JP2001125911A - Device and method for extracting individual name and recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide individual name extraction device and method and a recording medium. SOLUTION: In the individual name extraction device 151, an input receiving part 15 receives a morpheme string as an input, a 1st individual name storage part 153 previously stores a morpheme (string) expressing an individual name, a 2nd individual name storage part 154 temporarily stores the morpheme (string) expressing the individual name, and a rule storage part 155 previously stores a rule for extracting the morpheme (string) expressing the individual name from the morpheme string. A certification part 156 certifies the morpheme (string) stored in the 1st or 2nd individual name storage part 153 or 154 from the received morpheme string certifies the morpheme (string) expressing the individual name on the basis of the previously stored role, an updating part 157 updates the contents of the 2nd individual name storage part 154, temporarily stores the certified morpheme (string) expressing the individual name and an output part 158 outputs the morpheme (string) expressing the certified individual name.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an extraction device, an extraction method, and a recording medium suitable for extracting a morpheme representing an individual name or a sequence thereof from an input morpheme sequence.

[0002]

2. Description of the Related Art Conventionally, a search system for obtaining necessary information from documents has been used in computers, word processors, database systems, and the like. In particular, with the development of Internet technology in recent years, an infrastructure for obtaining necessary information from various information sources is being prepared.

[0003] Further, there has been proposed a system that allows a computer or the like to understand the contents of a sentence, summarize the contents, or make a selection as appropriate.

In such a system, for example, when a Japanese document is to be processed, the following method has been used.

[0005] First, when a character string identical to a word to be processed appears in a sentence, this is included in a search result or the content is summarized based on the appearance frequency of the character string having the same spelling. This method performs search and summarization only on the character surface, and is often used in Internet search engines and the like.

[0006] Second, first, a morphological analysis is performed to divide a Japanese sentence into words (also referred to as "morphemes"), and a sequence of information pairs of spellings of morphemes and their parts of speech is output and output. This is a method of performing processing based on a sequence of information pairs.

[0007]

However, in a method of processing a document only by using a character surface, for example, when searching,
Documents containing a character string with the same spelling that was accidentally included are also included in the search result, and there is a possibility that a problem that unnecessary information is buried in unnecessary information may occur.

On the other hand, when an individual name such as a person name, a region name, a country name, a company name, or a date and time is used as a search target, it is not possible to distinguish between a person name and a region name only with information output by morphological analysis. However, the same problem as described above may occur. In addition, information available from the WWW (World Wide Web) or the like has many unknown words such as new product names, and therefore, it is necessary to deal with these.

Therefore, there is a need for a method of identifying an individual name that can be used for information retrieval, summarizing concepts, and the like from morphemes obtained by morphological analysis, and extracting and outputting the individual name.

An object of the present invention is to extract a morpheme representing an individual name or a sequence of the morpheme from an input morpheme sequence.

[0011]

To achieve the above object, an apparatus for extracting an individual name according to the present invention comprises: an input receiving unit;
, A second individual name storage unit, a rule storage unit, a certification unit, an update unit, and an output unit.

Here, the input receiving unit receives a morpheme sequence as an input.

The first individual name storage unit stores in advance a morpheme representing the individual name or a sequence thereof in association with its attribute.

The second individual name storage unit temporarily stores a morpheme representing the individual name or a sequence thereof in association with the attribute.

The rule storage unit stores in advance rules for identifying morphemes representing individual names from the morpheme strings or the strings thereof in association with their attributes.

The recognition unit recognizes the morpheme or its row stored in the first or second individual name storage unit from the received morpheme string, and stores the previously stored morpheme string from the received morpheme string. The morpheme representing the individual name or its sequence is identified according to the rule.

The updating unit updates the second individual name storage unit and temporarily stores the morpheme or the column representing the recognized individual name in association with the attribute.

The output unit outputs a morpheme representing the recognized individual name or a sequence thereof as the extracted individual name.

In the present extraction device, when a morpheme or a string thereof representing an unknown word (hereinafter appropriately referred to as a “morpheme (string)”) is recognized as an individual name by a rule, the morpheme (string) representing the unknown word is assigned to an individual. The name is registered in the individual name storage unit. Therefore, it can be recognized as an individual name without using any rules. Furthermore, a morpheme (a part of the unknown word) included in the unknown word can be recognized as an individual name without using a rule.

[0020] Also, there is a case in which common nouns such as "information", "processing", and "society" continue to represent an individual name of "information processing society" although the morphological analysis does not result in unknown words. , "Information Processing Society of Japan" is recognized as an individual name and registered in the second individual name storage unit. In this case, it is not necessary to register the “Information Processing Society” itself in the first individual name storage unit. This is because recognition as an individual name is possible by the rules stored in the rule storage unit.

Thus, the individual name can be appropriately extracted.

[0022] The extraction device may be configured to further include a recertification control unit.

Here, when the updating unit updates the second individual name storage unit, the re-authorization control unit causes the authentication unit to again identify the morpheme representing the individual name or its sequence.

In the present extraction apparatus, as long as an unknown word is newly registered as an individual name, a repetition rule is applied to the same morpheme string input. Thus, for example, the rule is applied near the beginning of the document based on the unknown word recognized as the individual name near the end of the document, and the unknown word can be recognized as the individual name.

