JP6957388B2 - Business term discrimination device and business term discrimination method - Google Patents

Description

The present invention relates to a business term discriminating device and a business term discriminating method.

In large-scale system development, a wide variety of related parties such as customers and developers are involved, so the meanings of the terms used are different for each business, process, and affiliation of each related party. Is normal. For this reason, a great deal of cost is required for the development process and communication between related parties. One way to improve this situation is to create a glossary of technical terms that are commonly understood and used by customers and developers involved in system development work, that is, business terms. However, the current situation is that it has not been used well because it takes time to create it and requires continuous maintenance.

In this regard, with respect to various technical terms, a method of extracting terms specific to existing documents and design documents has been proposed. For example, Patent Document 1 extracts a morphological analysis means for recognizing a word unit and a part of an input document, and a compound word extraction means for extracting a compound word in the input document from an analysis result by the morphological analysis unit. Specialized word feature extraction means for extracting technical term feature word bases from the frequency of constituent words of compound words and the number of word combinations, and specialty for extracting compound words with extracted technical term feature word bases as proper nomenclature. It is described that candidates for technical terms are extracted using a term extraction means.

Further, Patent Document 2 includes a term data extraction means for extracting dependency information of a clause of an input document and a character string which is a nomenclature or a nomenclature phrase as term data by performing morphological analysis and syntactic analysis of the input document. Paraphrasing the dependency information and the term data Based on the concept data extraction means for extracting the conceptual data indicating the superordinate concept of the term data based on the specific paraphrase expression, and the term data based on the dependency information and the learning data. The term is used by concatenating the modification data with the modification data extraction means that determines whether or not the association modification clause is the definition of the term data and extracts the association modification clause determined to be the definition as the modification data. It is described that the candidate of the technical term and the commentary data of the technical term are generated by the commentary data generation means for generating the commentary data that defines the data.

Further, in Patent Document 3, among the compound words included in the document divided into the parts described by one author, unregistered compound words that are not registered in the glossary of terms specific to the document are extracted. Among the unregistered compound words extracted by the word extraction means, the technical terms whose character strings or meanings are similar to the technical terms registered in the glossary and whose appearance distribution in the part is not similar to the registered technical terms are selected. A document containing terms unique to a document by means for determining notational fluctuations that are determined to be notational fluctuations and terminology candidate extraction means for extracting unregistered compound words that are not determined to be notational fluctuations as candidates for technical terms. In the compound word extracted from, it is determined whether two similar words are technical terms having different meanings from each other, or whether one word is a technical term and the other word is a jargon of the technical term. It is stated that.

Japanese Unexamined Patent Publication No. 3-116374 Japanese Unexamined Patent Publication No. 2003-99429 Japanese Unexamined Patent Publication No. 2016-38596

In Patent Document 1, candidates for technical term feature groups are extracted from statistical information and semantic information of constituent words, and terms including technical term feature groups are extracted as technical terms. However, the business term does not necessarily include a specific technical term feature group, and the term including the specific technical term feature group does not always become a business term, so the extracted technical term may not be a business term. Not a few sexes. Further, in Patent Document 2, technical terms are specified by extracting dependency information and specific paraphrase expressions that paraphrase term data from a document. However, in documents in system development, business terms are often used without explaining their definitions. In such a case, it is difficult to correctly extract business terms in Patent Document 2. Further, in Patent Document 3, by paying attention to the creator of the document, it is determined whether the cause of the compound word not being registered in the glossary is due to the notational fluctuation, and the compound word is a business term. It cannot be determined whether or not. As described above, none of the patent documents has a structure capable of appropriately discriminating business terms.

The present invention has been made in view of such a current situation, and an object of the present invention is to correctly discriminate business terms existing in each document according to a mode of use thereof.

