referring to the prefix master. In the third example, since the first keyword does not have a prefix stored in the prefix master, as shown in FIG. 7(b), the first character string generation means 14 generates the same character string as the first keyword, "Kakide 3 (Lens reconstruction finger/Intra-capsule lens insertion/Other (One example))".

The second search means 15 searches the first medical practice/drug master for a character string that matches the second keyword, using the character string "Kakide 3 (Lens reconstruction finger/Intra-lens insertion/Other (example))" as the second keyword, and since there is no hit, the second character string generation means 16 generates the character string "Kakide 3" by removing the parentheses and the character string "(Lens reconstruction finger/Intra-lens insertion/Other (example))" enclosed by the parentheses from the second keyword, as shown in FIG. 7(c).

The similar string extraction means 17 attempts to extract a string containing the third keyword from the second medical practice/drug master using the string "Fence 3" as the third keyword, but in the third example, it is unable to extract a string containing the third keyword. Therefore, the second similarity calculation means 20 calculates the similarity between the third keyword "Fence 3" and each index string stored in the second medical practice/drug master.

In the third example, all of the similarities calculated by the second similarity calculation means 20 are less than the predetermined value of 0.8, and as shown in FIG. 7(d), the element string extraction means 21 refers to the element string master and extracts the element strings "lens" and "insertion" from the second keyword (see FIG. 7(b)).

Since there are multiple element strings extracted by the element string extraction means 21, the similar string extraction means 17 first selects the element string "lens" located near the center of the second keyword as the fourth keyword, and extracts strings (second similar strings) containing the fourth keyword "lens" from the first medical practice/drug master, as shown in Figure 8 (a).

As shown in FIG. 8(b), the third similarity calculation means 22 calculates the similarity between the extracted second similar character string and the second keyword "Kakide 3 (cataract reconstruction finger, intraocular lens insertion, other (example))".

The output means 19 outputs information on medical procedures or medicines corresponding to similar strings whose similarity calculated by the third similarity calculation means 22 is equal to or greater than a predetermined value of 0.8 among the second similar strings extracted by the similar string extraction means 17. Specifically, the output means 19 outputs information on the medical procedure name "Short 3 (Lens reconstruction, intraocular lens insertion, other, one side)" and the point table classification number "A4003ho", the medical procedure name "Short 3 (Lens reconstruction, intraocular lens insertion, other, both sides)" and the point table classification number "A4003he", and the medical procedure name "Short 3 (Lens reconstruction, intraocular lens insertion, other, one side) (lifestyle care)" and the point table classification number "A4003ho". The output means 19 may be configured to further output information on the similarity calculated by the third similarity calculation means 22.

Next, an example of the operation of the text data analysis system 1 of the first embodiment (a text data analysis method executed by the means provided in the computer constituting the text data analysis system 1) will be described with reference to a flowchart.

As shown in FIG. 9, the text data analysis system 1 first acquires text data (S110) (data acquisition step). Next, the text data analysis system 1 extracts a group of character strings representing one item from the text data acquired in the data acquisition step (S120) (character string extraction step).

Next, the text data analysis system 1 uses the character string extracted in the character string extraction step as a first keyword and searches the first medical practice/drug master for a character string that matches the first keyword (S131) (first search step). Then, if the search results in a character string that matches the first keyword (S132, Yes), the system proceeds to step S183.

On the other hand, if the search in the first search step does not result in a string that matches the first keyword (S132, No), the text data analysis system 1 refers to the prefix master and generates a string from the first keyword with the prefix removed (S140) (first string generation step).

Next, the text data analysis system 1 uses the character string generated in the first character string generation step as a second keyword and searches the first medical practice/drug master for a character string that matches the second keyword (S151) (second search step). If the search results in a character string that matches the second keyword (S152, Yes), the system proceeds to step S183.

On the other hand, if the search in the second search step does not result in a string that matches the second keyword (S152, No), the text data analysis system 1 generates a string from the second keyword by removing parentheses and the string enclosed by the parentheses (S160) (second string generation step).

Next, the text data analysis system 1 extracts a string containing the third keyword (first similar string) from the second medical practice/drug master and the first medical practice/drug master using the string generated in the second string generation step as the third keyword (S171) (similar string extraction step). Then, the text data analysis system 1 determines whether the first similar string has been extracted (S172).