Further, the extracting device may be configured to further include an erasing unit.

Here, when the specific condition is satisfied, the erasing unit erases the morpheme representing the temporarily stored individual name or its row from the second individual name storage unit.

In the present extracting apparatus, for example, the registered unknown words are deleted each time the type of document changes. Therefore,
Individual names can be extracted according to the context of the document.

Further, in the above-mentioned extraction device, the certifying section may determine, according to the previously stored rule, a first morpheme or a column thereof representing an individual name and a second morpheme or a column thereof representing an individual name. The interposed morpheme or its row may be configured to be recognized as a morpheme or a row representing an individual name.

In the present extracting apparatus, for example, an unknown kanji character string sandwiched between "Tokyo" and "Osaka" can be extracted as an individual name having a place name attribute.

Further, in the present extracting apparatus, the certifying unit certifies a predetermined morpheme or a morpheme connected to the sequence or a sequence thereof as a morpheme representing an individual name or a sequence thereof according to the previously stored rule.

In the present extracting apparatus, for example, when a katakana string follows "(share)", the katakana string can be extracted as an individual name having a company name attribute.

In order to achieve the above object, a method for extracting an individual name according to the present invention comprises: a first individual name storage unit for storing in advance a morpheme or a sequence thereof representing an individual name in association with its attribute; A second individual name storage unit for temporarily storing a morpheme or its column representing
A method for extracting an individual name from a morpheme string in a computer including a morpheme representing the individual name from the morpheme string or a rule storage unit for preliminarily storing a rule for certifying the string in association with its attribute. It is configured to include a receiving step, a certification step, an updating step, and an output step.

Here, in the input receiving step, a morphological sequence is received as an input.

In the certifying step, the morpheme or the sequence thereof stored in the first or second individual name storage unit is certified from the received morpheme sequence, and the previously stored morpheme sequence is stored in advance. The morpheme representing the individual name or its sequence is identified according to the rule.

In the updating step, the second individual name storage unit is updated, and the morpheme or the column representing the recognized individual name is temporarily stored in association with its attribute.

In the output step, a morpheme representing the recognized individual name or a sequence thereof is output as the extracted individual name.

Further, the extraction method can be configured to further include a recertification control step.

Here, in the recertification control step, when the second individual name storage unit is updated in the updating step, the morpheme representing the individual name or its column is recertified in the authorization step.

Further, the extraction method can be configured to further include an erasing step.

Here, in the erasing step, when the specific condition is satisfied, the morpheme representing the individual name or the column thereof which is temporarily stored is erased from the second individual name storage unit.

Further, in the above extraction method, in the certification step, the rule stored in advance may be:
A morpheme or a sequence sandwiched between a first morpheme or a sequence representing an individual name and a second morpheme or a sequence representing the individual name is identified as a morpheme or a sequence representing the individual name. Can be configured.

Further, in the above extraction method, in the certification step, the rule stored in advance may be:
A morpheme connected to a predetermined morpheme or a sequence thereof or a sequence thereof may be configured as a morpheme representing an individual name or a sequence thereof.

In order to achieve the above object, a computer-readable recording medium of the present invention comprises: a computer, an input receiving unit, a first individual name storage unit, a second individual name storage unit, a rule storage unit, a certification unit. , An update unit, and a program that functions as an output unit.

Here, the input receiving unit receives a morpheme sequence as an input.

The first individual name storage unit stores in advance a morpheme representing an individual name or a sequence thereof in association with its attribute.

The second individual name storage unit temporarily stores a morpheme representing the individual name or a sequence thereof in association with its attribute.

The rule storage unit stores in advance rules for identifying morphemes representing individual names from the morpheme strings or the strings in association with their attributes.

The certifying unit certifies the morpheme or the sequence thereof stored in the first or second individual name storage unit from the received morpheme sequence, and recognizes the previously stored morpheme sequence from the received morpheme sequence. Morpheme representing the individual name or its sequence according to the rules

The updating unit updates the second individual name storage unit and temporarily stores the morpheme or the column representing the recognized individual name in association with the attribute.

The output unit outputs a morpheme representing the recognized individual name or a sequence thereof as the extracted individual name.

Further, the program recorded on the computer-readable recording medium may be configured to further cause the computer to function as a recertification control unit.

Here, the recertification control unit, when the second individual name storage unit is updated by the updating unit, causes the certification unit to recertify the morpheme representing the individual name or its row.

Further, the program recorded on the computer-readable recording medium may be configured to cause the computer to further function as an erasing unit.

Here, when the specific condition is satisfied, the erasing unit erases the morpheme representing the individual name or the column thereof which is temporarily stored from the second individual name storage unit.

In the case where the computer functions as the certifying section by the program recorded on the computer-readable recording medium, the certifying section uses a first rule representing an individual name according to the previously stored rule.
The morpheme or its column sandwiched between the morpheme or its column and the second morpheme or its column representing the individual name can be configured as a morpheme or its column representing the individual name.

