JP7032582B1 - Information analysis program, information analysis method and information analysis device - Google Patents

Abstract

【Task】
By extracting the named entity and then analyzing the causal relationship of the sentence, the correspondence relationship of the keywords corresponding to the technical terms is analyzed.
SOLUTION:
The information analysis program analyzes the dependency from the text information, the text information acquisition step for acquiring the text information, the named entity extraction step for extracting the named entity from the text information and adding a tag, and the dependency source or the dependency. In the dependency extraction step that extracts the dependency whose one or both of the recipients are named entities, and the dependency extracted in the dependency extraction step from the text information, the degree of co-occurrence is calculated for the combination of the dependencies. It is a program that executes a co-occurrence extraction step that extracts a combination of dependencies based on the co-occurrence degree calculation step and the co-occurrence degree.
[Selection diagram] Fig. 2

Description

The present disclosure relates to an information analysis program, an information analysis method, and an information analysis device that analyze the relevance of a unique expression from text data.

With the progress of natural post-processing technology and machine learning technology in recent years, a method for automatically analyzing text data by a computer has been proposed. For example, if the causal relationship of text data can be analyzed, knowledge can be efficiently systematized by analyzing the main points of assertion of the document from a large amount of document data.

For example, Patent Document 1 proposes a method for analyzing a wide range of causal relationships by using natural language processing.

Japanese Patent Application Laid-Open No. 8-221415

However, the method disclosed in Patent Document 1 merely analyzes the causal relationship, and does not mention that it is applied to other correspondence relationships such as purpose and means, or method and effect.

For example, in the scene of analyzing literature related to specialized technology, not only the analysis of the causal relationship between cause and effect, but also the relationship between a certain event and the analysis method such as experiments, tests, inspections, and grounds to support the event. If it can be analyzed, it will be possible to extract more essential elements when systematizing the document.

Therefore, in the present disclosure, in the analysis of the causal relationship, the dependency including the keyword is extracted by extracting the named entity and then the dependency. Furthermore, in this disclosure, by extracting the co-occurrence relationship of the dependency, a more accurate causal relationship can be analyzed, and not only the analysis of the causal relationship but also the relationship between a certain event and its analysis method can be various. The purpose is to analyze the relationships between words.

It is a program to be executed by a computer equipped with a processor and a memory. The program analyzes the dependency from the text information, the text information acquisition step for acquiring the text information, the named entity extraction step for extracting the named entity from the text information and adding a tag, and the dependency source or the processor. The degree of co-occurrence is calculated for the combination of dependencies in the dependency extraction step that extracts the dependency whose one or both of the dependencies are named entities, and the dependency extracted in the dependency extraction step from the text information. It is a program that executes a co-occurrence degree calculation step and a co-occurrence extraction step that extracts a combination of dependencies based on the co-occurrence degree.

According to the present disclosure, it is possible to accurately analyze causal relationships even for text data including proper nouns, technical terms, etc. by using natural language processing. In addition, it is possible to analyze not only the causal relationship but also the relationship between various words desired by the user, such as the relationship between a certain event and the analysis method.

It is a figure which shows the whole structure of the server of Embodiment 1 of this disclosure. It is a figure which shows an example of the server configuration which concerns on Embodiment 1 of this disclosure. It is a figure which shows the specific example of the named entity extraction. It is a figure which shows the specific example which analyzes the dependency. It is a figure which shows the specific example of the dependency analysis using the information of a named entity. It is a figure which shows the specific example of a sentence including a scientific term. It is a figure which shows the specific example of the named entity extraction. It is a figure which shows the specific example which analyzes the dependency. It is a figure which shows the specific example finally extracted as a dependency. It is a flowchart which shows an example of the processing flow of Embodiment 1 in an information analysis apparatus. It is a figure which shows an example of the structure of the information analysis apparatus which concerns on Embodiment 2 of this disclosure. It is a figure which shows the specific example which displayed the dependency in the display part. It is a figure which shows the specific example which displayed the dependency by paying attention to the hierarchical structure in the display part. It is a flowchart which shows an example of the processing flow of Embodiment 2 in an information analysis apparatus. It is a figure which shows the specific example of the information analysis apparatus via a network.

Hereinafter, the information analysis device, the information analysis method, and the information analysis program according to the embodiment of the present disclosure will be described with reference to the drawings. The embodiments described below do not unreasonably limit the contents of the present disclosure described in the claims. Moreover, not all of the configurations described in the embodiments are essential constituent requirements of the present disclosure. Further, in all the drawings illustrating the embodiments, the common components are designated by the same reference numerals, and the repeated description will be omitted.

