JP2020154790A - Information processing device, information processing method, and program - Google Patents

Abstract

To extract a unique expression from a document with high accuracy.SOLUTION: An information processing device includes a division part for dividing a sentence into a character string including at least one or more characters, a calculation part for calculating a score in each character string divided by the division part on the basis of a plurality of queries inputted by a user, and an extraction part for extracting a unique expression from the sentence on the basis of the score calculated by the calculation part.SELECTED DRAWING: Figure 2

Description

The present invention relates to an information processing device, an information processing method, and a program.

Research is being conducted to replace the content of a sentence with an expression that can be used for machine learning by using a proper expression (for example, a proper noun) contained in the sentence as a feature quantity. In relation to this, the input text is divided into phrases by morphological analysis, and the impression element and score are divided for each phrase in which the input text is divided using an impression dictionary in which the impression element and its score are associated with the phrase in advance. Generate an associated phrase list, calculate the likelihood of each impression element from the entire phrase of the input text, calculate the objective impression associated with the likelihood as the score of each impression element, and make it a subjective impression. There is known a technique for calculating impression difference information by comparing the score of the impression element based on the score of the impression element based on the score of the impression element based on an objective impression (see, for example, Patent Document 1).

JP-A-2017-84015

New words that have not been used until now, such as new words and coined words, may become popular in the world. For example, if new content with a unique title is released and the content becomes a hot topic among people, the unique title will become popular as a new word. However, with the conventional technology, it is often difficult to update the dictionary frequently according to the fashion, and further, it is thoroughly examined from what kind of sentence the unique expression to be registered in the dictionary should be searched for. I wasn't. For this reason, conventional techniques may not be able to accurately extract named entities from a document.

The present invention has been made in view of the above problems, and an object of the present invention is to provide an information processing device, an information processing method, and a program capable of accurately extracting a named entity from a document.

One aspect of the present invention is a division unit that divides a sentence into character strings including at least one character, and a score for each character string divided by the division unit based on a plurality of queries input by the user. It is an information processing device including a calculation unit for calculation and an extraction unit for extracting a named entity from the sentence based on the score calculated by the calculation unit.

According to one aspect of the present invention, named entity can be accurately extracted from a document.

It is a figure which shows an example of the information processing system 1 including the information processing apparatus 100 in 1st Embodiment. It is a figure which shows an example of the structure of the information processing apparatus 100 in 1st Embodiment. It is a figure which shows an example of the search log 132. It is a flowchart which shows the flow of a series of processing of the control unit 110 in 1st Embodiment. It is a figure which shows typically the method of extracting a named entity from a three-character text. It is a figure which shows typically the method of extracting a named entity from a three-character text. It is a figure which shows typically the method of extracting a named entity from a three-character text. It is a figure which shows typically the method of extracting a named entity from a three-character text. It is a figure which shows typically the method of extracting a named entity from a four-character text. It is a figure which shows typically the method of extracting a named entity from a four-character text. It is a figure which shows typically the method of extracting a named entity from a four-character text. It is a figure which shows typically the method of extracting a named entity from a four-character text. It is a figure which shows typically the method of extracting a named entity from a four-character text. It is a figure which shows typically the method of extracting a named entity from a four-character text. It is a figure which shows typically the method of extracting a named entity from a four-character text. It is a figure which shows typically the method of extracting a named entity from a four-character text. It is a figure for demonstrating the method of determining the pattern of the score calculation target. It is a figure for demonstrating the method of determining the pattern of the score calculation target. It is a figure for demonstrating the method of determining the pattern of the score calculation target.

Hereinafter, an information processing apparatus, an information processing method, and a program to which the present invention is applied will be described with reference to the drawings.

[Overview]
The information processing device is realized by one or more processors. The information processing device divides a sentence into a character string including at least one character, and calculates a score for each character string based on a search log which is a history of a plurality of queries input by a user. Then, the information processing device extracts the named entity from the sentence based on the calculated score. As a result, the named entity can be accurately extracted from the document. As a result, for example, it is possible to obtain a distributed expression that accurately represents the content of the sentence.

<First Embodiment>
[overall structure]
FIG. 1 is a diagram showing an example of an information processing system 1 including an information processing device 100 according to the first embodiment. The information processing system 1 in the first embodiment includes, for example, one or more terminal devices 10, a service providing device 20, and an information processing device 100. Some or all of these devices are connected to each other via a network NW. A part of these devices may be included in other devices as a virtual device. For example, a part or all of the functions of the service providing device 20 are realized by the functions of the information processing device 100. The virtual machine may be a virtual machine, or conversely, a part or all of the functions of the information processing device 100 may be realized by the functions of the service providing device 20.