In the case where the computer functions as the certifying unit according to the program recorded on the computer-readable recording medium, the certifying unit determines a predetermined morpheme or a string of morphemes based on the previously stored rule. The morpheme or its column that is linked to is identified as a morpheme or its column that represents an individual name.

[0057]

Embodiments of the present invention will be described below with reference to the accompanying drawings.

(Summary Configuration) FIG. 1 shows a typical information processing device (including a dedicated device such as a general-purpose computer, various terminals, a portable terminal, a mobile phone, and a game device) which functions as an individual name extracting device of the present invention. It is a schematic diagram which shows schematic structure of (). Hereinafter, description will be made with reference to FIG.

The information processing apparatus 101 has a CPU (Central
It is controlled by a processing unit (central processing unit) 102. When the information processing apparatus 101 is powered on,
The CPU 102 executes an IPL (Initial Program) stored in a ROM (Read Only Memory) 103.
Loader (initial program loader).

The IPL includes the hard disk 104, the FD
(Floppy Disk) drive 110
FD, CD-ROM (Compact Disk ROM)
This is a program for reading and executing an OS (Operating System) program stored in a recording medium such as a CD-ROM mounted on the drive 111.

After the OS is started, the CPU 102 stores the data in the hard disk or the like in accordance with a user's instruction input with the keyboard 105 or the mouse 106 or in accordance with the contents of a setting file previously described in the hard disk or the like. Execute the application program.

In a small-scale information processing apparatus such as a portable terminal, an embodiment in which the IPL itself functions as an OS or an application program can be adopted.

The CPU 102 uses a RAM (Random Access Memory) 107 as a temporary work storage area when executing the program. In addition, a register or a cache (not shown) provided in the CPU 102 is used as a temporary work storage area.

In order to report the result to the user or to show the progress of the program, the CPU 102
Can display the information on a display device 108 such as a liquid crystal display or a CRT (Cathode Ray Tube). In the instruction operation using the mouse 106, the cursor displayed on the screen is moved by moving the mouse 106, and the menu item indicated by the cursor can be selected by clicking the mouse 106.

The information processing apparatus 101 is an NIC (Network
Interface 10 such as Interface Card) and modem
9 can communicate with a computer communication network such as the Internet. The document data received via the interface 109 may be processed, the processing result may be transmitted via the interface 109,
It can execute a program received via the interface 109, and the like.

FIG. 2 is a schematic diagram showing a schematic configuration of each function when the information processing apparatus 101 functions as the individual name extracting apparatus of the present invention.

The individual name extracting device 151 includes the input receiving unit 1
52, a first individual name storage unit 153, a second individual name storage unit 154, a rule storage unit 155, a certification unit 156, and an update unit 1
57, and an output unit 158. Further, a configuration may be provided in which a recertification control unit and an erasing unit are provided as elements not shown.

The input receiving unit 152 receives a morpheme string as an input. Here, the document data to be processed and the data in which the morpheme sequence resulting from the completion of the morphological analysis in advance can be obtained via the interface 109 as described above, and the hard disk 104,
It can be obtained from a file stored in an FD or a CD-ROM. Further, the document data input by the user operating the keyboard 105 or the mouse 106 may be subjected to morphological analysis, and then may be processed.
It is stored in the AM 107.

Therefore, the interface 109, the hard disk 104, the FD drive 110, the CD-RO
The M drive 111 and the RAM 107 function as an input receiving unit 152 that receives a morpheme string as input.

The first individual name storage unit 153 and the second individual name storage unit 154 store morphemes (columns) representing individual names in association with their attributes. The former is stored in advance, and the latter is stored in advance. Memorize temporarily. Also, the rule storage unit 15
5 stores in advance a rule for identifying a morpheme (column) representing an individual name from the morpheme string in association with its attribute.

Therefore, the hard disk 104, the FD
The FD mounted on the drive 110, the CD mounted on the CD-ROM drive, the ROM 103, the RAM 107, and the other computer connected via the interface 109 correspond to these readable information storage devices. It functions as one individual name storage unit 153 and rule storage unit 155. The first individual name storage unit 153 and the rule storage unit 155 do not necessarily need to be writable.

On the other hand, the hard disk 104, the FD mounted on the FD drive 110, and the RAM 107
Those corresponding to these readable and writable information storage devices of another computer connected via the interface 109 function as the second individual name storage unit 154.

The certification section 156 is provided with the first individual name storage section 15.
3. Using the information stored in the second individual name storage unit 154 and the rule storage unit 155, a morpheme (column) representing the individual name is recognized from the morpheme sequence whose input has been received. In the present embodiment, the CPU 102 functions as the certification unit 156.

The updating unit 157 additionally stores a morpheme (column) representing the newly recognized individual name in the second individual name storage unit 154. In the present embodiment, the CPU 102 functions as the updating unit 157.

The output unit 158 outputs a morpheme (column) representing the recognized individual name as the extracted individual name. Note that a form in which the attribute associated with the morpheme (column) representing the individual name is also output can be adopted.

The display device 108, the hard disk 104,
FD mounted on FD drive 110, RAM 107
And a morpheme (column) representing an individual name extracted from those corresponding to these readable / writable information storage devices of another computer connected via the interface 109
Can be selected as the output destination. Therefore, they function as the output unit 158.