If a first similar string is extracted (S172, Yes), the text data analysis system 1 determines whether multiple first similar strings are extracted (S173). If only one first similar string is extracted (S173, No), the process proceeds to step S183.

On the other hand, if there are multiple strings extracted in the similar string extraction step (S173, Yes), the text data analysis system 1 calculates the similarity between each of the first similar strings extracted in the similar string extraction step and the second keyword (S181) (first similarity calculation step). Then, the text data analysis system 1 extracts, from among the first similar strings extracted in the similar string extraction step, first similar strings whose similarity calculated in the first similarity calculation step is equal to or greater than a predetermined value (S182), and proceeds to step S183.

For example, a medical care statement or a prescription statement may contain multiple items related to medical procedures such as the name of surgery or items related to medicines, or may contain items other than the items in question, so that the text data may contain multiple groups of character strings representing one item. Therefore, in step S183, the text data analysis system 1 determines whether there is a next group of character strings representing one item. If there is a next character string (S183, Yes), the text data analysis system 1 returns to step S120 and executes the subsequent processes, and if there is no next character string (S183, No), the text data analysis system 1 proceeds to step S191.

In step S191, the text data analysis system 1 extracts information on medical procedures and medicines that correspond to the character strings found in the first search step or the second search step, and outputs the extracted information on medical procedures and medicines (S192). Alternatively, the text data analysis system 1 extracts information on at least one medical procedure and medicine that corresponds to the first similar character string extracted in the similar character string extraction step (S191), and outputs the extracted information on medical procedures and medicines (S192) (output step).

In step S172, if a string containing the third keyword (first similar string) cannot be extracted from the second medical practice/drug master and the first medical practice/drug master (No), as shown in FIG. 10, the text data analysis system 1 calculates the similarity between the third keyword and each index string stored in the second medical practice/drug master (S201) (second similarity calculation step).

Then, the text data analysis system 1 determines whether there is an index string whose similarity calculated in the second similarity calculation step is equal to or greater than a predetermined level (S202). If there is an index string whose similarity is equal to or greater than a predetermined level (S202, Yes), the text data analysis system 1 extracts a string that corresponds to the index string (corresponding string) from the second medical practice/drug master and the first medical practice/drug master (S203).

Then, the text data analysis system 1 proceeds to step S173 in FIG. 9 to determine whether there are multiple extracted corresponding character strings. If there is one (S173, No), it proceeds to step S183. If there are multiple extracted corresponding character strings (S173, Yes), it calculates the similarity between each of the extracted corresponding character strings and the second keyword (S181) and executes the subsequent processes.

Returning to FIG. 10, in step S202, if there is no index string whose similarity calculated in the second similarity calculation step (S201) is equal to or greater than a predetermined value (S202, No), the text data analysis system 1 refers to the element string master and extracts element strings from the second keyword (S211) (element string extraction step).

Next, the text data analysis system 1 determines whether multiple element strings have been extracted (S212). If multiple element strings have been extracted (S212, Yes), the text data analysis system 1 first determines the element string from which the second similar string will be extracted, specifically, the element string located close to the center of the second keyword (S213), and extracts a string containing the fourth keyword (second similar string) from the first medical practice/drug master using the determined element string as the fourth keyword (S214). If only one element string has been extracted (S212, No), the text data analysis system 1 extracts the second similar string from the first medical practice/drug master using the element string as the fourth keyword (S214).

Next, the text data analysis system 1 calculates the similarity between the extracted second similar string and the second keyword (S221) (third similarity calculation step). Next, the text data analysis system 1 determines whether there is a second similar string whose calculated similarity is equal to or greater than a predetermined level (S222). If there is no second similar string whose similarity is equal to or greater than a predetermined level (S222, No), the process returns to step S213 to determine the next element string from which the second similar string is to be extracted first, and executes the subsequent processes. On the other hand, if there is a second similar string whose similarity is equal to or greater than a predetermined level in step S222 (Yes), the text data analysis system 1 proceeds to step 182 in FIG. 9, extracts a second similar string whose similarity is equal to or greater than a predetermined level (S182), and executes the subsequent processes.

According to the first embodiment described above, character strings representing medical procedures or medicines can be extracted from text data without adding more keywords than necessary to the master. In addition, character strings representing medical procedures or medicines can be extracted from text data by converting them into a format included in the basic master defined by the Ministry of Health, Labor and Welfare.