Each device shown in FIG. 1 transmits and receives various information via the network NW. The network NW includes, for example, a radio base station, a Wi-Fi access point, a communication line, a provider, the Internet, and the like. It should be noted that not all combinations of the devices shown in FIG. 1 need to be able to communicate with each other, and the network NW may partially include a local network.

The terminal device 10 is a terminal device including an input device, a display device, a communication device, a storage device, and a calculation device, such as a mobile phone such as a smartphone, a tablet terminal, and various personal computers. The communication device includes a network card such as a NIC (Network Interface Card), a wireless communication module, and the like. In the terminal device 10, a UA (User Agent) such as a web browser or an application program is activated, and a request corresponding to a user's input is transmitted to the service providing device 20. Further, the terminal device 10 in which the UA is activated causes the display device to display various images based on the information acquired from the service providing device 20.

The service providing device 20 is, for example, a web server that provides a web page to the terminal device 10 in response to a request from a web browser started as a UA. The web page may be, for example, a web page constituting various websites such as a shopping site, an auction site, and a flea market site. Further, the service providing device 20 may provide the terminal device 10 with a web page that provides various services such as a search site, an SNS (Social Networking Service), and a mail service. Further, the service providing device 20 is an application server that provides the same services as various websites such as sales sites by providing contents to the terminal device 10 in response to a request from an application started as a UA. May be good.

The information processing device 100 acquires a search log from the service providing device 20, and uses the search log to extract a unique expression from the text. The proper expression according to the present embodiment includes, for example, not only one word (word) such as a noun, but also one phrase (phrase) in which a noun and a noun are connected by another part of speech (for example, a particle). A sentence containing various parts of speech such as nouns, verbs, particles, and particles is included. That is, any word used by humans as a named entity can be a named entity, no matter how long the sentence.

[Information processing device configuration]
FIG. 2 is a diagram showing an example of the configuration of the information processing apparatus 100 according to the first embodiment. As shown in the figure, the information processing device 100 includes, for example, a communication unit 102, a control unit 110, and a storage unit 130.

The communication unit 102 includes, for example, a communication interface such as a NIC (Network Interface Card) and a DMA (Direct Memory Access) controller. The communication unit 102 communicates with the service providing device 20 and another web server via the network NW.

The control unit 110 includes, for example, an acquisition unit 112, a text segmentation unit 114, a phrase score calculation unit 116, and a named entity extraction unit 118. The components of the control unit 110 are realized, for example, by a processor such as a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit) executing a program stored in the storage unit 130. In addition, some or all of the components of the control unit 110 are realized by hardware (circuit unit; circuitry) such as LSI (Large Scale Integration), ASIC (Application Specific Integrated Circuit), FPGA (Field-Programmable Gate Array). It may be done, or it may be realized by the cooperation of software and hardware.

The storage unit 130 is realized by, for example, an HDD (Hard Disc Drive), a flash memory, an EEPROM (Electrically Erasable Programmable Read Only Memory), a ROM (Read Only Memory), a RAM (Random Access Memory), or the like. The storage unit 130 stores the search log 132 in addition to various programs such as firmware and application programs.

FIG. 3 is a diagram showing an example of the search log 132. As shown in the illustrated example, the search log 132 is historical information associated with each query input by the user to the search engine, the number of times the query is input, and the like for each aggregated period. The number of inputs may be, for example, the number of unique browser cookies. In this case, even if the same query is input many times through the same browser, the number of times the query is input is counted as one.

[Processing flow]
Hereinafter, a series of processing flows of the control unit 110 in the first embodiment will be described with reference to a flowchart. FIG. 4 is a flowchart showing a series of processing flows of the control unit 110 according to the first embodiment. The processing of this flowchart may be repeated, for example, at a predetermined cycle.

First, the acquisition unit 112 acquires a text, which is one of the contents, from the service providing device 20 via the communication unit 102 (S100). For example, a text is a text that can be searched by a user using a search engine. Specifically, an article or an organization posted on an official website operated by an organization, an organization, a celebrity, or the like. Articles posted on general sites (unofficial sites) operated by third parties unrelated to organizations, celebrities, etc.