<Embodiment 1>
In the first embodiment, text information is acquired and analyzed, a named entity is extracted, then the dependency of the named entity and the co-occurrence relationship are analyzed, and the degree of co-occurrence is calculated for each arbitrary named entity.

<Overall configuration>
FIG. 1 is a diagram showing an overall configuration of the server 10 of the present embodiment. The server 10 is a general-purpose computer. The server 10 is realized by, for example, a stationary PC (Personal Computer), a laptop PC, or the like. Further, the server 10 may be a portable computer such as a smartphone or a tablet terminal.

As shown in FIG. 1, the server 10 includes a processor 11, a memory 12, a storage 13, a communication IF 14, and an input / output IF 15.

The processor 11 is hardware for executing an instruction set described in a program, and is composed of an arithmetic unit, registers, peripheral circuits, and the like. The memory 12 is for temporarily storing a program, data processed by the program or the like, and is realized by a volatile memory such as a DRAM (Dynamic Random Access Memory) or the like. The storage 13 is a storage device for storing data, and is realized by, for example, a flash memory or an HDD (Hard Disk Drive). The communication IF 14 is an interface for transmitting and receiving signals because the server 10 communicates with an external device. The input / output IF 15 functions as an interface with an input device for receiving input from the user and an output device for presenting information to the user.

<Server configuration>
Hereinafter, the configuration of the server 10 will be described in detail. As shown in FIG. 2, the server 10 according to the present embodiment includes a communication unit 101, a storage unit 102, and a control unit 103. The communication unit 101 performs a process for the server 10 to communicate with another device. The communication unit 101 performs transmission processing on the signal generated by the control unit 103, and transmits the signal to the outside. The communication unit 101 performs reception processing on the signal received from the outside and outputs the signal to the control unit 103.

The storage unit 102 is configured by, for example, a flash memory or the like, and stores data and programs used by the server 10.

The control unit 103 is realized by the processor 11 reading a program stored in the storage unit 102 and executing an instruction included in the program. The control unit 103 controls the operation of the server 10. Specifically, for example, the control unit 103 functions as a text information acquisition unit 1031, a named entity extraction unit 1032, a dependency extraction unit 1033, a co-occurrence degree calculation unit 1034, and a co-occurrence extraction unit 1035. Demonstrate.

Hereinafter, the functional configuration of the control unit 103 of the server 10 will be described.

The text information acquisition unit 1031 acquires the text data to be analyzed. The text data to be analyzed is assumed to be text data consisting of a plurality of sentences. Further, the text information acquisition unit 1031 is not necessarily limited to acquiring the text data itself, and may acquire the image data of the document and then convert it into text by OCR (Optical Character Recognition) or the like.

The text information acquisition unit 1031 records computer-readable data such as a USB (Universal Serial Bus) memory, a CD-ROM (Compact Disc-Read Only Memory), a DVD (Digital Versatile Disc), and an HDD via the input / output interface 15. Text data may be acquired from the medium.

The text information acquisition unit 1031 may acquire text data from the network via a communication IF 15 provided with a NIC (Network Interface Card controller) for performing wired communication or wireless communication.

The named entity extraction unit 1032 extracts the named entity and attaches a tag to the text data acquired by the text information acquisition unit 1031.

Named entity is an expression that is a key element when extracting information from data. Named entities include named entities such as personal names, place names, organization names, product names, numerical expressions such as quantities, amounts, and percentages, and time expressions such as dates and times. Named entities may also include specific substances, structures, functions, treatment methods, analytical methods, terminology, and the like. The range of the named entity may be defined in advance from all the expressions.

The tag indicates the identification information for classifying the named entity. For tags, for example, tags representing proper nouns such as "personal name" and "place name", tags representing numerical expressions such as "quantity", "date", and "time" are prepared and assigned to the proper expression. May be good.

In addition, for the purpose of analyzing specialized literature and literature in the field of natural science, tags such as an event tag indicating an "event" and an analysis tag indicating an "analysis method" for an event may be prepared. good. Prepare multiple tags in advance.

Further, the tag may be defined with a hierarchical structure.

For example, in specialized literature such as papers, especially in the field of natural science, there are "analytical methods" such as experiments, tests, tests, and grounds that support certain "events" such as hypotheses, states, and results. Expected to be shown. Then, in the "event", it is assumed that it is shown in the form of "the object is a state" such as "A is B". Further, in the "analytical method", in order to support the event, it is assumed that it is indicated in the form of "A is B" or "the result is obtained by a test or the like".