Furthermore, by further providing a first similarity calculation means 18, it is possible to narrow down and output information on medical procedures and medicines.

In addition, by further providing a second similarity calculation means 20, character strings representing medical procedures or medicines can be extracted more reliably from the text data.

In addition, by further providing an element string extraction means 21, it is possible to more reliably extract strings representing medical procedures or medicines from text data.

The system further includes a third similarity calculation means 22, which, if there is a second similar string whose similarity is equal to or greater than a predetermined level among the second similar strings whose similarity has been calculated previously, outputs information on the name of the injury or illness corresponding to the second similar string and terminates the process, thereby reducing the amount of processing required to output information on the medical procedure or medicine, thereby increasing the processing speed.

In addition, the second character string generating means 16 not only removes parentheses and character strings enclosed by those parentheses from the second keywords, but also removes the above character strings (1) to (5), making it easier to narrow down information on medical procedures and medicines.

In the first embodiment, the second character string generating means 16 removes the above character strings (1) to (5) from the second keyword in the second character string generating step, but it is sufficient that the second character string generating means 16 is configured to remove at least one of the above character strings (1) to (5) from the second keyword. Also, for example, the second character string generating means 16 may be configured to remove only parentheses and character strings enclosed by the parentheses from the second keyword in the second character string generating step.

In the first embodiment, the similar string extraction means 17 extracts the first similar string by referring to the second medical practice/drug master and the first medical practice/drug master, but, for example, the first similar string may be extracted by referring to only the first medical practice/drug master. In other words, the text data analysis system may not be provided with the second medical practice/drug master. Also, the index string may be stored in the first medical practice/drug master in association with a string representing a medical practice/drug, etc.

In addition, in the first embodiment, the medical procedure/drug master stored the fee schedule category number and drug price standard code, but it may also store other codes, such as the medical procedure code from the medical procedure master or the drug code from the drug master.

Next, the second embodiment will be described. Note that, below, differences from the first embodiment will be described in detail, and explanations of the same points will be omitted as appropriate by assigning the same reference numerals to the same elements.

As shown in FIG. 11, the text data analysis system 1 according to the second embodiment is a system that converts and extracts items related to the names of illnesses and injuries from text data created based on items written on a medical certificate or the like into a format included in the ICD10-compatible standard disease name master. The text data analysis system 1 includes a data acquisition means 11, a character string extraction means 12, a first search means 23, a first character string generation means 24, a second search means 25, a second character string generation means 26, a third search means 27, an element character string extraction means 28, a similar character string extraction means 29, a similarity calculation means 30, an output means 31, and a storage device 90.

In the second embodiment, the computer program stored in the ROM or storage device 90 causes the computer constituting the text data analysis system 1 to function as a data acquisition means 11, a string extraction means 12, a first search means 23, a first string generation means 24, a second search means 25, a second string generation means 26, a third search means 27, an element string extraction means 28, a similar string extraction means 29, a similarity calculation means 30, and an output means 31.

The storage device 90 stores an injury/disease name master, a suffix master, a prefix master, and an element character string master.
As shown in Fig. 12, the injury/disease name master is configured as a table that stores character strings representing injury/disease names. The injury/disease name master stores character strings representing injury/disease names included in the ICD10-compatible standard illness name master in association with ICD codes.

As shown in FIG. 13(a), the suffix master is configured as a table that stores suffixes, which are specific character strings that are added to the end of a character string. Suffixes are character strings that are sometimes added to the end of the name of an injury or illness written on a medical certificate, etc., such as "suspected of," "post-surgery of," "pre-surgery of," "exacerbation of," "post-treatment of," and "secondary infection of." Suffixes can be determined, for example, by referring to modifiers used for suffixes listed in the modifier master.

As shown in FIG. 13(b), the prefix master is configured as a table that stores prefixes, which are specific character strings that are added to the beginning of a character string. Prefixes are character strings such as "lower", "acute", "upper", "left", "left side", "right", and "right side" that are sometimes added to the beginning of the name of an injury or illness written on a medical certificate, etc. Prefixes can be determined, for example, by referring to modifiers used for prefixes listed in the modifier master.

