JP4039635B2 - Language information processing device - Google Patents

Description

The present invention relates to a language for performing a rule registration or knowledge registration process or a search process using a keyword based on a character string representing a language expression input by specifying a section in an information processing system that handles language expressions. The present invention relates to an information processing apparatus.
Information processing systems that handle linguistic expressions include machine translation systems, sentence selection systems, and sentence feature extraction systems. In such information processing systems, linguistic expressions such as idioms and phrases are expressed by linguistic information processing devices. By inputting a character string by specifying a section and analyzing the character string, rules and knowledge about the language are registered and utilized.

  The language information processing apparatus can also be used for searching a database using a specific language expression in a document as a keyword.

FIG. 7 shows a configuration example of a conventional language information processing apparatus.
In FIG. 7, the extraction processing unit 301 extracts the character string indicated by the section designation from the entire character string representing the document in accordance with an instruction from the user, and the decomposition processing unit 302 extracts the extracted character The sequence is divided into processing units and sent to the analysis processing unit 311.
The decomposition processing unit 302, for example, converts a character string representing a linguistic expression into a morpheme based on a morpheme dictionary 303 that stores morphemes composed of stems of independent words and ending information such as inflection endings. Has been broken down.

In response to the input of the morpheme collection obtained in this way, the analysis processing unit 311 analyzes the connection between morphemes, and using this result, the registration processing unit 312 uses the morpheme linked according to a predetermined rule. As a series of phrases, idioms and phrases are registered in the idiom dictionary 313.
Here, in the registration work to the idiom dictionary 313, an enormous number of idioms are often registered together.

  In such a case, the user designates corresponding sections in the document one after another, and accordingly, the character strings extracted by the extraction processing unit 301 are sequentially accumulated, After the designation is completed, the decomposition processing by the decomposition processing unit 302 and the registration processing to the idiom dictionary 313 are batch processed.

As described above, since the analysis processing unit 311 analyzes the relationship between morphemes, the input to the analysis processing unit 311 must be decomposed into a series of morphemes.
For this reason, conventionally, in the disassembly processing unit 302, when all the character strings in the specified section could not be decomposed into morphemes, the processing for the character string in the corresponding specified section is stopped at that time, and an error message or the like is issued. The user was notified that the specification of the character string could not be accepted.

In addition, in order to register a idiom or phrase as a verb or adjective stem, a constraint condition such as “the specified section must start with an independent word and end with an independent word” is necessary.
Conventionally, such a constraint condition has been examined by the analysis processing unit 311. If the sequence of morphemes received from the decomposition processing unit 302 does not satisfy the constraint condition, the character string in the designated section Processing is stopped immediately. Also in this case, the user is notified that the designation of the character string has not been accepted, as in the case where the character string cannot be decomposed.

Thus, the conventional linguistic information processing apparatus is based on the premise that the user is aware of the processing units and restrictions in the linguistic information processing apparatus and accurately specifies the character string section so as to match them. .
Therefore, in order to make full use of the language information processing apparatus, the user needs to have sufficient knowledge and experience regarding processing units in the language information processing apparatus such as morphemes.

However, general users often do not have such knowledge and experience, and the processing units such as morphemes described above are not the same as common sense language units. It is very difficult to specify exactly the interval that matches the constraints.
In addition, even if a user with sufficient knowledge inputs a specified section, if a large number of idioms and phrases are to be registered at once, an error will occur in the specification by the user. Since it becomes easy, many designation | designated areas will be excluded without being accepted.

In a conventional language information processing apparatus, in order to register a idiom or phrase corresponding to a specified section that has been rejected, the user has to correct the input of the specified section and repeat the registration work. However, this operation is troublesome for the user and increases the burden on the user.
An object of the present invention is to provide a language information processing apparatus that flexibly accepts input of a language expression including inconsistency with processing conditions.

FIG. 1 shows a principle block diagram of a current information processing apparatus according to the present invention.
The language information processing apparatus shown in FIG. 1 includes a character string input unit 111, a designated section input unit 112, a decomposition unit 113, a first determination unit 114, a first correction unit 115, and a second determination. It comprises means 121, second correction means 122, and extraction means 116.
The principle of the language information processing apparatus according to the present invention is as follows.