The recertification control unit (not shown) controls to repeat the certification process in the certification unit 156 as long as there is a morpheme (column) representing the newly certified individual name.
In this case, the CPU 102 functions as a recertification control unit.

The erasing unit (not shown) erases the morpheme (column) of the individual name stored in the second individual name storage unit 154 when a specific condition is satisfied. In this case, the CPU 10
2 functions as an erasing unit.

In addition, the hard disk 104, the FD mounted on the FD drive 110, the CD mounted on the CD-ROM drive, the ROM 103, the RAM 107, and the other computers connected via the interface 109 can read these data. The information storage device functions as a recording medium of the present invention by recording a program executed by the CPU 102.

As described above, the information processing apparatus 1 shown in FIG.
01 is a typical embodiment for functioning as the individual name extracting device of the present invention shown in FIG.

For example, a robot server that automatically obtains documents in the Internet, extracts individual names, and creates a document database of individual names, executes the individual name extraction method of the present invention, and executes the robot server. Can function as the individual name extracting device of the present invention. In this case, the display device 108, the FD drive 110, the keyboard 105, the mouse 106, etc. are unnecessary, while the interface 109 is essential.

In the portable terminal, a nonvolatile semiconductor memory (flash memory card) is used as the external storage device.
Alternatively, a battery-backed RAM can be used. In this case, the hard disk 104,
The FD drive 110 and the CD-ROM drive 111 are not required.

(Processing Procedure) The processing executed by the above information processing apparatus to realize the function of the individual name extracting apparatus of the present invention, that is, the outline of the individual name extracting method of the present invention will be described below. I do. FIG. 3 is a flowchart showing the flow of the processing.

First, the input receiving unit 152 receives an input morpheme sequence (step S201). As described above, the morpheme sequence may be received from a recording medium such as the hard disk 104, or may be received from an input device such as the keyboard 105. Further, an embodiment in which the data of the morpheme sequence is stored in a predetermined area in the RAM 107 and the step S201 is started in the form of a function call or a system call using the address of the area storing the data as a parameter is adopted. Can also.

Next, the recognition unit 156 recognizes morphemes (columns) stored in the first individual name storage unit 153 and the second individual name storage unit 154 from the input morpheme sequence (step S202). . Details of the individual name used in step S202 will be described later.

Further, the recognition unit 156 is stored in the rule storage unit 155 by using information such as a feature word, morphemes arranged in the vicinity (particularly adjacent), and recognized individual names from the input morpheme sequence. By applying the rule, the morpheme (column) of the unregistered individual name is recognized (step S203).
Details of the certification rules will be described later.

Next, among the morphemes (columns) representing the recognized individual names, there are morphemes (columns) not stored in the second individual name storage unit 154, that is, morphemes (columns) of the newly recognized individual names. Is determined (step S20).
4).

If there is a morpheme of the newly identified individual name (step S204; Yes), the updating unit 157 additionally stores and updates the second individual name storage unit 154 (step S206), and proceeds to step S202. Return. As a result, the rule can be applied again using the information of the newly recognized individual name. The control of the process returning to step S202 is performed by the recertification control unit.

On the other hand, if there is no morpheme (column) of the newly identified individual name (step S204; No), all the individual names that can be identified by the method of the present invention have been identified, The morpheme (column) representing the name is output as the individual name extracted from the output unit 158 (step S205), and the process ends. Details of the output will be described later.

(Type of Individual Name) In the individual name extraction device and the extraction method of the present invention, a morpheme (string) representing an individual name is extracted from a morpheme string output by executing a morphological analysis in advance. . In the following, an example of the type of morpheme adopted as the individual name will be described.

In this embodiment, the following four types of attributes are recognized in association with morphemes (columns) of individual names.・ Proper noun ・ time ・ quantity ・ index

The attribute “proper noun” is further classified into a product name, an organization name, a place name (area name, natural object name, facility name), and a person name.

The attribute “time” is further classified into economic time, calendar time (including day of the week), non-calendar time, and special affixes.

The attribute “quantity” is further classified into an amount, a ratio, and an ordinal.

The attribute “index” further includes a URL
(Universal Resource Locater), e-mail address, telephone number and facsimile number. These are character strings composed of alphanumeric characters and symbols, but can be considered according to the names of persons and organizations in terms of specifying individuals and companies.

As described above, the attributes of the morphemes (columns) of the individual names handled in the present invention are hierarchically structured. The attribute hierarchy can be changed and configured as appropriate.

In the present invention, morphemes (sequences) of the individual names to which these attributes are assigned are classified and certified from the morpheme strings to be processed, and the certification result is output as the extracted individual names.

A description will be given with a specific example. Hereinafter, the morphemes will be appropriately separated using the symbol “/”. "I,
I am the Managing Director of Yamada Shokai Tanaka. "Is" I / ha /, /
Co., Ltd. / Yamada / Shokai / Tanaka / Senior Managing Director / ] And can be analyzed into a morpheme sequence.

Some of these morphemes can be assigned the above attributes. For example, "Tanaka" has the attribute "Last Name" and "Senior Managing" has the "Title"
Can be assigned respectively. The attribute assigned to the morpheme itself is called "role feature".