As shown in FIG. 13(c), the element string master is configured as a table that stores element strings. An element string is a specific string contained in the name of an injury or illness, such as "pearl," "tendon," "room," "ear," "shrink," "crush," and "middle ear." As element strings, strings are used that do not generate too many candidates when narrowing down the injury or illness names, for example, strings that can narrow down the candidates to around 500 or less.

Returning to FIG. 11, the first search means 23 uses the character string extracted by the character string extraction means 12 as a first keyword and searches the injury/illness name master for a character string that matches the first keyword. If the search finds a character string that matches the first keyword, the first search means 23 outputs information on the injury/illness name that corresponds to the hit character string.

Here, in the second embodiment, the first search means 23, the second search means 25, the third search means 27 and the output means 31 output information on the injury or illness name and the ICD code.

When the search by the first search means 23 does not find a string matching the first keyword, the first character string generating means 24 refers to the suffix master and generates a string from the first keyword by removing the suffix. Also, when the first keyword does not have a suffix stored in the suffix master, the first character string generating means 24 sets the generated string as the first keyword.

The second search means 25 uses the character string generated by the first character string generation means 24 as a second keyword and searches the injury/illness name master for a character string that matches the second keyword. If the search finds a character string that matches the second keyword, the second search means 25 outputs information on the injury/illness name that corresponds to the matched character string.

If the search by the second search means 25 does not find a string that matches the second keyword, the second string generation means 26 refers to the prefix master and generates a string from the second keyword with the prefix removed.

In addition, when the first keyword and the second keyword are the same character string, the second search means 25 may not search for a character string that matches the second keyword, and the second character string generation means 26 may generate a character string from the second keyword by removing the suffix.

In addition, if the second keyword does not have a prefix stored in the prefix master, the second character string generating means 26 treats the generated character string as the second keyword.

The third search means 27 uses the character string generated by the second character string generation means 26 as a third keyword and searches the injury/illness name master for a character string that matches the third keyword. If the search finds a character string that matches the third keyword, the third search means 27 outputs information on the injury/illness name that corresponds to the matched character string.

When the search by the third search means 27 does not find a string that matches the third keyword, the element string extraction means 28 refers to the element string master and extracts at least one element string from the third keyword.

In addition, if the second keyword and the third keyword are the same character string, the third search means 27 may not search for a character string that matches the third keyword, and the element string extraction means 28 may extract the element string from the third keyword.

The similar string extraction means 29 uses the element string extracted by the element string extraction means 28 as the fourth keyword and extracts strings (hereinafter also referred to as "similar strings") that include the fourth keyword from the injury/illness name master.

The similarity calculation means 30 calculates the similarity between the similar string extracted by the similar string extraction means 29 and the third keyword. As an example, the similarity calculation means 30 calculates the similarity between the similar string and the third keyword based on at least one of the Levenshtein distance and the Jaro-Winkler distance. In the second embodiment, the similarity is also calculated as a numerical value between 0 and 1.

The output means 31 outputs information on at least one injury or illness name corresponding to the similar string extracted by the similar string extraction means 29. In detail, the output means 31 outputs information on the injury or illness name corresponding to a string having a similarity calculated by the similarity calculation means 30 equal to or greater than a predetermined value, among the similar strings extracted by the similar string extraction means 29.

In the second embodiment, when there are multiple element strings extracted by the element string extraction means 28, the similar string extraction means 29 first extracts similar strings from the injury/illness name master using the element string (hereinafter, also referred to as the "priority element string" in the second embodiment) that is located close to the center of the third keyword as the fourth keyword.

The similarity calculation means 30 calculates the similarity between the priority element string and the third keyword before other element strings for the similar strings extracted by the similar string extraction means 29 using the priority element string as the fourth keyword.

If there is a string whose similarity is equal to or greater than a predetermined level among the similar strings whose similarity has been calculated previously, the output means 31 outputs information on the name of the injury or illness corresponding to the similar string. When the output means 31 outputs information on the name of the injury or illness corresponding to the similar string whose similarity has been calculated previously, the similar string extraction means 29 does not thereafter extract similar strings for other element strings, and the similarity calculation means 30 does not calculate similarity.

If there is no string with a similarity equal to or greater than a predetermined level among the similar strings whose similarity has been calculated earlier, the similar string extraction means 29 next extracts a similar string from the injury/illness name master using an element string located close to the center of the third keyword as the fourth keyword, and the similarity calculation means 30 calculates the similarity between the extracted similar string and the third keyword. Then, if there is a string with a similarity equal to or greater than a predetermined level among the similar strings whose similarity has been calculated, the output means 31 outputs information on the injury/illness name corresponding to the similar string.