In a language information processing apparatus that receives a language expression to be processed and executes a predetermined process, the character string input unit 111 inputs a character string including the language expression to be processed. The designated section input means 112 inputs a designated section indicating the range of language expression included in the character string. The decomposition unit 113 decomposes the character string input by the character string input unit 111 into processing units which are units of language expression semantic analysis processing. The first determination unit 114 determines the validity of the specified section based on whether or not the boundary of the specified section matches any boundary of a series of processing units obtained by the decomposition unit 113. The first correcting means 115 is designated by moving the boundary position of the designated section so as to coincide with one of the boundaries of the processing unit when the first determining means 114 determines that the designated section is not valid. Correct the interval. The second determination unit 121 determines whether the first and last processing units are independent words for the order and type of processing units included in the designated section modified by the first modification unit 115 , or the designated section is an independent word. The validity of the specified section is determined based on whether or not the constraint condition that only one is included is satisfied. When the second determination unit 121 determines that the specified section is not valid when the second determination unit 121 determines that the first processing unit is a non-independent word, the specified section toward the head of the sentence until an independent word appears. According to the modification rule that the specified section is reduced toward the beginning of the sentence until the independent word appears, or the specification other than the first independent word is ignored if the processing unit at the end is a non-independent word Correct the specified section. The extracting unit 116 extracts a processing unit corresponding to the character string included in the designated section obtained by the correction from the series of processing units obtained by the decomposing unit 113.

The operation of the language information processing apparatus configured as described above is as follows.
In the invention of claim 1, since all of the character strings input by the character string input means 111 are subjected to the decomposition processing by the decomposition means 113, the character string of the designated section indicated by the designated section input means 112 is Information about the preceding and following character strings can be obtained.
Therefore, when the first determination unit 114 obtains a determination result that the specified section is not valid, the first correction unit 115 determines the specified section based on the information about the specified section and the character string before and after the specified section. Can be moved within the range including the character string before and after the specified section.

  At this time, the first correcting unit 115 moves the boundary of the designated section according to an appropriate rule for determining whether to include or exclude the processing unit divided by the boundary of the designated section in the designated range. For example, the inconsistency between the boundary of the designated section and the boundary of the processing unit can be eliminated, and a plurality of processing units included in the modified designated section can be used for subsequent processing as information on the language expression to be processed.

That is, a plurality of processing units included in the designated section corrected in this way are input to the second determination unit 121, and the determination process by the second determination unit 121, that is, the arrangement of these processing units is determined. It is subjected to a process for determining whether or not the constraint condition is satisfied.
When the second determination unit 121 determines that the specified section is not valid in light of the constraint condition, the second correction unit 122 includes the boundary of the specified section in the specified section based on the constraint condition described above. Move by processing unit.

At this time, the second correcting means 122 ignores designations other than the first independent word , for example, when the determination result by the second determining means indicates that there are a plurality of independent words in the designated section. If the boundary is moved , the inconsistency between the arrangement of the processing units included in the designated section and the structure in the language expression indicated by the constraint condition described above can be eliminated. As a result, the extraction unit 116 can extract information related to the language expression based on the legitimate designated section and can be used for processing such as registration processing.

  According to the language information processing apparatus according to the present invention, based on the information about the character string of the section specified by the user and the character string before and after the section, the inconsistency between the boundary of the specified section and the boundary of the morpheme and the constraint condition By detecting inconsistencies and moving the boundary position of the corresponding designated section, these inconsistencies can be eliminated. As a result, it is possible to flexibly accept specified sections including inconsistencies and perform language expression analysis and registration processing, and the work burden on the user can be greatly reduced.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
FIG. 2 shows a block diagram of an embodiment of the language information processing apparatus of the present invention.
In FIG. 2, the linguistic expression holding unit 201 accumulates linguistic expressions such as sentences including phrases and phrases to be registered as a unit, and the display data creation unit 202 accumulates in the linguistic expression holding unit 201. Display data for displaying the displayed language expression is created and sent to the display device 204 via the display memory 203.

Here, the language expression holding unit 201 described above may store, for example, one sentence divided by punctuation as one unit of language expression, and add a serial number to each sentence.
At this time, the display data creation unit 202 sequentially reads at least one sentence from the language expression holding unit 201 and displays the display data in which character codes are arranged according to the number of lines and the number of digits on the display screen by the display device 204. May be created and stored in the display memory 203.