On the other hand, attributes can also be assigned to morpheme strings such as "Yamada / Shokai", "(Co) / Yamada / Shokai", and the like. These can be assigned the attribute "organization name". Similarly, "Tanaka / Senior Managing Director" has "person name"
Attribute can be assigned.

"/ Yamada / Shokai / Tanaka / Senior Managing Director"
In this example, this is the maximum morpheme sequence that can be identified as an individual name, and an attribute of “person name” can be assigned to this. The attribute assigned to the maximum morpheme sequence that can be identified as an individual name is called a "category".

The first individual name storage unit 153 stores in advance morphemes (columns) selected as these individual names,
The attribute of the individual name is stored in association with the attribute.

In the above example, attributes are simply assigned.
For example, even if the proper noun "Yamada" is used, there are various suitable attributes such as a person name, an organization name (company name), and a region name. Once the phrase "Yamada Museum" is written in front of the sentence, simply write "Yamada" at the end of the sentence.
In this case, the name of the facility will be displayed.

Therefore, the first individual name storage unit 153 may store a plurality of types of attributes that can be taken by morphemes (columns) representing individual names in association with each other. For example,
For "Yamada", "last name", "organization name", "company name", "region name" and the like are stored.

The time, quantity, and index can be determined by a pattern of a character string representing a morpheme.
pression), and can be stored in the first individual name storage unit 153 or the rule storage unit 155.

FIG. 4 shows the morphemes (columns) representing the individual names stored in the first individual name storage unit 153 and their attributes and categories.

In the example shown in FIG. 4, “Pajero Mini Duke”, “Yamamoto Foods”, “Yamamoto Foods”, “Tanaka Housing Hokuriku”, “Iyakei”, “Yakezanji”, and “Heiseimon” are the morpheme strings that can be identified as individual names. It is remembered. For example, the attribute (category) of "Pajero Mini Duke" is "Product", and the attributes (role features) of "Pajero", "."
"Product name body""Symbol""Product name body""Seriesname"
It is.

As described above, the first individual name storage section 153 stores morphemes (columns) representing predetermined individual names. The unknown word recognized as the individual name by the recognition rule can be stored in the second individual name storage unit 154 in a similar format.

(Certification Rule) First Individual Name Storage Unit 153
The individual name can be identified from the morpheme sequence received as input using the morphemes (columns) stored in, but this may not be sufficient.

As described above, coined words, new words, and the like are often used in proper nouns such as product names, so that there are many unknown words that are not stored in the first individual name storage unit 153. This is particularly noticeable when available from the WWW. We want to be able to extract unregistered individual names even in such a case.

[0111] These are dealt with by the certification rules described below. The following information is used in these certification rules. -Attributes of the morphological analysis result. Part of speech, character type, character face, etc. -Attributes of morphemes (rows) of already recognized individual names. Categories, role features, etc. The morpheme (column) of the recognized individual name is stored in the second individual name storage unit 154.

When a plurality of attributes of an individual name are recognized by assigning weights to the recognition rules, the attributes of the individual names, and their connection relationships, a score is calculated from these weights, and which is the most likely one. By estimating, the accuracy of document search and summarization can be improved.

(Authentication Rule Based on Characteristic Word) First, an example of an authentication rule using a characteristic word or a co-occurrence word will be described. Prefixes and suffixes can be considered as characteristic words. For example, "manager" and "(stock)". When “Yamada / Ichiro / Manager” is input as the morpheme sequence, it is understood from the characteristic word “Manager” that the morpheme sequence “Yamada / Ichiro” represents a person's name.

Therefore, the morphemes (columns) of “Yamada”, “Ichiro”, and “Yamada / Ichiro” are updated and registered as personal names in the second individual name storage unit 154. As will be described later, since the first individual name storage unit 153 and the second individual name storage unit 154 are both used for the identification of the individual name, “Yamada / Ichiro / Manager” appears in the text. After that, even if "Yamada" appears alone, it can be estimated that this "Yamada" represents a person's name.

Similarly, when a morpheme sequence such as “(stock) / just” or “system / (stock)” is input,
“Just” or “System” is the first individual name storage unit 15
It is found that even the unregistered proper noun that is not stored in No. 3 indicates the company name from the characteristic word “(stock)”.

Similarly, when a morpheme string such as “12 / month / 28 / day” is input, “12” represents the month,
It can be seen that “28” is an individual name representing a day.

The characteristic word is “Yamada / Ichiro / Manager”
Like “(share) / just”, a category is included in a morpheme (column) representing an individual name, and a category can be assigned to the entire morpheme (column) including this.

On the other hand, a co-occurrence word has a function similar to that of a characteristic word, but is not included in a morpheme (string) representing an individual name. For example, when a morpheme sequence of “alto / wo / release” is input, “alto” is a morpheme (column) representing an individual name and a category “product name” is given from the co-occurrence word of “release” . “Release” is not included in the morpheme (column) representing the individual name.

Information on such characteristic words and co-occurrence words is stored in the rule storage unit 155. FIG. 5 shows the rule storage unit 1
It is a schematic diagram which shows the aspect of the characteristic word and co-occurrence word memorize | stored in 55.