When the element string extraction means 28 extracts multiple element strings, and the positions of the third keyword from the center of two element strings are the same, the similar string extraction means 29 extracts similar strings from the illness/injury name master with each element string as the fourth keyword, and calculates the similarity between the extracted similar strings and the third keyword. Then, if there is a string with a predetermined similarity or higher among the similar strings whose similarity has been calculated, the output means 31 outputs information on the illness/injury name corresponding to the similar string. Note that in this case, the similarity may be calculated for the string with fewer extracted similar strings before the string with a larger number, and if there is a string with a predetermined similarity or higher among the similar strings whose similarity has been calculated first, information on the illness/injury name corresponding to the similar string may be output and the process terminated.

Here, the processing in the text data analysis system 1 of the second embodiment will be described with reference to a specific example.
As a first example, as shown in FIG. 14(a), the first search means 23 uses the character string “suspected acute influenza pneumonia” acquired by the data acquisition means 11 and extracted by the character string extraction means 12 as a first keyword, and searches the injury/illness name master for a character string that matches the first keyword.

If the search by the first search means 23 does not result in a string matching the first keyword, the first string generation means 24 refers to the suffix master and generates the string "acute influenza pneumonia" by removing the suffix "suspected of" from the first keyword, as shown in FIG. 14(b).

The second search means 25 uses the character string "acute influenza pneumonia" generated by the first character string generation means 24 as a second keyword and searches the injury/disease name master for a character string that matches the second keyword.

If the search by the second search means 25 does not result in a string matching the second keyword, the second string generation means 26 refers to the prefix master and generates the string "influenza pneumonia" by removing the prefix "acute" from the second keyword, as shown in FIG. 14(c).

The third search means 27 uses the character string "influenza pneumonia" generated by the second character string generation means 26 as a third keyword and searches the illness name master for a character string that matches the third keyword. If a character string that matches the third keyword is found as a result of the search, the third search means 27 outputs information on the illness name corresponding to the hit character string. Specifically, the third search means 27 outputs information on the illness name "influenza pneumonia" and the ICD code "J110", as shown in FIG. 14(d).

As a second example, as shown in FIG. 15(a), the first search means 23 uses the character string "Postoperative period for right pearlescent otitis media" acquired by the data acquisition means 11 and extracted by the character string extraction means 12 as the first keyword, and searches the injury/disease name master for a character string that matches the first keyword.

If the search by the first search means 23 does not result in a string matching the first keyword, the first string generation means 24 refers to the suffix master and generates the string "right pearlescent otitis media" by removing the suffix "post-operative" from the first keyword, as shown in FIG. 15(b).

The second search means 25 uses the character string "right pearlescent otitis media" generated by the first character string generation means 24 as a second keyword and searches the injury/disease name master for a character string that matches the second keyword.

If the search by the second search means 25 does not result in a string matching the second keyword, the second string generation means 26 refers to the prefix master and generates the string "pearlogenic otitis media" by removing the prefix "right" from the second keyword, as shown in Figure 15 (c).

The third search means 27 uses the character string "pearlogenic otitis media" generated by the second character string generation means 26 as a third keyword and searches the injury/disease name master for a character string that matches the third keyword.

If the search by the third search means 27 does not find a string that matches the third keyword, the element string extraction means 28 refers to the element string master and extracts the element strings "pearl" and "middle ear" from the third keyword, as shown in FIG. 15(d).

Since there are multiple element strings extracted by the element string extraction means 28, the similar string extraction means 29 first selects the element string "middle ear" located near the center of the third keyword as the fourth keyword, and extracts strings (similar strings) containing the fourth keyword "middle ear" from the injury/illness name master, as shown in FIG. 16(a).

As shown in FIG. 16(b), the similarity calculation means 30 calculates the similarity between each of the extracted similar strings and the third keyword "pearlogenic otitis media."

The output means 31 outputs information on the disease name corresponding to the similar character strings whose similarity calculated by the similarity calculation means 30 is a predetermined value, for example, 0.8 or more, among the similar character strings extracted by the similar character string extraction means 29. Specifically, the output means 31 outputs information on the disease name "cholesteatomatous otitis media" and the ICD code "H71", the disease name "acute otitis media" and the ICD code "H669", and the disease name "chronic otitis media" and the ICD code "H669". The output means 31 may be configured to further output information on the similarity calculated by the similarity calculation means 30.