In this case, the user can operate the mouse 205 and the keyboard 206 while looking at the sentences displayed on the display device 204 and designate a section of a character string representing an idiom or phrase included in these sentences. Good.
Information indicating the section designated in this way is first sent to the display data creation unit 202 via the input control unit 207 as information indicating the range of positions on the display screen, for example, by lines and digits. .

Based on this information, the display data creation unit 202 changes the attribute information corresponding to the character included in the specified section of the corresponding sentence, for example, the character included in the specified section is underlined, As a result, the user can confirm the designation of the section.
At this time, the input control unit 207 notifies the reading processing unit 211 that the section has been designated by the user, and in response to this, the reading processing unit 211 reads the corresponding sentence from the display memory 203. The character codes corresponding to all the character strings included in the character string are read together with the attribute information, and the character code string is sent to and held in the character string holding unit 212, and the attribute information is sent to the section information detecting unit 213.

The section information detection unit 213 detects attribute information indicating section specification from the received attribute information, creates section information indicating the range of the specified section based on the detection result, and stores section information. What is necessary is just to send to the part 214.
At this time, for example, the section information detection unit 213 receives information on the display format of the sentence such as the number of digits in one line from the display data creation unit 202, and based on this information, the position in the sentence of the specified section is determined from the beginning of the sentence. What is necessary is just to calculate as the number of characters. Further, when there are a plurality of designated sections in one sentence, each designated section may be numbered and held together with the number in the section information holding unit 214.

  Therefore, as shown in Fig. 3 (a), when the section indicated by the underline of the sentence "He is angry" is specified, this sentence " "He was angry." Is sent to the character information holding unit 212, and the section information detecting unit 213 detects section information indicating the range of the section indicated by the underline. As shown in Table 1, section information corresponding to section number 0 is stored.

ここで、表１においては、文の先頭文字から各文字に順に第０番から番号を付し、区間に含まれる番号の範囲を示すことにより、その区間の範囲を文字位置の範囲として示している。

Here, in Table 1, each character is numbered from the first character in order from the first character of the sentence, and the range of the number is indicated by indicating the range of the number included in the interval. Yes.

In this way, the content of the display memory 203 is read by the read processing unit 211 and separated into character code strings and attribute information, and the section information detection unit 213 extracts section information from the attribute information, so that idioms and the like can be obtained. It is possible to accept input of character information related to the included language expression itself and section information indicating a range of idioms to be registered.
That is, the input control unit 206 controls the display data creation unit 202 and the read processing unit 211 in accordance with the operation of the mouse 205 and the keyboard 206 to start the above-described operation, so that these units can perform character string input means. 111 and the function of the designated section input means 112 can be realized.

  Since the character information input in this way corresponds to the entire language expression of one unit stored in the language expression holding unit 201, the decomposition processing unit 221 corresponding to the disassembling unit 113 performs the morpheme dictionary 222. By disassembling this character information into morphemes in the same manner as in the past, it is possible to obtain information related to the language expression before and after the language expression of the designated section.

  Here, in the morpheme dictionary 222 described above, as shown in FIG. 4, “ha”, “wo”, “masashi”, together with morphemes that are stems of independent words such as “he”, “belly”, “stand”, etc. Morphemes that are non-independent words such as "" and "." Are stored together with information such as their attributes. However, in FIG. 4, as a part of information corresponding to each morpheme, a circle mark is attached when it is an independent word, and a cross mark is attached when it is a non-independent word.

For example, when the example sentence “He is angry” shown in FIG. 3A is decomposed into morphemes by the decomposition processing unit 221, each morpheme is obtained as shown in FIG. And sent to the mismatch detection unit 224 via the morpheme holding unit 223.
The inconsistency detection unit 224 corresponds to the first determination unit 114, and the decomposition result obtained by the decomposition processing unit 221 is collated with the corresponding section information, and the boundary of the specified section is the boundary of the morpheme. It is determined whether or not they match, and when a determination result indicating that they do not match is obtained, it is determined that a mismatch has been detected and the correction processing unit 225 is activated.