One of the present inventions for solving the above problems is a compound word extraction unit that extracts compound words from a plurality of documents, and the frequency of appearance of the extracted compound words in the document related to business is other than that. A business term extraction unit that extracts the extracted compound word as a business term when the frequency of occurrence of the extracted compound word in the document is higher, and a value of a parameter indicating the frequency of occurrence of the compound word that is the business term in the document. Is calculated for each attribute of the document, and if there is an attribute having the value of the parameter higher than the predetermined threshold, it is determined that the compound word which is the business term is a special business term , and the predetermined threshold is determined. When there is no attribute having a higher value of the parameter, the special business term discriminating unit that determines that the compound word that is the business term is a general-purpose business term, and the compound word that is the business term are special. business terms determination apparatus and an output unit for outputting the business terms or generic results of the information is determined either der Luke business terms, and that.
Further, another one of the present invention for solving the above problems is a compound word extraction process in which the business term discriminating device extracts a compound word from a plurality of documents, and the extracted compound in the document related to the business. When the frequency of appearance of words is higher than the frequency of appearance of the extracted compound words in other documents, the business term extraction process for extracting the extracted compound words as business terms and the compound words that are the business terms The value of the parameter representing the frequency of appearance in the document is calculated for each attribute of the document, and when there is an attribute having the value of the parameter higher than a predetermined threshold, the compound word which is the business term is a special business term. When there is no attribute having the value of the parameter higher than the predetermined threshold value, the compound word which is the business term is determined to be a general-purpose business term. It is a business term discrimination method for executing an output process for outputting information as a result of discriminating whether the compound term which is the business term is either a special business term or a general-purpose business term.

According to the present invention, it is possible to correctly discriminate business terms existing in each document according to the mode of use thereof.

It is a figure which shows an example of the structure of the business term discriminating apparatus which concerns on this embodiment. It is a figure which shows an example of the business document attribute information in this embodiment. It is a flowchart explaining the business term discrimination process in this embodiment. It is a figure which shows an example of the frequency information by attribute in this embodiment. It is a flowchart which shows an example of the general term determination process in this embodiment. It is a flowchart explaining an example of the business term determination processing in this embodiment. It is a figure which shows an example of the business term list in this embodiment.

Hereinafter, the business term discriminating device according to the present embodiment will be described with reference to the drawings. The business term discriminating device of the present embodiment is a business performed using a large number of business documents and various other documents (for example, a large-scale system development business involving a large number of parties such as customers and developers. Hereinafter, the business. In (), compound words (described later) are extracted from these documents, and the compound words are general or everyday terms (hereinafter referred to as general terms) without specializing in a specific business. Is it a general-purpose business term commonly used by the parties involved in the Business (hereinafter referred to as the general-purpose business term), or only for persons with specific attributes in the Business (for example, only those who belong to a specific department). It is a device that determines whether the term is used (hereinafter referred to as a special business term).

<Configuration of business term discrimination device>
FIG. 1 is a diagram showing an example of the configuration of the business term discriminating device according to the present embodiment. As shown in the figure, the business term discriminating device 1 includes a processor 10 such as a CPU (Central Processing Unit), a memory 11 of a RAM (Random Access Memory) and a ROM (Read Only Memory), and an HDD (Hard Disk Drive). ), SSD (Solid State Drive) and other storage devices 12, an input device 13 consisting of a keyboard, mouse, touch panel, etc., an output device 14 consisting of a monitor (display), etc., and a communication device 15 that communicates with other devices. These are connected by an internal bus or the like.

Further, the business term determination device 1 has the functions of the compound word extraction unit 101, the business term extraction unit 107, and the business term determination unit 105. Further, the business term discriminating device 1 stores a plurality of documents, and specifically, at least one or more business documents related to the business (for example, used specifically for a specific business or inside the business document 1). Business document information 111, which is information in which documents described based on terms that are used differently depending on the department, etc., and at least one or more other documents (for example, not sentences specific to a specific business) It stores general document information 112, which is information in which documents (documents mainly described based on general terms) are stored.

Further, the business term discriminating device 1 stores the respective information of the business document attribute information 113, the attribute-specific frequency information 114, and the business term list 115. Of these, the business document attribute information 113 is information on the attributes of each document in the business document information 111 and the general document information 112 (for example, the type of business targeted by the document, or the department in which the document is created or managed. Or department information. Hereinafter referred to as attribute information) is included. The frequency information 114 by attribute and the business term list 115 will be described later.