Next, the acquisition unit 112 acquires the search log 132 from the service providing device 20 via the communication unit 102 (S102). For example, the acquisition unit 112 acquires the search log 132 including the history of queries entered by the user in the last few months.

Next, the text segmentation unit 114 divides the text acquired by the acquisition unit 112 into one or a plurality of phrases (S104). The phrase according to the present embodiment may be a single character such as particles such as "ha", "ga", and "o", or may be a character string including a plurality of characters. That is, in the description of the present embodiment, the term phrase is not used in the dictionary sense of being a collection of a plurality of words, but is referred to as one character, one word, one phrase, one sentence. , Used in a broader sense.

For example, when the number of characters included in the text is N, the text segmentation unit 114 divides the text by a pattern of a combination of phrases according to 2 (N-1) power. For example, if N = 1, the text is divided into one pattern, if N = 2, it is divided into two patterns, if N = 3, it is divided into four patterns, and N = 4. If so, it is divided into eight patterns.

Next, the phrase score calculation unit 116 quantifies the certainty that the phrase is a named entity for each phrase divided by the text segmentation unit 114 based on the search log 132 (hereinafter, phrase). (Referred to as a score) is calculated (S106). For example, the phrase score calculation unit 116 calculates the phrase score based on the mathematical formula (1).

In the formula, S represents the phrase score, β represents the number of times a query is input that matches the phrase for which the score is to be calculated (the number of searches using that query), and α is a real number greater than 1 (for example, 10). ), And L represents the length of the phrase for which the score is to be calculated, that is, the number of characters included in the phrase.

For example, if the phrase to be scored matches any of the queries included in the search log 132, that is, if the phrase to be scored is entered as a query more than once, the phrase score calculation unit. In 116, the larger the number of query inputs β and the larger the phrase length L, the larger the phrase score S of the target phrase, and the smaller the number of query inputs β and the smaller the phrase length L, the larger the phrase score S of the target phrase. Decrease the phrase score S. If the phrase to be scored does not match any of the queries included in the search log 132, that is, if the phrase to be scored is not entered as a query, the phrase score calculation unit 116 is the target. Set the phrase score S of the phrase to 0.

Next, the named entity extraction unit 118 extracts the named entity from the text based on the phrase score calculated for each phrase by the phrase score calculation unit 116 (S108). This ends the processing of this flowchart.

5 to 8 are diagrams schematically showing a method of extracting a named entity from a three-character text. These figures schematically show that the three-letter text "ABC" is divided into one or more phrases. Therefore, the text is divided into 4 ( ²² ) patterns.

For example, FIG. 5 shows pattern 1 in which one text "ABC" is divided into a one-letter phrase "A", a one-letter phrase "B", and a one-letter phrase "C". Represents. In the case of pattern 1, the phrase score calculation unit 116 calculates the phrase score S _A for the phrase "A", calculates the phrase score S _B for the phrase "B", and calculates the phrase score S _B for the phrase "C". is to calculate the phrase score S _C.

FIG. 6 represents pattern 2 in which one text "ABC" is divided into a two-letter phrase "AB" and a one-letter phrase "C". Phrase score calculating unit 116, the phrase "AB" computes the phrase score S _AB, the phrase "C", calculates the phrase score S _C.

FIG. 7 represents pattern 3 in which one text "ABC" is divided into a one-letter phrase "A" and a two-letter phrase "BC". The phrase score calculation unit 116 calculates the phrase score S _A for the phrase "A" and calculates the phrase score S _BC for the phrase "BC".

FIG. 8 shows a pattern 4 in which one text "ABC" is used as it is as one phrase. The phrase score calculation unit 116 calculates the phrase score S _ABC for the phrase "ABC".

When the phrase score calculation unit 116 calculates the phrase score S of each phrase for each pattern as described above, the phrase score calculation unit 116 calculates the sum of the phrase scores S for each pattern. In pattern 1 illustrated in FIG. 5, the sum of the phrase score _{S, (S A + S B} + S C) next to, in the pattern 2 illustrated in FIG. 6, the sum of the phrase score S _{is, (S} AB + _{S C),} and the in pattern 3 illustrated in FIG. 7, the sum of a phrase score S _is, in the pattern 4 illustrated (S a _{+ S BC),} and the 8, the sum of a phrase score S _{becomes (S ABC).}

The named entity extraction unit 118 selects a pattern having the maximum sum of the phrase scores S from these four patterns, and extracts the phrase represented by the pattern as a named entity. For example, if the sum S _ABC phrase score S of the pattern 4 is the largest, named entity extraction unit 118 extracts one phrase "ABC" as a named entity. Further, for example, if the sum of the phrase score S of the pattern 2 (S _{AB + S C)} is the maximum, named entity extraction unit 118 extracts a phrase and phrase "AB", "C" as the respective named entities ..