What is actually stored in the rule storage unit 155 is that in the table shown in FIG.
There are a "table classification" column and a "registered word" column.

"Table classification" is an identifier for identifying characteristic words and co-occurrence words having similar characteristics.
“Registered words” are characteristic words and co-occurring words having the characteristics.

For example, “industry”, “industry”, and “construction” belonging to the classification “T15” are all “subsequent characteristic words”, and the attribute (category) of the morpheme (column) of the individual name including these characteristic words is "Product name".

On the other hand, “release” belonging to the classification “T11”
"Sales", "development" and "announcement" are all "afterwords"
The attribute (category) of the morpheme (column) arranged before these characteristic words is “product name”.

(Rules for Identifying Parallel Noun Phrases) There may be a rule in which unregistered morphemes (strings) are estimated and identified as individual names by the parallel noun phrases.

For example, "Tokyo /, / Ndjamena /, /
When a morphological sequence such as "Paris" is input and it is determined that "Tokyo" and "Paris" are individual names representing a region name, "Ndjamena" is not stored in the first individual name storage unit 153. Even an unregistered word can be inferred to be an individual name representing a region name.

Since “Ndjamena” is composed of only katakana spelling, even if this is an unregistered word, it can be understood that “Ndjamena” is a noun equivalent word by morphological analysis.
No further information is available. It is possible to presume that this seems to be a region name by rules using parallel noun phrases.

(Case Pattern Recognition Rule) A rule for estimating and recognizing an individual name using the co-occurrence relationship between a noun and a verbal in the same sentence can be considered.

For example, in the sentence "FULLBAND / wo / 10 / month / ni / just / ga / release / do / ta", "FULLBA /
If ND is an unregistered word, morphological analysis will use FULLBA
We only know that "ND" is a noun equivalent.

In this rule, “release” in the same sentence
Utilizing the co-occurrence relationship that the target case of "FULLBAND" is "FULLBAND", "FULLBAND" is presumed to be a trade name and is recognized as an individual name.

FIG. 6 shows a state in which an authorization rule including a co-occurrence relationship authorization rule is stored in the rule storage unit 155. The authorization rules shown in FIG. 6 are defined with reference to the information on characteristic words and co-occurrence words shown in FIG.

For example, "new product" is "T11", open brackets are "T12", and closed brackets are "T1".
"3", "wo" is "T14", "Release" is "T15"
Attribute (category) in the morpheme sequence "new product /" / gin / "" (double quotation marks are used for clarity) because of the co-occurrence relationship
"Product name" will be added.

(Inter-sentence anaphor relationship recognition rule) In addition, the morpheme (column) representing the once-recognized individual name is stored in the second individual name storage unit 154 along with its attribute, so that a different sentence It is possible to estimate the individual name by using the anaphor relationship and to certify the individual. While the morpheme (column) representing the individual name is newly discovered and the morpheme (column) representing the newly discovered individual name is additionally updated in the second individual name storage unit 154, the above rule is repeated. It can be realized by applying.

For example, “Suzuki / san / no /
If there is a sentence such as “develop / do / ta”, “Suzuki” is stored in the second individual name storage unit 1 as the individual name of the role feature “person name”.
54. As a result, even if there is a sentence such as “[/ product /// development / wa / major / de / shi / ta /] / (/ suzuki /)” at the back of the sentence, “suzuki” in this sentence is It can be estimated and estimated as the individual name of the role feature "person name".

In addition, URL, e-mail address,
If a rule that recognizes a part that matches a regular expression corresponding to the telephone number, facsimile number, etc. is set as an individual name, it is possible to set a rule based on the type (pattern) of the character and to perform the recognition. .

Note that “Tokushima” is stored in the first individual name storage unit 15.
3 and "prefecture" appears after the place name, a rule indicating that it is "prefecture name" is stored in the rule storage unit 154, and "Tokushima prefecture" is stored in the first individual name storage unit 153. There is no need to memorize. By using the authorization rules in this way, it is possible to save the storage area and make the management of terms easy.

These rules can be changed, added, or deleted as appropriate, and the changed, added, or deleted embodiments are also included in the scope of the present invention.

(Second individual name storage unit) As described above,
The morpheme (row) representing the individual name once recognized is temporarily stored in the second individual name storage unit 154 along with its attributes (part of speech, character face, character type, category, role feature, etc.).
At this time, the format of the storage area may be the same as that of the first individual name storage unit 153.

The above-described recognition rule is for certifying an individual name by a morpheme (row) arranged in the vicinity. Therefore, the certification rule alone may result in insufficient certification. For example, "We released Fullband.
Fullband is wonderful. In a sentence like "
Even if “Fullband” is not registered, the feature word / co-occurrence word “release” stored in the first individual name storage unit 153
From the authorization rules stored in the rule storage unit 155, it can be seen that “Fullband” is an individual name representing a product name.

The information is stored in the second individual name storage unit 154.
If the second sentence appears alone, "Fullband"
Is difficult to guess, but in the present invention, the certification result of the first sentence is stored in the second individual name storage unit 154.
Is stored. Therefore, also in the second sentence, “Fu
It is understood that "llband" is an individual name representing a product name.