Here, the details of the business document attribute information 113 will be described.
(Business document attribute information 113)
FIG. 2 is a diagram showing an example of business document attribute information 113. The business document attribute information 113 stores information that specifies the type of business (target business) targeted by the document with the document name 1131 and the document name 1131 in which the information that identifies each document (business document or general document) is stored. This is information having each item of the target business 1132 to be performed and the creation department 1133 in which information for specifying the department (creation department) in which the document with the document name 1131 is created or managed is stored. In the example of the figure, the business related to the document of "Document 01" is "Business G1", and the document creation department of "Document 01" is "Department A1". In the present embodiment, the target business 1132 and the target business 1132 correspond to the attribute information, but information such as the creator of the document and the creation date and time may be used as other attribute information.

Next, the details of each function of the business term discriminating device 1 will be described.
First, the compound word extraction unit 101 shown in FIG. 1 extracts compound words from each document. Specifically, the compound word extraction unit 101 extracts compound words from business documents and general documents. In this embodiment, the compound word means a word composed of a plurality of roots (morphemes).

The business term extraction unit 107 extracts a compound word that is a business term by specifying the mode of appearance of the compound word extracted by the compound word extraction unit 101 in each document. For example, when the frequency of the extracted compound words in the business-related document (business document) is higher than the frequency of the extracted compound words in other documents (general documents), the business term extraction unit 107 may use the business term extraction unit 107. The above-extracted compound word is extracted as a business term. Further, in the business term extraction unit 107, the position and the number of appearances of the extracted compound words in the document (business document) related to the business are the appearances of the extracted compound words in the other documents (general documents). When it is determined that the position and the number of times are different from the above-mentioned compound words, the above-extracted compound words are extracted as business terms.

Specifically, the business term extraction unit 107 includes a document attribute-based frequency analysis unit 102, a term usage method analysis unit 103, and a general term determination unit 104. The document attribute frequency analysis unit 102 analyzes the frequency of occurrence of compound words based on the document attributes. The term usage analysis unit 103 analyzes whether or not the usage of compound words is different between each document. The general term determination unit 104 determines whether or not the compound word is a general term based on the frequency of appearance of the compound word in the business document and the general document and the usage of the compound word. This determination result is stored in the frequency information 114 for each attribute.

Next, the business term determination unit 105 determines whether the compound word, which is a business term, is either a general-purpose business term or a special business term. Specifically, the business term determination unit 105 has a special business term determination unit 108 and an output unit 109.

The special business term discriminating unit 108 specifies the distribution of compound words for each type of business as an attribute of a document or for each department in the business, so that the compound word as a business term is either a special business term or a general-purpose business term. Is determined. These determination results are stored in the business term list 115.

The output unit 109 outputs information as a result of determining whether the compound word which is a business term is either a special business term or a general-purpose business term.

The functions of the business term discriminating device 1 described above are the programs stored in the storage device 12 (or the memory 11) by the hardware of the business term discriminating device 1 or by the processor 10 of the business term discriminating device 1. It is realized by reading and executing. In addition, these programs are, for example, a secondary storage device, a non-volatile semiconductor memory, a hard disk drive, a storage device such as an SSD, or a non-temporary data storage that can be read by a computer such as an IC card, an SD card, or a DVD. Stored in the medium.

Next, the processing performed by the business term discriminating device 1 will be described.
<Processing of business term discriminator>
FIG. 3 shows, among the processes performed by the business term discriminating device 1, compound words are extracted from business documents and general documents, and it is determined whether these compound words are general terms, general-purpose business terms, or special business terms. It is a flowchart explaining the process (hereinafter, referred to as a business term discrimination process) which outputs the result. Note that this process is started, for example, when a predetermined input is input to the input device 13 by the user.

First, the compound word extraction unit 101 of the business term discriminating device 1 extracts compound words from each document stored in the business document information 111 and the general document information 112 (S1). Specifically, for example, the compound word extraction unit 101 extracts morphemes from the text of each document by performing morphological analysis, and based on the extracted morphemes, a word (compound word) having a predetermined sequence of part of speech is extracted. Identify. The morphological analysis is a method of extracting a morpheme, which is the smallest meaningful unit, from a sentence. The sequence of part of speech means a plurality of nouns that are adjacent to each other without interposing particles and the like. In addition, the extracted morpheme contains information on characters and part of speech. For example, when there is a sentence "applying morphological analysis to a business document", the compound word extraction unit 101 uses morphological analysis to perform "business", "document", "morpheme", "analysis", "apply", etc. Each morpheme is extracted, and among them, "business document" and "morpheme analysis", which are a sequence of a plurality of nomenclatures, are extracted as compound words.