Based on the above, for example, the tag may be defined with a hierarchical structure such as "event: target", "event: state", "analysis: test method", "analysis: test result", and the like. .. In this case, "event" and "analysis" are superordinate concepts, and "object" and "test method" are defined as tags having a hierarchical structure as subordinate concepts.

As a method of extracting the named entity, the named entity extraction unit 1032 may prepare a dictionary representing the named entity in advance, extract the named entity from the input sentence using the dictionary, and perform tagging. ..

The named entity extraction unit 1032 may use a generally known method such as NER (Named Entry Recognition) as the named entity extraction method. Further, the named entity extraction unit 1032 may extract the named entity after creating a model for extracting the named entity by machine learning using the learning data.

In general NER, the named entity to be extracted is a named entity, a numerical expression, or the like. Therefore, when it is desired to extract other expressions such as events and analysis methods as named entities, a model for named named entity extraction is prepared after preparing learning data for the expressions to be extracted. To create.

For example, the named entity extraction unit 1032 prepares learning data with a predetermined tag attached to the text data prepared in advance, and uses these to prepare a model for machine learning for extracting the named entity. Keep it. Then, after morphological analysis of the text data to be analyzed acquired by the text information acquisition unit 1031, the NER method is used using the created model. As a result, the named entity extraction unit 1032 extracts the named entity and then attaches one of the predetermined tags to the extracted named entity.

By extracting and tagging the named entity in the named entity extraction unit 1032, words having a particularly high possibility of being particularly meaningful are tagged from the text data acquired by the text information acquisition unit 1031 in advance. It becomes possible to extract.

The dependency extraction unit 1033 extracts the dependency. A commonly used dependency analyzer may be used for this.

Dependency refers to the relationship between two word pairs that have a relationship between a subject and a predicate, and a modifier and a modifier when parsing is performed. In the case of a subject and a predicate, the subject is the dependency source and the predicate is the dependency destination, and in the case of the modifier and the modifier, the modifier is the dependency source and the modified word is the dependency destination. In this way, the dependency consists of a pair of the dependency source and the dependency destination.

In a general dependency analyzer, the relationship between a subject and a predicate and the relationship between a modifier and a modifier are extracted as a dependency. The dependency extraction unit 1033 may extract both or either of these as a dependency.

Further, as a commonly used method of dependency analysis, there is also a method of analyzing clue expressions such as "because" and "by", and performing dependency extraction by focusing on the keywords before and after that. The dependency extraction unit 1033 may perform extraction using a clue expression and / or using a subject / predicate relationship and a modifier / modifier relationship.

Further, in the case of performing general dependency extraction, when W1, W2, W1W2 (for example, W1 is "precision", W2 is "machine", W1W2 is "precision machine", etc.) are named entity, W1- W2 may be extracted as a dependency. At this time, when the noun phrase is extracted as a dependency, the dependency may be analyzed by searching further ahead of the noun phrase and extracting the dependency such as W1-W3. The noun phrase here is composed of a plurality of words and has the same fluttering as one noun as a whole.

The dependency extraction unit 1033 analyzes the dependency on the text data acquired by the text information acquisition unit 1031, and then extracts the dependency including the named entity tagged by the named entity extraction unit 1032. At this time, it is more desirable to extract only the dependencies whose dependency source and dependency destination are both named entities.

By extracting the dependency in the dependency extraction unit 1033, it is possible to extract the relationship between words having some meaning such as cause and effect, purpose and means, event and analysis method, and the like.

In addition, by extracting only the dependencies that include named entities, in particular the dependencies whose both the dependency source and the recipient are named entities, it is possible to extract particularly important expressions in the text.

The dependency extraction unit 1033 uses the first word W1 as the dependency source when, for example, W1, W2, W3, and W4 are all unique expressions for the dependency including the extracted named entity. When there is a dependency between the first dependency W1-W3 and the second dependency W2-W4 whose dependency source is the second word W2, and the compound word W1W2 is also a named entity. , W1W2-W3, W1W2-W4, etc., with compound words that are W1W2 as the dependency source, and also for the first dependency dependency word W3 and the second dependency dependency word W4. Dependents may be added and expanded as if there were dependencies. A compound word is a word in which two or more originally independent words are combined to have a new meaning and function as one word.

The extension of the dependency in the dependency extraction unit 1033 is not limited to the compound word of two words, and may be extended to the compound word of three or more words.

Regarding the dependency in the named entity, in the compound word, by extending the dependency of the segmented named entity to the compound word, it becomes possible to appropriately grasp the content to be expressed in the text.