At this time, the inconsistency detection unit 224 determines whether the start position of the specified section matches the front boundary of the morpheme and whether the end position of the specified section matches the back boundary of the morpheme, respectively. What is necessary is just to judge.
For example, if inconsistency detection processing is performed for the example shown in FIG. 3, the start position of the specified section matches the boundary of the morpheme, but the end position of the specified section is a non-independent word “Dai” It can be seen that it does not coincide with the morpheme boundaries.

In this case, the inconsistency detection unit 224 activates the correction processing unit 225 by designating the boundary of the designated section in which the inconsistency is detected, and performs a process for correcting inconsistency between the boundary of the corresponding designated section and the boundary of the morpheme. Ask.
The correction processing unit 225 is configured to perform correction processing described later in accordance with the correction rules in the correction rule holding unit 226.

Here, the modified rule holding unit 226 holds, for example, the following two rules (1) and (2), and the morpheme including the boundary of the specified section is an independent word. It may be applied depending on whether or not.
Rule (1) When the corresponding morpheme is an independent word, the specified section is extended to the entire corresponding morpheme.

Rule (2) If the corresponding morpheme is a non-independent word, the morpheme is excluded from the designated section.
When the example of the specified section shown in Fig. 3 is modified, rule (2) is applied because the corresponding morpheme "Sat" is a non-independent word. Eliminated. In this case, the correction processing unit 225 corrects the section information corresponding to the corresponding section number in the section information holding unit 214 to the character position “2 to 5”, and as shown in FIG. May be moved to the rear side of the “stand” which is the morpheme immediately before the morpheme “sata”.

As described above, the correction processing unit 225 operates according to the correction rule in the correction rule holding unit 226, thereby realizing the function of the first correction unit 115 shown in FIG. Inconsistency can be resolved.
In FIG. 2, the transfer processing unit 227 operates as the extraction unit 116, and in response to a detection result indicating that there is no inconsistency or a notification indicating that the correction processing by the correction processing unit 225 described above has ended. According to the section information held in 214, the morphemes included in the designated section may be read from the morpheme holding unit 223 and sequentially sent to the analysis processing unit 311.

In this way, it is possible to flexibly accept section designations including inconsistencies, decompose the corresponding character string into morphemes, and use the decomposition results for analysis processing and registration processing.
In this case, since all character strings input to the analysis processing unit 311 are decomposed into morphemes, the analysis processing unit 311 and the registration processing unit 312 perform analysis processing and registration processing similar to those in the past, and specify The idiom or phrase expressed by the character string of the section may be registered in the idiom dictionary 313.

As described above, it is possible to automatically correct the boundary of the designated section by the user, so that the user needs to be aware of the processing unit in the language information processing apparatus when registering the language expression. It is possible to receive the section of the character string intuitively determined by the user and reliably input the corresponding language expression.
Therefore, it is possible to realize a language information processing apparatus that is easy to use even for a user with less specialized knowledge by reducing the burden on the user by eliminating the trouble of repeatedly inputting the same language expression.

The unit for accumulating language expressions in the language expression holding unit 201 is not limited to a so-called “sentence” that is grammatically complete, but may be a part of a sentence including idioms to be registered. However, one unit of language expression to be accumulated must be decomposable into morphemes.
Further, as a rule for correcting the boundary of the designated section, an example such as the following rule (3) is also conceivable.

Rule (3) If there is a boundary of a specified section in the middle of a character string that could not be decomposed into morphemes, the specified section is extended to the entire character string.
This rule (3) is to capture a user's intention by capturing a character string that could not be decomposed into morphemes as a proper noun and including the entire character string in a specified section.
Thereby, even when a proper noun that is not stored in the morpheme dictionary 222 provided in the language expression input device is included, incomplete section designation can be flexibly accepted.

In addition, as in the above-described embodiment, when the input processing, analysis, and registration processing of language expressions are performed interactively, the correction result by the correction processing unit 225 is displayed on the display device 204 via the display data creation unit 202. By doing so, it is also possible to let the user acquire specialized knowledge empirically.
On the other hand, there is a case where a user inputs a large number of language expressions at once, and the analysis and registration processes related to these language expressions are processed in batches.