Next, the document attribute-based frequency analysis unit 102 calculates the frequency of occurrence of each of the compound words extracted in S1 in each document (business document or general document) for the processing of S3 and S4 described later (). S2). Then, the document attribute-based frequency analysis unit 102 stores each compound word, the document in which the compound word exists, and the attribute information of the document in association with each other.

Specifically, for example, the document attribute-based frequency analysis unit 102 selects one of the compound words extracted in S1 and records the business document attribute information 113 in which the selected compound word is recorded in the document name 1131. All are specified, and the number of occurrences of the compound word existing in the document indicated by the document name 1131 of those records (hereinafter, referred to as a corresponding document at this stage. There are a case of a business document and a case of a general document) is calculated. At the same time, the attributes of the corresponding documents are specified by referring to the target business 1132 and the creation department 1133 of those records. These processes are repeated for all the compound words extracted in S1. Then, the document attribute-based frequency analysis unit 102 converts the calculated number of occurrences into the appearance frequency. In the calculation of the frequency of occurrence, for example, the average number of compound words appearing in one document may be used as the frequency of occurrence, or the number of compound words appearing in each document or the average number of compound words appearing per fixed number of characters may be used as the frequency of occurrence. It may be the frequency of appearance. The document attribute-based frequency analysis unit 102 stores this information in the attribute-based frequency information 114.

Here, the frequency information 114 for each attribute will be described.
(Frequency information by attribute 114)
FIG. 4 is a diagram showing an example of the frequency information 114 for each attribute. The attribute-specific frequency information 114 includes a target business 1141 which is information for specifying a target business as attribute information, a creation department 1142 which is information for specifying a creation department as attribute information, and a target business and creation department of the target business 1141. Information having each item of compound word 1143, which is a list of compound words associated with the creation department of 1142. In the figure, the target business and the creation department are used as the attribute information, and the compound words are "compound word W1", "compound word W2", "compound word W3", and "compound word W4". The frequency of appearance of the "compound word W1" corresponding to the target business of the "business G1" and the creation department of the "department A1" is 0.2. Further, when the compound word appears in the general document, the attribute information is not stored in the target business 1141 and the creation department 1142, but the information indicating that it is a general document is stored instead.

Next, as shown in FIG. 3, the general term determination unit 104 executes a process (general term determination process) of determining whether each compound word extracted in S1 is a general term based on the attribute-specific frequency information 114. (S3).

Here, the details of the general term determination process will be described.
(General term judgment processing)
FIG. 5 is a flowchart showing an example of the general term determination process. First, the general term determination unit 104 selects one of the compound words extracted in S1 (S31). Next, the frequency analysis unit 102 by document attribute compares the frequency of appearance in the general document and the frequency of appearance in the business document with respect to the compound word (hereinafter referred to as the selected word) selected in S31 (S32).

Specifically, for example, the document attribute-based frequency analysis unit 102 acquires all the values of the frequency of occurrence of the selected word among the values (frequency of occurrence) of the compound word 1143 of each record in the frequency information 114 by attribute. Further, the frequency analysis unit 102 for each document attribute identifies whether the record is a general document record or a business document record by acquiring the contents of the target business 1141 and the creation department 1142 of those records. When the appearance frequency of the selected word is stored in a plurality of records, the maximum value among them may be set as the appearance frequency, and the average value or the sum of the appearance frequencies for each attribute (for example, target business or creation department). The frequency of appearance for each attribute may be calculated individually by calculating. In this embodiment, the former method shall be adopted. In the example of FIG. 4, when the selected word is "compound word W1", the frequency of appearance in a general document is "1.5" and the frequency of appearance in a business document is "0.2".

Next, as shown in FIG. 5, the general term determination unit 104 determines whether or not the frequency of appearance of the selected word in the general document is higher than the frequency of appearance in the business document (S33). Specifically, for example, the general term determination unit 104 determines whether or not the frequency of appearance of selected words in a general document is higher than a predetermined value or more than the frequency of appearance in a business document. If the frequency of appearance of the selected word in the general document is higher than the frequency of appearance in the business document (S33: YES), the process proceeds to S37 to determine that the selected word is a general term, and the frequency of appearance of the selected word in the general document is increased. If it is not higher than the frequency of appearance in the business document (S33: NO), the process proceeds to S34.