9 to 16 are diagrams schematically showing a method of extracting a named entity from a four-character text. These figures schematically show that the four-letter text "ABCD" is divided into one or more phrases. Therefore, the text is divided into 8 (2 ³ ) patterns.

For example, FIG. 9 shows a pattern 1-1 in which one text "ABCD" is divided into phrases for each character. FIG. 10 represents a pattern 1-2 that divides one text "ABCD" into three phrases "A", "B", and "AB". FIG. 11 represents a pattern 2-1 that divides one text "ABCD" into three phrases "AB", "C", and "D". FIG. 12 represents a pattern 2-2 that divides one text "ABCD" into two phrases "AB" and "CD". FIG. 13 represents a pattern 3-1 that divides one text "ABCD" into three phrases "A", "BC", and "D". FIG. 14 represents a pattern 3-2 that divides one text "ABCD" into two phrases "A" and "BCD". FIG. 15 represents a pattern 4-1 that divides one text "ABCD" into two phrases "ABC" and "D". FIG. 16 shows a pattern 4-2 in which one text "ABCD" is used as it is as one phrase. Similarly to the above, when the phrase score calculation unit 116 calculates the phrase score S of each phrase for each pattern, the phrase score calculation unit 116 calculates the sum of the phrase scores S for each pattern. Then, the named entity extraction unit 118 selects the pattern having the maximum sum of the phrase scores S from these eight patterns, and extracts the phrase represented by the pattern as the named entity.

According to the first embodiment described above, the text is divided into one or a plurality of phrases in a pattern of a combination of numbers based on the number of characters N of the text, and a phrase score is calculated for each divided phrase in each pattern. Then, the sum of the phrase scores is calculated for each pattern, and the phrase of the pattern having the maximum calculated sum is extracted as a named entity. As a result, the named entity can be accurately extracted from the document.

Traditionally, named entities have been extracted from text using a dictionary in which named entities are registered in advance, but named entities such as new words are appearing every day, and it is necessary to update the dictionary frequently. is there. However, it is practically difficult to update the dictionary daily. In addition, it is difficult for niche terms that have begun to be used as new words in some communities to be registered in the dictionary as named entities.

Therefore, for example, the title of the new content is a sentence such as "○○○ Official Guidebook ・ How to walk from ◇◇ to △△", and a named entity is extracted by applying a dictionary to the text containing this title. In this case, multiple words such as "○○○", "official", "guidebook", "◇◇", "△△", and "how to walk" are extracted as named entities, and "○" should be extracted. 〇〇 Official guidebook ・ The sentence "How to walk from ◇◇ to △△" will not be extracted as a named entity.

On the other hand, when considering the nature of queries, if the title of a new content is as long as a sentence, the user copies the character string representing the title from the official website or the website of a third party, and then It is assumed that the copied character string will be pasted in the input field of the search site. In this case, the query that is the same word by word as the title of the content that is a named entity is collected as the search log 132. In particular, the most recent search log 132, such as the last few months, tends to include new words that are currently popular as queries. Therefore, in the present embodiment, by comparing the phrase in which the text is divided with the query of the search log 132, the named entity is accurately extracted from the text even if the named entity is long or is a brand new named entity. be able to.

It is also conceivable to extract a part of the text surrounded by punctuation marks (descriptive symbols) such as parentheses, apostrophes, and primes as named entities. However, in this method, a person's dialogue or quotation may be extracted as a named entity, and it is possible to distinguish whether it is a named entity used as one noun or simply a dialogue or quotation. Can not. In addition, although the character name and personal name of the content are unique expressions, they cannot be extracted from the text because they are usually not enclosed in parentheses.

On the other hand, in the present embodiment, named entity can be extracted regardless of punctuation marks. Also, movie and book titles often have subtitles, which may be surrounded by punctuation marks. Even if the named entity is surrounded by punctuation marks, if the user inputs the named entity surrounded by punctuation marks as a query, the named entity including the punctuation marks is also extracted by the method of the present embodiment. be able to.