When the same character appears, the result of morphological analysis and the attributes of the morpheme (column) are checked instead of being unconditionally reused. For example, in a sentence such as “I want to travel to the moon. Today is November 1”, the first “moon” is recognized as an individual name representing a place name because “travel” is a co-occurrence word. Remember, the next month
Is not recognized as a place name because a part-of-speech of time is added as a result of morphological analysis.

Since the second individual name storage unit 154 contains information on the individual name unique to the text currently being processed, the second individual name storage unit 154 needs to be deleted each time the type of document changes.

As for the erasing timing, the following method can be considered.

When the processing of a morpheme sequence derived from a different document is started.

When the number of morphemes processed after the morpheme (column) representing the individual name is stored in the second individual name storage unit 154 exceeds a predetermined number.

The period during which the morpheme (column) representing the individual name is stored in the second individual name storage unit 154 and which has not been used by the certification rule is a predetermined period (such as the number of processed morphemes). If exceeded.

When there is an instruction from the user.

As described above, according to the present invention, by storing the recognized individual names in the second individual name storage unit 154 temporarily, the individual names recognized for one sentence using a certain recognition rule can be obtained. Can be used in other sentences. As a result, even in a sentence in which the recognition rule cannot be directly applied, the individual name can be recognized.

(Output Format) FIG. 7 is an explanatory diagram showing a display example in which the result of individual name recognition by the individual name extraction device of the present invention is output. In the text shown in this figure, the part highlighted in italics and underlined is the morpheme (column) extracted as the individual name. Morphological elements (columns) such as "Korea", "99.9.10", and "ten pages" are extracted.

In the output example shown in this figure, the output unit outputs the morpheme (column) of the extracted individual name to the screen, but further adds information and passes the extraction result to another application or the like. be able to.

For example, a morpheme (column) of the extracted individual name and an attribute (including a category, a role feature, etc.) corresponding thereto can be output together. The attribute of the morpheme (column) can be obtained from information stored in the first individual name storage unit 153, the second individual name storage unit 154, and the rule storage unit 155.

According to this embodiment, the following output can be performed. For "1 / month / 26 / day", the role feature "number of months" is set to "1", the role feature "days" is set to "26", the category "date" is set for "January 26" as a whole,
These are output in association with each other. For “Yamada / Manager”, the role feature “surname” is output to “Yamada”, and the category “person name” is output to “Yamada” as a whole. "Tanaka" for "Corporation / Tanaka / Shokai / Yamada / President"
The role feature "company name", the company "role name" for "shokai", the attribute "company name" for "Tanaka Shokai", the attribute "company name" for "Tanaka Shokai Co., Ltd." and "Yamada". The role feature "person name; surname" and the "president" the role feature "title"
The attribute “person name” is output in association with “President Yamada”, and the category “person name” is output in association with “President Tanaka Corporation Yamada”.

By outputting such attribute information together, it is possible to search for documents based on attributes, such as extracting only individual names considered to be “surname”, and to use the information in machine summarization.

[0153]

As described above, according to the present invention,
An extraction device, an extraction method, and an extraction method suitable for extracting a morpheme representing an individual name or a sequence thereof from an input morpheme sequence.
A recording medium can be provided.

[Brief description of the drawings]

FIG. 1 is a schematic diagram showing a schematic configuration of an information processing device for realizing an individual name extracting device of the present invention.

FIG. 2 is a schematic diagram showing a schematic configuration of each function of the individual name extracting device of the present invention.

FIG. 3 is a flowchart showing a flow of processing executed by the individual name extracting apparatus of the present invention, that is, the processing flow of the individual name extracting method of the present invention.

FIG. 4 is an explanatory diagram showing morphemes (columns) representing individual names stored in a first individual name storage unit of the present invention and examples of attributes thereof.

FIG. 5 is an explanatory diagram showing a state in which characteristic words and co-occurrence words used in the authorization rule of the present invention are stored in a rule storage unit.

FIG. 6 is an explanatory diagram showing a state in which co-occurrence relation recognition rules stored in a rule storage unit of the present invention are stored in a rule storage unit.

FIG. 7 is an explanatory diagram showing output results of the individual name extracting device and the extracting method of the present invention.

[Explanation of symbols]

 Reference Signs List 101 information processing device 102 CPU 103 ROM 104 hard disk 105 keyboard 106 mouse 107 RAM 108 display device 109 interface 110 FD drive 111 CD-ROM drive 151 individual name recognition device 152 input receiving unit 153 first individual name storage unit 154 second Name storage unit 155 rule storage unit 156 certification unit 157 update unit 158 output unit

Claims (15)