In S34, the general term determination unit 104 determines whether or not the appearance frequency of the selected word in the general document and the appearance frequency of the selected word in the business document are close to each other. Specifically, for example, the general term determination unit 104 determines whether or not the difference in appearance frequency between the two is within a predetermined threshold value. If the frequency of appearance of the selected word in the general document and the frequency of appearance in the business document are similar (S34: YES), the process proceeds to S35, and the frequency of appearance of the selected word in the general document and the frequency of appearance in the business document are different. If they are not similar (S34: NO), the process proceeds to S38 to determine that the selected word is a business term.

In S35, the term usage analysis unit 103 determines whether or not the usage (distribution) of the selected words in the general document and the usage (distribution) of the selected words in the business document are similar. Specifically, for example, the term usage analysis unit 103 has the distribution of the selected words in the general document corresponding to the selected words and the distribution of the selected words in the business document corresponding to the selected words, which are associated in S2. Judge the similarity of.

In this similar judgment, for example, the selected words in each document are expressed in a distributed manner (for example, the selected words are represented by a low-dimensional real-valued vector (Word2vec, etc.)), and the difference in the distributed expressions in each document determines a general document. The similarity between the usage of the selected word in the business document and the usage of the selected word in the business document shall be evaluated. Further, for example, the similarity of the usage of the selected word may be determined based on the description position of the selected word in the figure or table in each document, or the usage of the selected word may be determined based on the similarity of the sentences including the selected word. You may judge the similarity of.

If the usage of the selected word in the general document is similar to the usage of the selected word in the business document (S36: YES), the process proceeds to S37 to determine that the selected word is a general term, and the general document If the usage of the selected word in is not similar to the usage of the selected word in the business document (S36: NO), the process proceeds to S38 to determine that the selected word is a business term.

In S37, the general term determination unit 104 determines that the selected word is a general term, and registers the selected word and the fact that the selected word is a general term in the business term list 115 described later. After that, the processing of S39 is performed.

In S38, the general term determination unit 104 determines that the selected word is (candidate) for the business term, and registers the selected word and the selected word as (candidate for) the business term in the business term list 115 described later. .. After that, the processing of S39 is performed.

In S39, the general term determination unit 104 determines whether or not there is a compound word that has not been selected so far, and if there is a compound word that has not been selected so far (S39: YES), the compound word. If one of the above is selected and the process of S31 is repeated thereafter, and there is no compound word that has not been selected so far (S39: NO), the general term determination process ends (S40).

Next, as shown in S4 of FIG. 3, when the general term determination process is completed, the business term determination unit 105 is based on the business term specified by the general term determination process and the attribute-specific frequency information 114, and the business term. Executes a process (business term determination process) for determining whether is a general-purpose business term or a special business term.

Here, the details of the business term determination process will be described.
(Business term judgment processing)
FIG. 6 is a flowchart illustrating an example of the business term determination process. First, the business term determination unit 105 selects one compound word determined to be a business term in the general term determination process (S41). Then, the business term determination unit 105 creates all combinations of each attribute, and for each of the created combinations, identifies all the business documents (business document group) associated with the attributes constituting the combination. In the specified business document group, the relative importance (importance to all business documents) of the business term selected in S41 (hereinafter referred to as the selected business term) is calculated (S42). The importance of this is, for example, TF-IDF (Term Frequency- Inverse Document).
Frequency) is used. In this case, TF-IDF is calculated by, for example, the product of the frequency of occurrence of selected business terms in a certain business document group and the logarithm of the value obtained by dividing the number of business document groups by the number of all business documents. As a result, for example, when the value of TF-IDF in a certain business document group is large, it can be considered that the selected business term is often used in the business document group and not so much in other business documents. ..

As an example of S42, in FIG. 4, the business term determination unit 105 has an attribute of "business G1", an attribute of "business G2", an attribute of "department A1", an attribute of "department A2", and "business G1". Corresponds to each attribute of "department A1", "business G1" and "department A2", "business G2" and "department A1", and "business G2" and "department A2". The selected business term TF-IDF is calculated for the set of business documents.