Also, if you simply compare the phrase in the text with the query in the search log 132, relatively short phrases such as "ha", "o", "desu", and "masu" are likely to appear in the text. The phrase coincides with the query by chance, and the phrase score S tends to increase. On the other hand, in the present embodiment, the product of an arbitrary radix α with the phrase length L as the exponent and the number of inputs β is the phrase score S. Therefore, even if the phrase has a small number of inputs β, the phrase length L If is large, the phrase score S can be increased, and even if the phrase has a large number of inputs β, the phrase score S can be decreased if the phrase length L is small. As a result, it is possible to accurately extract a phrase or sentence in which a plurality of nouns are connected by particles or the like as one named entity while suppressing the extraction of particles or the like as a named entity.

<Second Embodiment>
Hereinafter, the second embodiment will be described. In the first embodiment described above, the sum of phrase scores is calculated for all patterns that are a combination of phrases, and the phrase of the pattern having the maximum sum is extracted as a named entity. On the other hand, in the second embodiment, the sum of the phrase scores is calculated after rationally determining the pattern to be verified, instead of calculating the sum of the phrase scores for all the patterns. Different from the embodiment. Hereinafter, the differences from the first embodiment will be mainly described, and the points common to the first embodiment will be omitted. In the description of the second embodiment, the same parts as those of the first embodiment will be described with the same reference numerals.

17 to 19 are diagrams for explaining a method of determining a pattern to be scored. For example, when a four-character text "ABCD" is given, the text segmentation unit 114 according to the second embodiment has the first character "A" at the beginning of the text followed by the first character "A" as illustrated in FIG. The two-letter "B" is split to generate the phrase "A" and the phrase "B", and the two-letter phrase "AB" is generated without splitting between these letters. .. Phrase score calculating unit 116 according to the second embodiment, by using the search log 132, calculates the sum of the phrase score _{S B} phrase phrase score _{S A} and phrases "A", "B" _(S A + _{S B)} as well as, to calculate the phrase score _{S AB} of the phrase "AB". The text segmentation unit 114 compares these and deletes the pattern derived from the pattern having the smaller score from the pattern candidates to be verified next. The first character "A" is an example of "first character string", the second character "B" is an example of "second character string", and the phrase "AB" is "third character string". This is an example.

In the example of FIG. _17, the larger _{S AB} than _(S A + S _B). In this case, the pattern of the phrase combination including the third character "C" following the second character is a total of four types of the above-mentioned pattern 1, pattern 2, pattern 3, and pattern 4. Of these four types of patterns, at least pattern 1 and pattern 2 are the same as the results before including the third character "C". For example, the sum of the phrase score pattern 1 is _{(S A + S B + S} C), the sum of the phrase scores pattern 2 because it is _(S AB + S _C), than the previous time _(S A + S _B) If the result is that S _AB is larger, the magnitude relationship of the sum of the phrase scores of pattern 1 and pattern 2 does not change. Therefore, the phrase score calculation unit 116 does not calculate the score for the pattern 1 in which the first character "A" at the beginning of the text and the second character "B" following it are divided.

Next, the text segmentation unit 114 compares the scores of the remaining three types of patterns 3 as illustrated in FIG. In the example of FIG. 18, the phrase score S _{ABC of} pattern 4 is the largest. Therefore, the phrase score calculation unit 116 has a pattern 3-1 that divides between the first character "A" of the text and the combination "BCD" of the following three characters, and the combination "AB" of the first two characters of the text. Pattern 2-2 that divides between and the following two-letter combination "CD", and pattern 4-1 that divides between the first three-letter combination "ABC" and the following one-letter "D". And, the score is calculated only for a total of 4 patterns of pattern 4-2, which is a single phrase without dividing the text. As described above, in this method, when examining the last k-th character of the character string, k patterns are compared.

Next, the phrase score calculation unit 116 calculates the sum of the phrase scores of each of the above four patterns (3-1, 2-2, 4-1 and -2). For example, when the sum of the phrase scores of the pattern 4-1 (S _ABCD + _SD ) is the largest, the named entity extraction unit 118 extracts the phrase "ABC" and the phrase "D" as named entities, respectively. In this way, the optimum combination can be searched for by combining the characters in order from the first character of the text and selecting the candidates for each combination according to the size of the score.