Further, the dependency extraction unit 1033 can also generate a dependency of a higher-level concept by extending the dependency by using a named entity.

The co-occurrence degree calculation unit 1034 analyzes the co-occurrence relationship in which the dependency appears in the sentence for each dependency of the dependencies extracted by the dependency extraction unit 1033, and the text data acquired by the text information acquisition unit 1031. Within, the number of sentences in which any of the dependencies appears and the number of sentences in which the co-occurrence of the dependency occurs are analyzed, and the ratio is calculated as the degree of co-occurrence. For this, a commonly used method such as the Jaccard index may be used.

The co-occurrence degree calculation unit 1034 may, for example, use W1-W2 or W1-W2 or from the text data acquired by the text information acquisition unit 1031 when the dependencies W1-W2 and W3-W4 are extracted by the dependency extraction unit 1033. The ratio of the number of sentences including the dependencies of W3-W4 and the sentences including the dependencies of W1-W2 and W3-W4 is calculated as the co-occurrence degree.

By calculating the degree of co-occurrence regarding the dependency by the co-occurrence degree calculation unit 1034, it is possible to analyze the combination of the dependencies having a high degree of co-occurrence.

The co-occurrence extraction unit 1035 extracts a combination of dependencies based on the degree of co-occurrence regarding the dependency.

The co-occurrence extraction unit 1035 may extract, for example, a combination of dependencies having a co-occurrence degree exceeding a threshold value with respect to the co-occurrence degree for any combination of dependencies calculated by the co-occurrence degree calculation unit 1034. good.

For example, if the co-occurrence extraction unit 1035 has a threshold value of T and the co-occurrence degree of the dependency Ki and Kj is Vij, the co-occurrence extraction unit 1035 may extract a combination of the dependency Ki and Kj such that Vij> T.

The co-occurrence extraction unit 1035 starts from the one with the higher co-occurrence degree with respect to the co-occurrence degree for any combination of dependencies calculated by the co-occurrence degree calculation unit 1034, starting from a certain selected dependency. The specified number, that is, the upper N pieces may be extracted in order.

For example, assuming that the degree of co-occurrence between the dependent Ki and Kj is Vij, Kj that is the top N cases may be extracted from the one with the highest Vij, starting from Ki.

The co-occurrence extraction unit 1035 pays attention to the co-occurrence degree with a specific tag including the hierarchical relationship with respect to the co-occurrence degree for any combination of dependencies calculated by the co-occurrence degree calculation unit 1034, for example. It may be extracted.

A combination of co-occurrence relationships that is particularly meaningful in text data by extracting the top N co-occurrence relationships that have a co-occurrence degree exceeding the threshold value or that start from a certain dependency by the co-occurrence extraction unit 1035. It becomes possible to extract relationships that are presumed to be particularly meaningful and important from a large amount of text data.

(Specific example of processing in named entity extraction unit 1032)
Hereinafter, a specific example of named entity extraction by the named entity extraction unit 1032 will be shown. For example, as shown in Fig. 3, there is a sentence "On June 29, 2006, Prime Minister Koizumi visited Washington and attended a summit meeting with President Bush." , "Place name", "quantity", "date", "time".

First, the named entity extraction unit 1032 performs morphological analysis and decomposes it into a sequence of morphemes. Then, the named entity extraction unit 1032 prepares a learning model for extracting tags of "person name", "place name", "quantity", "date", and "time" defined in advance. Then, the named entity extraction unit 1032 tags the sentences decomposed into columns of morphemes.

In the example of FIG. 3, "June 29, 2006" represents a date, "Koizumi" represents a person's name, "Washington" represents a place name, and "Bush" represents a person's name, so they are tagged respectively.

As shown in FIG. 3, the named entity extraction unit 1032 extracts a word to be a named entity and then tags it. This makes it possible to extract and classify words corresponding to tags whose named entities are defined in advance.

(Specific example of processing of dependency extraction unit 1033)
A specific example of the dependency extraction of the dependency extraction unit 1033 is shown. As shown in FIG. 4, for example, it is assumed that there are two sentences such as "a system stop occurs" and "a possibility of a power failure is inferred." The dependency extraction unit 1033 extracts the dependency. At this time, when the dependency is analyzed using a general dependency analyzer, "system-stop", "stop-occurrence", "power supply-abnormality", "power supply-possibility", "abnormality-possibility" , "Abnormality-guess" and "possibility-guess" can be extracted.