Next, as shown in FIG. 5, the business term determination unit 105 determines whether or not there is a particularly high importance of the respective importance calculated in S42 (S43). Specifically, for example, the business term determination unit 105 determines whether or not the highest value of each TF-IDF value calculated in S42 is higher than the threshold value specified in advance.

If there is something of high importance (S43: YES), the process proceeds to S44, and if there is no item of high importance (S43: NO), the process proceeds to S45.

In S44, the business term determination unit 105 determines the selected business term as a special business term, identifies the combination of attributes for which the high importance is calculated, and registers these results (discrimination results) in the business term list 115. do. After that, in S46, the business term determination unit 105 determines whether or not there is a business term that has not been selected so far, and if there is a business term that has not been selected so far (S46: YES), that business. One of the terms is selected and the process of S41 is repeated. On the other hand, if there is no business term that has not been selected so far (S46: NO), the business term determination process ends (S47).

In S45, the business term determination unit 105 determines the selected business term as a general-purpose business term, identifies the combination of attributes for which the high importance is calculated, and registers these results (discrimination results) in the business term list 115. do. After that, the processing of S46 is performed.

Here, the business term list 115 will be described.
(List of business terms 115)
FIG. 7 is a diagram showing an example of the business term list 115. The business term list 115 includes compound word 1151 which is information indicating a compound word, term type 1152 which is information indicating a compound word type (general term, general-purpose business term, or special business term) of compound word 1151, and compound. Each piece of category 1153 is information that identifies a combination of attributes determined to be of high importance in the compound word of word 1151. In the figure, for example, the type of compound word of "compound word W03" is a special business term ("business individual term") and its attribute is "business G1" (that is, for example, "compound word W03". The use of compound words is limited to the business of "business G1"). In addition, the type of compound word of "compound word W02" is a general-purpose business term ("business common term").
No information is stored in category 1153 because there is no combination of attributes determined to be of high importance. Further, the type of the compound word of the compound word "compound word W01" which is a compound word determined to be a general term in the general term determination process is a general term ("general term"), and therefore information is not stored in the category 1153.

The above description of the embodiment is for facilitating the understanding of the present invention, and does not limit the present invention. The present invention can be modified and improved without departing from the spirit of the present invention, and the present invention includes its equivalents.

As described above, the business term discriminating device 1 of the present embodiment is a business term by focusing on the compound word that tends to be a business term and specifying the appearance mode in the compound word document extracted from a plurality of documents. By extracting compound words and specifying the distribution of the compound words in each document for each attribute of the document, it is possible to determine whether the compound term is either a special business term or a general-purpose business term. Is the compound term of the included business term a business term that is used in relation to a specific attribute and is understood or used differently depending on the type of business or department, or is it a business term that is generally used in business? Can be determined. As described above, according to the business term discriminating device 1, the business terms existing in each document can be correctly discriminated according to the usage mode thereof.

The description herein reveals at least the following: That is, when the frequency of the extracted compound words in the document related to the business is higher than the frequency of the extracted compound words in the other documents, the business term extraction unit uses the extracted compound words as business terms. It may be extracted as.

In this way, business terms are extracted by comparing the frequency of appearance of compound words in business documents with the frequency of compound words in other general documents. Accurate extraction is possible.

Further, in the business term extraction unit of the present embodiment, the position and number of appearances of the extracted compound words in the document related to the business are the positions and times of appearance of the extracted compound words in other documents. If it is determined that they are different, the extracted compound words may be extracted as business terms.

In this way, business terms are extracted by comparing the appearance position and number of compound words in business documents with the appearance position and number of compound words in other general documents. Can be accurately extracted.

Further, the special business term discriminating unit of the present embodiment specifies the distribution for each type of business as an attribute of the document or for each department in the business, so that the compound word which is the business term is a special business term. Alternatively, it may be determined whether it is one of general business terms.

In this way, by specifying the distribution of compound words according to the attributes related to the business such as the type of business and the department, it is possible to determine when the business term is used in the business, and each business related person's business. It is possible to specify the business term according to.

Further, in the present embodiment, it may be provided with an output unit that outputs information as a result of determining whether the compound word which is the business term is either a special business term or a general-purpose business term.

In this way, by outputting the discrimination results of special business terms and general-purpose business terms, it is possible to provide information useful for the execution of business for the persons concerned in each business.