[Claims] An input receiving unit for receiving a morpheme string as an input; a first individual name storage unit for storing in advance a morpheme or a string representing an individual name in association with its attribute; a morpheme representing an individual name; A second individual name storage unit for temporarily storing the column in association with the attribute, and a rule for preliminarily storing a morpheme representing the individual name from the morphological sequence or a rule for identifying the column in association with the attribute. A rule storage unit, and the first or second from the received morpheme sequence
A certification unit that recognizes a morpheme or a sequence thereof stored in the individual name storage unit, and also recognizes a morpheme or a sequence thereof representing an individual name from the received morpheme sequence by the previously stored rule. An updating unit that updates the individual name storage unit 2 and temporarily stores a morpheme or a row of the identified individual name in association with its attribute; and a morpheme or a morpheme representing the identified individual name. An output unit that outputs the column as an extracted individual name. 2. A re-authorization control unit for re-authorizing a morpheme representing an individual name or a column thereof, when the second individual name storage unit is updated by the updating unit. The extraction device according to claim 1. 3. The image processing apparatus according to claim 2, further comprising an erasing unit for erasing a morpheme or a column of the temporarily stored individual name from the second individual name storage unit when a specific condition is satisfied. The extraction device according to claim 1 or 2. 4. The certifying section according to claim 1, wherein said morpheme is sandwiched between a first morpheme or a string thereof representing an individual name and a second morpheme or a string thereof representing an individual name according to the rule stored in advance. 4. The extraction device according to claim 1, wherein the sequence is identified as a morpheme representing an individual name or a sequence thereof. 5. 5. The certifying unit according to claim 1, wherein the certifying unit certifies a morpheme connected to a predetermined morpheme or a string thereof or a string thereof as a morpheme representing an individual name or a string thereof according to the previously stored rule. Item 4. The extraction device according to any one of Items 1 to 3. 6. A first individual name storage unit for storing a morpheme representing an individual name or its column in advance in association with its attribute, and temporarily storing a morpheme or its column representing an individual name in association with its attribute. A computer comprising a second individual name storage unit that stores the morphemes representing the individual names from the morpheme sequence or a rule for associating the morphemes with the attributes thereof in advance. An input method for receiving a morpheme string as an input, and an input method for extracting an individual name from the received morpheme string;
Certifying a morpheme or a column thereof stored in the individual name storage unit of the individual, and certifying a morpheme or a column thereof representing the individual name by the previously stored rule from the received morphological sequence; An update step of updating the individual name storage unit 2 and temporarily storing a morpheme or a column thereof representing the recognized individual name in association with its attribute; and a morpheme or a morpheme representing the recognized individual name. An output step of outputting the column as an extracted individual name. 7. A re-certification control step of re-certifying a morpheme representing an individual name or a column thereof in the certification step when the second individual name storage unit is updated in the updating step. The extraction method according to claim 6, characterized in that: 8. The method according to claim 1, further comprising an erasing step of erasing a morpheme or a column of the temporarily stored individual name from the second individual name storage unit when a specific condition is satisfied. The extraction method according to claim 6. 9. In the certifying step, a first morpheme or a string representing an individual name and a second morpheme or a string representing an individual name are sandwiched by the rule stored in advance. The method according to any one of claims 6 to 8, wherein the morpheme or a sequence thereof is recognized as a morpheme or a sequence representing an individual name. 10. The certifying step according to claim 1, wherein a morpheme connected to a predetermined morpheme or a sequence thereof or a sequence thereof is recognized as a morpheme representing an individual name or a sequence thereof according to the rule stored in advance. The extraction method according to any one of claims 6 to 8. 11. An input receiving unit for receiving a morpheme string as an input, a morpheme representing an individual name or a first individual name storage unit for storing a string thereof in association with its attribute, a morpheme representing an individual name. Or a second individual name storage unit for temporarily storing the column in association with the attribute, and storing in advance a rule for identifying the morpheme representing the individual name or the column from the morphological sequence in association with the attribute. A rule storage unit, the first or second from the received morpheme sequence
A morpheme stored in the individual name storage unit or a column thereof, and a certifying unit that certifies a morpheme representing the individual name or a column thereof based on the previously stored rule from the received morpheme sequence, An update unit that updates the individual name storage unit and temporarily stores the morpheme or the column representing the recognized individual name in association with the attribute, and the morpheme or the morpheme representing the recognized individual name. A computer-readable recording medium for recording a program for causing a column to function as an output unit that outputs the extracted individual name. 12. The re-authorization control for causing the computer to re-authorize a morpheme representing an individual name or a column thereof when the second individual name storage unit is updated by the updating unit. The computer-readable recording medium according to claim 11, wherein the recording medium further functions as a unit. 13. The program according to claim 1, wherein the computer deletes a morpheme representing the temporarily stored individual name or a sequence thereof from the second individual name storage unit when a specific condition is satisfied. 13. The computer-readable recording medium according to claim 11, further comprising: 14. When the computer functions as the certification unit according to the program, the certification unit, according to the rule stored in advance, stores a first morpheme or a string thereof representing an individual name and a first morpheme representing the individual name. 14. The computer-readable medium according to claim 11, wherein the morpheme or its row sandwiched between the second morpheme or its row is identified as a morpheme or its row representing an individual name. Possible recording medium. 15. When the computer functions as the recognition unit according to the program, the recognition unit specifies a morpheme or a row connected to a predetermined morpheme or a row thereof by an individual name according to the previously stored rule. 14. The computer-readable recording medium according to claim 11, wherein the morpheme is represented as a morpheme or a sequence thereof.