Next, an example of the operation of the text data analysis system 1 of the second embodiment will be described with reference to a flowchart.

As shown in FIG. 17, the text data analysis system 1 uses the character string extracted in the character string extraction step (S120) as a first keyword and searches the injury/disease name master for a character string that matches the first keyword (S231) (first search step). Then, if the search results in a character string that matches the first keyword (S232, Yes), the process proceeds to step S304.

On the other hand, if the search in the first search step does not result in a string that matches the first keyword (S232, No), the text data analysis system 1 refers to the suffix master and generates a string from the first keyword with the suffix removed (S240) (first string generation step).

Next, the text data analysis system 1 uses the character string generated in the first character string generation step as a second keyword and searches the injury/illness name master for a character string that matches the second keyword (S251) (second search step). If the search results in a character string that matches the second keyword (S252, Yes), the system proceeds to step S304.

On the other hand, if the search in the second search step does not result in a string that matches the second keyword (S252, No), the text data analysis system 1 refers to the prefix master and generates a string from the second keyword with the prefix removed (S260) (second string generation step).

Next, the text data analysis system 1 uses the character string generated in the second character string generation step as a third keyword and searches the injury/illness name master for a character string that matches the third keyword (S271) (third search step). If the search results in a character string that matches the third keyword (S272, Yes), the system proceeds to step S304.

On the other hand, if the search in the third search step does not yield a string that matches the third keyword (S272, No), the text data analysis system 1 refers to the element string master and extracts an element string from the third keyword (S280) (element string extraction step).

Next, the text data analysis system 1 determines whether multiple element strings have been extracted (S291). If multiple element strings have been extracted (S291, Yes), the text data analysis system 1 first determines the element string from which to extract a similar string, specifically, the element string located close to the center of the third keyword (S292), and extracts strings containing the fourth keyword (similar strings) from the injury/illness name master using the determined element string as the fourth keyword (S293). If only one element string has been extracted (S291, No), the text data analysis system 1 extracts similar strings from the injury/illness name master using the element string as the fourth keyword (S293) (similar string extraction step).

Next, the text data analysis system 1 calculates the similarity between the extracted similar string and the third keyword (S301) (similarity calculation step). Next, the text data analysis system 1 determines whether there is a similar string whose calculated similarity is equal to or greater than a predetermined level (S302). If there is no similar string whose similarity is equal to or greater than a predetermined level (S302, No), the process returns to step S292 to determine the next element string from which a similar string will be extracted first, and executes subsequent processing. On the other hand, if there is a similar string whose similarity is equal to or greater than a predetermined level in step S302 (Yes), the text data analysis system 1 extracts a similar string whose similarity is equal to or greater than a predetermined level (S303) and proceeds to step S304.

In step S304, the text data analysis system 1 determines whether there is a next group of character strings representing an item. If there is a next character string (S304, Yes), the process returns to step S120 and executes the subsequent processes. If there is no next character string (S304, No), the process proceeds to step S311.

In step S311, the text data analysis system 1 extracts information on the injury or illness name corresponding to the character string hit in the first search step, the second search step, or the third search step, and outputs the extracted injury or illness name information (S312). Alternatively, the text data analysis system 1 extracts information on at least one injury or illness name corresponding to the similar character string extracted in the similar character string extraction step (S311), and outputs the extracted injury or illness name information (S312) (output step).

According to the second embodiment described above, it is possible to extract character strings representing names of illnesses and injuries from text data without adding more keywords than necessary to the master. In addition, character strings representing names of illnesses and injuries can be extracted from text data by converting them into a format included in the ICD-10 standard illness name master.

In addition, by further providing an element string extraction means 28, a similar string extraction means 29, and an output means 31, it is possible to more reliably extract strings representing the names of injuries and illnesses from text data.

Furthermore, by further providing a similarity calculation means 30, it is possible to narrow down and output information on injury or illness names.

In addition, if there is a similar string whose similarity is equal to or greater than a predetermined level among the similar strings whose similarity has been calculated previously, information on the name of the injury or illness corresponding to that similar string is output and processing ends, so the amount of processing required to output the information on the name of the injury or illness can be reduced, thereby increasing processing speed.