When the phrase score of a certain pattern becomes 0, it is not necessary to consider the derived pattern in which one character is added to the pattern. When the phrase score is 0, it means that the number of inputs β of the query matching the phrase is 0. That is, it means that no user has input the phrase represented by the pattern as a query, and it is extremely unlikely that the phrase of the pattern is a named entity.

According to the second embodiment described above, the characters are combined in order from the first character of the text, the phrase scores of each pattern are compared at that time, and the patterns having a smaller phrase score are excluded from the subsequent processing targets.

For example, some contents such as movies, dramas, and animations have proper nouns whose titles are particles such as "no" and "to" between a certain word A and a certain word B. Specifically, the titles are "○○ and □□" and "○○ △△". When the phrase score is obtained for all patterns of the title of such content, the score is also calculated for the phrase in which particles such as "no" and "to" appear at the beginning of the noun. However, in the real world, particles rarely appear at the beginning of a phrase, and it can be considered that the phrase itself does not exist in the world. Therefore, by combining in order from the first character of the text, comparing the phrase scores of each pattern at that point, and excluding patterns with smaller phrase scores from the subsequent processing targets, proper nouns are efficiently extracted. be able to.

When the method of separating all patterns is tried as in the first embodiment described above, if the number of characters included in the text is N, the number of times the score is calculated increases as in 2 ^N-1 . In contrast, in the second embodiment, in consideration of the character of continuity, to reduce the number of patterns, it is possible to suppress the number of calculations of the score ² times N.

<Hardware configuration>
The information processing device 100 of the above-described embodiment is realized by, for example, a hardware configuration as shown in FIG. FIG. 19 is a diagram showing an example of the hardware configuration of the information processing apparatus 100 of the embodiment.

The information processing device 100 includes NIC100-1, CPU100-2, RAM100-3, ROM100-4, secondary storage devices 100-5 such as flash memory and HDD, and drive device 100-6, which are internal buses or dedicated communication lines. It is configured to be interconnected by. A portable storage medium such as an optical disk is mounted on the drive device 100-6. A program stored in a portable storage medium mounted on the secondary storage device 100-5 or the drive device 100-6 is expanded into the RAM 100-3 by a DMA controller (not shown) or the like, and executed by the CPU 100-2. As a result, the control unit 110 is realized. The program referred to by the control unit 110 may be downloaded from another device via the network NW.

Although the embodiments for carrying out the present invention have been described above using the embodiments, the present invention is not limited to these embodiments, and various modifications and substitutions are made without departing from the gist of the present invention. Can be added.

1 ... Information processing system, 10 ... Terminal device, 20 ... Service providing device, 100 ... Information processing device, 102 ... Communication unit, 110 ... Control unit, 112 ... Acquisition unit, 114 ... Text segmentation unit, 116 ... Phrase score calculation unit , 118 ... Named entity extraction unit, 130 ... Storage unit

Claims (6)

A division part that divides a sentence into a character string containing at least one character,
A calculation unit that calculates a score for each of the character strings divided by the division unit based on a plurality of queries input by the user.
An extraction unit that extracts a named entity from the sentence based on the score calculated by the calculation unit.
Information processing device equipped with. The calculation unit calculates the score based on the number of times the query is input that matches the character string and the length of the character string.
The information processing device according to claim 1. The extraction unit extracts one or more of the character strings obtained when the division unit divides the sentence at a position where the sum of the scores for each character string is maximized, as the named entity.
The information processing device according to claim 1 or 2. The division portion divides the sentence at the first position and
The calculation unit includes a score of a first character string containing at least one character appearing immediately before the first position and a score of a second character string containing at least one character appearing immediately after the first position. , The score of the third character string which is a combination of the first character string and the second character string is calculated.
The extraction unit compares the sum of the score of the first character string and the score of the second character string with the score of the third character string, and expresses the character string having the smaller score as the unique expression. Exclude from the extraction target of
The information processing device according to any one of claims 1 to 3. The computer
Divide the sentence into a string containing at least one character,
Based on a plurality of queries entered by the user, a score is calculated for each of the divided character strings.
Based on the calculated score, the named entity is extracted from the sentence.
Information processing method. On the computer
The process of dividing a sentence into a character string containing at least one character,
A process of calculating a score for each of the divided character strings based on a plurality of queries entered by the user, and
A process of extracting a named entity from the sentence based on the calculated score,
A program to execute.

Images