FIG. 5 is a block diagram showing another embodiment of the language information processing apparatus according to the present invention.
5, the language information processing apparatus includes a sentence information holding unit 215 and a read processing unit 228 instead of the character information holding unit 212 shown in FIG.
In this case, when a section is designated by the user, sentence information indicating the storage location in the language expression holding unit 201 of the sentence for which the section is specified is held in the sentence information holding unit 215 for analysis and registration. When performing the processing, the read processing unit 228 may read the corresponding sentence from the language expression holding unit 201 based on the sentence information and send all the character strings to the decomposition processing unit 221.

For example, when a sentence number is given to each linguistic expression including idioms to be registered and stored in the linguistic expression holding unit 201 corresponding to the sentence number, the sentence information holding unit 215 displays The corresponding sentence number may be received from the data creation unit 202 and the sentence number may be held as the sentence information described above.
If the reading processing unit 228 searches the linguistic expression holding unit 201 based on the sentence number, all the character strings constituting the corresponding sentence can be read, and the character string before and after the character string in the specified section can be read. The column can be subjected to a morpheme decomposition process by the decomposition processing unit 221.

Therefore, using the information on the character string included in the specified section and the information on the character strings before and after it, the consistency between the boundary of the specified section and the boundary of the morpheme is determined, and the detected inconsistency is identified as the boundary of the specified section. Can be eliminated, and the registration process can be performed by flexibly receiving the designation of the section including the inconsistency.
Further, in this case, since the decomposition process and the correction process can be performed together with the analysis process and the registration process, the processing capability of the processor of the information processing apparatus can be effectively utilized.

Furthermore, it is possible to check the consistency of constraints required when registering idioms and phrases as verbs and adjective stems, and to modify the boundaries of the designated section according to the check results.
FIG. 6 is a block diagram showing still another embodiment of the language information processing apparatus according to the present invention.
In FIG. 6, the language information processing apparatus adds a condition holding unit 231, a condition check unit 232, a correction processing unit 233, and a correction rule holding unit 234 to the language information processing apparatus shown in FIG. And the morpheme boundary are detected and resolved, and the processing result is sent to the analysis processing unit 311 via the transfer processing unit 227.

  In FIG. 6, the condition holding unit 231 holds a constraint condition such as “the first and last morphemes are independent words” with respect to the order and types of morphemes included in the specified section, and the condition check unit 232 It is only necessary to determine whether or not the received series of morphemes satisfies this constraint condition. That is, the function of the second determination unit 121 is fulfilled by the condition holding unit 231 and the condition check unit 232.

For example, when the correction result shown in FIG. 3 (c) is input, since both the first morpheme “belly” and the last morpheme “stand” are independent words, the condition check unit 232 determines the specified section. Determines that the constraint condition is satisfied, and sends these morphemes to the analysis processing unit 311.
On the other hand, as shown in Fig. 3 (d), if the character string "I got angry" is set as the specified section, the start position and end position of the specified section match the morpheme boundary. The check unit 232 receives four morphemes from the first morpheme “belly” to the last morpheme “sata”.

In this case, since the last morpheme is not an independent word, the condition check unit 232 determines that the constraint condition is not satisfied, and requests the correction processing unit 233 to perform the correction process for the specified section.
Here, the correction rule holding unit 234 holds, for example, the following two rules (4) and (5), which are used for correction processing by the correction processing unit 233.
Rule (4) If the first morpheme is a non-independent word, the designated section is expanded toward the beginning of the sentence until the independent word appears.

Rule (5) If the last morpheme is a non-independent word, the specified section is reduced toward the beginning of the sentence until the independent word appears.
For example, in the case of the example shown in FIG. 3D, the correction processing unit 233 corrects the end position of the specified section by applying the rule (5), and the specified section as shown in FIG. By deleting the morpheme “sata” from the end and setting the end position of the specified section behind the “stand”, it is possible to obtain a series of morphemes that satisfy the constraint conditions described above.

As described above, the correction processing unit 233 performs the correction process according to the correction rule in the correction rule holding unit 234, thereby realizing the function of the second correction unit 115 shown in FIG.
As a result, it is possible to flexibly accept specified sections that include inconsistencies with the constraint conditions, and to proceed with analysis and registration processing. Therefore, it is possible to reliably register the input language expression. This saves you the trouble of repeatedly entering language expressions.