Further, in the present embodiment, the business term discriminating device is a business term by specifying a compound word extraction process for extracting a compound word from a plurality of documents and a mode of appearance of the extracted compound word in the document. By specifying the business term extraction process for extracting the compound word and the distribution of the compound word as the business term in the document for each attribute of the document, the compound word as the business term is a special business term or general purpose. It is also possible to execute a special business term discrimination process for discriminating which one of the business terms is used.

Further, in the business term discrimination method in the present embodiment, the business term discriminating device extracts the frequency of the extracted compound words in the document related to the business in the business term extraction process in the other documents. When the frequency of the compound word is higher than the frequency, the extracted compound word may be extracted as a business term.

Further, in the business term discrimination method in the present embodiment, the business term discriminating device determines the position and number of appearances of the extracted compound word in the business-related document in the business term extraction process. When it is determined that the position and the number of appearances of the extracted compound word are different from those in the above, the extracted compound word may be extracted as a business term.

Further, in the business term discrimination method in the present embodiment, the business term discriminating device specifies the business type as an attribute of the document or the distribution for each department in the business in the special business term discriminating process. , It may be determined whether the compound word which is the business term is either a special business term or a general-purpose business term.

Further, in the business term discrimination method in the present embodiment, the information as a result of the business term discriminating device determining whether the compound word which is the business term is either a special business term or a general-purpose business term is obtained. The output processing to be output may be further executed.

1 Business term discriminator, 101 compound word extraction unit, 107 business term extraction unit, 108 special business term discriminator, 109 output unit

Claims (6)

A compound word extractor that extracts compound words from multiple documents,
Business term extraction unit that extracts the extracted compound word as a business term when the frequency of appearance of the extracted compound word in the document related to the business is higher than the frequency of appearance of the extracted compound word in the other documents. When,
The value of the parameter representing the frequency of appearance of the compound word, which is the business term, in the document is calculated for each attribute of the document, and if there is an attribute having the value of the parameter higher than a predetermined threshold, the business term is used. there compound word is determined to be a special business terms, determined when there is no attribute having a value higher than the predetermined threshold value the parameter, compound word which is the operational term is generic business terms and special business terms determination unit that,
An output unit that outputs information as a result of determining whether the compound word that is the business term is a special business term or a general-purpose business term.
A business term discriminator equipped with. The business term extraction unit is similar to the distribution of the appearance position and the number of times of the extracted compound word in the document related to the business and the distribution of the appearance position and the number of times of the extracted compound word in the other documents. The business term discriminating device according to claim 1, wherein the extracted compound word is extracted as a business term when it is determined that the property is low. The special business term discriminating unit calculates a value of a parameter representing the type of business as an attribute of the document or the frequency of appearance for each department in the business, so that the compound word that is the business term is a special business term or The business term discriminating device according to claim 1, which discriminates whether the term is one of general-purpose business terms. The business term discriminator
Compound word extraction processing that extracts compound words from multiple documents,
Business term extraction process for extracting the extracted compound word as a business term when the frequency of appearance of the extracted compound word in the document related to the business is higher than the frequency of appearance of the extracted compound word in the other documents. When,
The value of the parameter representing the frequency of appearance of the compound word, which is the business term, in the document is calculated for each attribute of the document, and if there is an attribute having the value of the parameter higher than a predetermined threshold, the business term is used. there compound word is determined to be a special business terms, determined when there is no attribute having a value higher than the predetermined threshold value the parameter, compound word which is the operational term is generic business terms and special business terms determination process that,
Output processing that outputs information as a result of determining whether the compound word that is the business term is a special business term or a general-purpose business term, and
How to determine business terms to execute. The business term discriminating device
In the business term extraction process, the distribution of the appearance position and the number of times of the extracted compound word in the document related to the business is similar to the distribution of the appearance position and the number of times of the extracted compound word in the other documents. The business term determination method according to claim 4 , wherein the extracted compound word is extracted as a business term when it is determined that the property is low. The business term discriminating device
In the special business term discrimination process, the compound word that is the business term is a special business term or by calculating the value of the parameter indicating the frequency of appearance for each business type or department in the business as an attribute of the document. The business term determination method according to claim 4 , wherein the business term is determined to be one of general-purpose business terms.

Images