In the second embodiment, the injury/illness name master stored ICD codes, but it may also store other codes, such as the disease name management number of the ICD10-compatible standard disease name master. In the second embodiment, the injury/illness name master was created based on the ICD10-compatible standard disease name master, but it may also be created based on the injury/illness name master of the basic master, for example.

Although the embodiment has been described above, the present invention is not limited to the above embodiment, and can be modified as appropriate as shown in the following examples.

For example, in the above embodiment, when multiple element strings are extracted, the similarity is calculated for element strings that are closer to the center of a specified keyword before other element strings, but this is not limited to the above. For example, when multiple element strings are extracted, similar strings are extracted for each of the extracted element strings, and the similarity is calculated first for those with fewer extracted similar strings than the others. If there is a string with a similarity equal to or higher than a specified level among the similar strings whose similarity is calculated first, information on the medical procedure or the like corresponding to the similar string is output and the process is terminated. Also, the element string master may store element strings in correspondence with numbers or the like that indicate the priority when calculating the similarity.

In the above embodiment, the similarity calculation means calculates the similarity between the similar strings extracted by the similar string extraction means and a predetermined keyword, and the output means outputs information corresponding to the similar strings whose similarity calculated by the similarity calculation means is equal to or greater than a predetermined value, but this is not limited to the above. For example, the output means may be configured to output information corresponding to the similar strings extracted by the similar string extraction means without calculating the similarity if the number of similar strings extracted by the similar string extraction means is equal to or less than a predetermined value.

The elements described in the above-mentioned embodiments and variations may be implemented in any combination. For example, when the first embodiment and the second embodiment are implemented in combination, the prefix master of the first embodiment and the prefix master of the second embodiment may be a common master, and the element string master of the first embodiment and the element string master of the second embodiment may be a common master.

REFERENCE SIGNS LIST 1 Text data analysis system 11 Data acquisition means 12 Character string extraction means 13 First search means 14 First character string generation means 15 Second search means 16 Second character string generation means 17 Similar character string extraction means 18 First similarity calculation means 19 Output means 20 Second similarity calculation means 21 Element character string calculation means 22 Third similarity calculation means 23 First search means 24 First character string generation means 25 Second search means 26 Second character string generation means 27 Third search means 28 Element character string extraction means 29 Similar character string extraction means 30 Similarity calculation means 31 Output means

Claims (8)