In addition, since it eliminates the need for the user to be aware of the constraints, it is possible to greatly reduce the workload of the user and provide a language information processing device that is easy to use even for users with less specialized knowledge. it can.
Here, if the linguistic expression to be registered satisfies the above-mentioned restriction condition “the morphemes at the beginning and end are independent words”, the linguistic expression is used as it is, or a prefix is added. And can make effective use of the corresponding language expression. In particular, when it is desired to register a linguistic expression as a verb or an adjective, it is desirable that the above-described constraints are satisfied.

Therefore, the check and correction functions for the constraint conditions described above are particularly effective when registering linguistic expressions utilized by endings such as verbs and adjectives.
In the case of keyword search or the like, for example, a constraint condition that “the specified section includes only one independent word” can be considered.
In this case, it is only necessary to store “ignore designations other than the first independent word” as the rule (6) in the correction rule holding unit 234 and send only the first independent word as a keyword to the search processing unit. .

  According to the language information processing apparatus according to the present invention, language expression registration processing, such as a machine translation system, a sentence search system, a sentence feature extraction system, and a database search using a specific language expression in a document as a keyword. In a system that is expected to specify the language expression to be registered or searched in accordance with the rules required for each processing in the search processing of keywords and keywords, a specified section including inconsistencies with such rules is used. It is possible to flexibly accept and analyze and register language expressions.

  Therefore, the user's workload is greatly reduced in machine translation systems, sentence review systems, sentence feature extraction systems, and information processing systems that handle linguistic expressions such as searching a database using a specific linguistic expression in a document as a keyword. Therefore, it is extremely useful in the field of information processing systems that handle such language expressions.

It is a principle block diagram of the language information processing apparatus concerning this invention. It is an Example block diagram of the language information processing apparatus concerning this invention. It is a figure explaining the correction operation | movement of a designated area. It is explanatory drawing of a morpheme dictionary. It is another Example block diagram of the language information processing apparatus concerning this invention. It is another Example block diagram of the language information processing apparatus concerning this invention. It is a figure which shows the structural example of the conventional language information processing apparatus.

Explanation of symbols

111 Character string input means 112 Designated section input means 113 Decomposition means 114 First determination means 115 First correction means 116 Extraction means 121 Second determination means 122 Second correction means 201 Language expression holding section 202 Display data creation section 203 Display Memory 204 Display Device 205 Mouse 206 Keyboard 207 Input Control Unit 211 Reading Processing Unit 212 Character String Holding Unit 213 Section Information Detection Unit 214 Section Information Holding Unit 215 Sentence Information Holding Units 221 and 302 Decomposition Processing Units 222 and 303 Morphological Dictionary 223 Morphological holding unit 224 Inconsistency detecting unit 225, 233 Correction processing unit 226, 234 Correction rule holding unit 227 Transfer processing unit 228 Reading processing unit 231 Condition holding unit 232 Condition checking unit 301 Extraction processing unit 311 Analysis processing unit 312 Registration processing unit 313 Idioms dictionary

Claims (1)

In a language information processing apparatus that receives an input of a language expression to be processed and executes a predetermined process,
A character string input means for inputting a character string including a language expression to be processed;
A designated section input means for inputting a designated section indicating a range of language expressions included in the character string;
Decomposing means for decomposing the character string input by the character string input means into processing units which are units of semantic analysis processing of language expression;
First determination means for determining the validity of the specified section based on whether or not the boundary of the specified section matches any boundary of a series of processing units obtained by the decomposing means;
When the first determination means determines that the specified section is not valid, the specified section is corrected by moving the boundary position of the specified section so as to coincide with any boundary of the processing unit. 1 correction means;
Restriction that the first and last processing units are independent words or that the specified section contains only one independent word for the order and type of processing units included in the specified area modified by the first modification means A second determination means for determining the validity of the specified section based on whether or not a condition is satisfied;
When the second determination means determines that the specified section is not valid, if the first processing unit is a non-independent word, the specified section is expanded toward the beginning of the sentence until the independent word appears, or at the end When the processing unit is a non-independent word, the specified section is corrected according to a correction rule that reduces the specified section toward the beginning of the sentence until the independent word appears or ignores specifications other than the first independent word . Corrective means,
An linguistic information processing apparatus comprising: extraction means for extracting a processing unit corresponding to a character string included in a specified section obtained by correction from a series of processing units obtained by the decomposing means.

Images