A data acquisition means for acquiring text data;
a character string extraction means for extracting a group of character strings representing one item from the text data acquired by the data acquisition means;
a first search means for searching a medical practice/drug master storing character strings representing medical practices or drugs, using the character string extracted by the character string extraction means as a first keyword, to see if a character string matching the first keyword is found, and for outputting information on the medical practice or drug corresponding to the character string that is found as a result of the search;
a first character string generating means for generating a character string obtained by removing the prefix from the first keyword when a character string matching the first keyword is not found as a result of a search by the first search means, by referring to a prefix master storing a prefix which is a specific character string added to the beginning of a character string;
a second search means for searching the medical practice/drug master for a character string that matches the second keyword, using the character string generated by the first character string generation means as a second keyword, and for outputting information on the medical practice or drug corresponding to the character string that matches the second keyword when the search results in a hit;
a second character string generating means for generating a character string by removing at least parentheses and the character string enclosed by the parentheses from the second keyword when a character string matching the second keyword is not found as a result of the search by the second search means;
a similar character string extracting means for extracting a character string including the third keyword from the medical practice/drug master using the character string generated by the second character string generating means as a third keyword;
and an output unit that outputs information on at least one medical practice or drug corresponding to the character string extracted by the similar character string extraction unit. a first similarity calculation means for calculating a similarity between each of the character strings extracted by the similar character string extraction means and the second keyword when the similar character string extraction means extracts a plurality of character strings,
The text data analysis system according to claim 1, characterized in that the output means outputs information on medical procedures or medicines corresponding to character strings among the character strings extracted by the similar character string extraction means, the character strings having a similarity calculated by the first similarity calculation means that is equal to or greater than a predetermined value. The medical practice/drug master stores character strings representing one or more medical practices or drugs in association with index character strings that are character strings included in the character strings representing the one or more medical practices or drugs,
The text data analysis system further includes a second similarity calculation means for calculating a similarity between the third keyword and the index string when the similar string extraction means is unable to extract a string including the third keyword from the medical practice/drug master,
3. The text data analysis system according to claim 1, wherein when there is an index string whose similarity calculated by the second similarity calculation means is equal to or greater than a predetermined value, the similar string extraction means extracts a string corresponding to the index string from the medical procedure/drug master. and an element string extracting means for extracting at least one element string from the second keyword when there is no index string having a similarity equal to or greater than a predetermined value calculated by the second similarity calculating means, by referring to an element string master storing element strings that are specific character strings included in character strings representing medical procedures or medicines;
The text data analysis system according to claim 3, characterized in that the similar string extraction means extracts strings including the fourth keyword from the medical procedure/drug master, using the element string extracted by the element string extraction means as a fourth keyword. a third similarity calculation means for calculating a similarity between the second keyword and a character string extracted by the similar character string extraction means, the third similarity calculation means being configured to calculate a similarity between the second keyword and a character string extracted by the similar character string extraction means, the fourth keyword being a fourth keyword, before calculating a similarity between the second keyword and other character strings, when a plurality of element character strings are extracted by the element character string extraction means;
The text data analysis system according to claim 4, characterized in that, when a character string whose similarity is equal to or greater than a predetermined value is included among the character strings whose similarity has been previously calculated by the third similarity calculation means, the output means outputs information about a medical procedure or medicine corresponding to the character string. The text data analysis system according to any one of claims 1 to 5, characterized in that the second character string generation means generates a character string by further removing at least one of the following character strings (1) to (5) from the second keyword.
(1) A space at the beginning or end of a word; (2) A space in the middle of a word and the characters following that space; (3) A dot; (4) A comma; (5) A number and the characters immediately following that number that represent the unit. The computer includes:
A data acquisition step of acquiring text data;
a character string extraction step of extracting a group of character strings representing one item from the text data acquired in the data acquisition step;
a first search step of searching a medical practice/drug master storing character strings representing medical practices or drugs, using the character string extracted in the character string extraction step as a first keyword, to see if a character string matching the first keyword is found, and outputting information on the medical practice or drug corresponding to the hit character string when a character string matching the first keyword is found as a result of the search;
a first character string generating step of generating a character string obtained by removing the prefix from the first keyword by referring to a prefix master that stores a prefix, which is a specific character string added to the beginning of a character string, when a character string matching the first keyword is not found as a result of the search in the first search step;
a second search step of searching the medical practice/drug master for a character string that matches the second keyword using the character string generated in the first character string generation step as a second keyword, and outputting information on the medical practice or drug corresponding to the hit character string when the character string matches the second keyword as a result of the search;
a second character string generating step of generating a character string by removing at least parentheses and the character string enclosed by the parentheses from the second keyword when a character string matching the second keyword is not found as a result of the search in the second search step;
a similar string extraction step of extracting a string including the third keyword from the medical practice/drug master using the string generated in the second string generation step as a third keyword;
and an output step of outputting information on at least one medical practice or drug corresponding to the character string extracted in the similar character string extraction step. Computer,
A data acquisition means for acquiring text data;
a character string extraction means for extracting a group of character strings representing one item from the text data acquired by the data acquisition means;
a first search means for searching a medical practice/drug master storing character strings representing medical practices or drugs, using the character string extracted by the character string extraction means as a first keyword, to see if a character string matching the first keyword is found, and for outputting information on the medical practice or drug corresponding to the character string that is found as a result of the search;
a first character string generating means for generating a character string obtained by removing the prefix from the first keyword when a character string matching the first keyword is not found as a result of a search by the first search means, by referring to a prefix master storing a prefix which is a specific character string added to the beginning of a character string;
a second search means for searching the medical practice/drug master for a character string that matches the second keyword, using the character string generated by the first character string generation means as a second keyword, and for outputting information on the medical practice or drug corresponding to the character string that matches the second keyword when the search results in a hit;
a second character string generating means for generating a character string by removing at least parentheses and the character string enclosed by the parentheses from the second keyword when a character string matching the second keyword is not found as a result of the search by the second search means;
a similar character string extracting means for extracting a character string including the third keyword from the medical practice/drug master using the character string generated by the second character string generating means as a third keyword;
A computer program that functions as an output means for outputting information on at least one medical practice or drug corresponding to the character string extracted by the similar character string extraction means.

Images