When the dependency extraction unit 1033 has a noun phrase, the dependency extraction unit 1033 extracts the dependency destination after that, so that the dependency of "power supply-abnormality" is skipped and only the "abnormality-possibility" is extracted.

The dependency extraction unit 1033 extracts only the dependencies with tags as named entity by the named entity extraction unit 1032.

For example, when there are two types of tags, "performance" and "state", as named entities, the sentence in FIG. 4 is "system" = performance, "stop" = state, "occurrence" = as shown in FIG. Tags such as state, "power" = performance, "abnormal" = state, "possibility" = state, "guess" = state are attached. In addition, the compound word "power supply error" can also be tagged as "power supply error" = status.

When the dependency extraction unit 1033 extracts only the dependencies with tags attached to both words of the dependency, both the dependency source and the dependency recipient are tagged as named entities, so that "system" Extract "stop", "stop-occurrence", "power supply-possibility", "abnormality-possibility", "abnormality-guess", and "possibility-guess".

Further, the dependency extraction unit 1033 has a relationship of W1-W3 and W2-W4, and when W1W2 is a named entity, it extends to the dependency of W1W2-W3 and W1W2-W4. Here, since there is a relationship of W1 = "power supply" and W2 = "abnormality", "power supply abnormality-possibility" and "power supply abnormality-guess" are extracted as dependencies.

(Specific processing in the co-occurrence degree calculation unit 1034)
The co-occurrence degree calculation unit 1034 calculates the co-occurrence degree for the dependency extracted by the dependency extraction unit 1033. For example, it is assumed that the dependency extraction unit 1033 extracts the dependency of W1-W2 and the dependency of W3-W4. At this time, the total number of sentences including the dependency of either or both of W1-W2 and W3-W4 is 100, and the total number of sentences including the dependency of both W1-W2 and W3-W4 is all. When there are 10 sentences in, the co-occurrence degree is (the number of sentences including the dependency of W1-W2 or / and W3-W4) / (the number of sentences including the dependency of W1-W2 and W3-W4) = Calculated as 0.1.

(Specific processing in the co-occurrence extraction unit 1035)
The co-occurrence extraction unit 1035 extracts a combination of dependencies having a co-occurrence degree exceeding a threshold value with respect to the co-occurrence degree for any combination of dependencies calculated by the co-occurrence degree calculation unit 1034.

For example, when there are dependencies K1, K2, and K3, the co-occurrence degree of K1 and K2 is 0.5, K1 and K3 are 0.1, and K2 and K3 are 0.2 by the co-occurrence degree calculation unit 1034. When calculated and the co-occurrence extraction unit 1035 is set to 0.3, the co-occurrence extraction unit 1035 extracts the co-occurrence relationship of K1-K2 having a co-occurrence degree exceeding the threshold.

(Modification example 1:)
Hereinafter, a modified example of extracting not only the co-occurrence of the dependency but also the co-occurrence of the dependency and the word or the co-occurrence of the word will be shown.

The co-occurrence degree calculation unit 1034 calculates the co-occurrence ratio of the dependencies extracted by the dependency extraction unit 1033 as the co-occurrence degree, and also the dependency extracted by the dependency extraction unit 1033 and the unique expression extraction unit. The degree of co-occurrence with the word extracted as the eigenexpression in 1032 and / or the degree of co-occurrence between the words extracted as the eigenexpression by the eigenexpression extraction unit 1032 are calculated.

The co-occurrence extraction unit 1035 exceeds the threshold value for the co-occurrence degree in any combination of dependencies, the combination of the dependency and the word, and / or the combination of words calculated by the co-occurrence degree calculation unit 1034. Extract the combinations of dependencies with co-occurrence, the combinations of dependencies and words, and / or the combinations of words. At this time, the threshold value may be set to a different value in the combination of the dependency, the combination of the dependency and the word, and the combination of the words.

The co-occurrence extraction unit 1035 assigns a specific tag to the tag assigned by the unique expression extraction unit 1032 when extracting the combination of the dependency, the combination of the dependency and the word, and / or the combination of the words. Co-occurrence relationships may be extracted in a limited manner.

By extracting the co-occurrence relationship as described above, in addition to the co-occurrence relationship of the dependency, the co-occurrence of the dependency and the word and / or the word can be extracted and used for analysis.

(Variation example 2: Analysis example of specialized literature)
The following is an example of modification when analyzing a particularly specialized document. In specialized literature, especially in the field of science and technology, in literature such as papers, events such as hypotheses, conclusions, assertions, states, performances, and properties, and analytical methods (experimental methods, inspection methods) to support them. Is often shown. Therefore, by extracting the event and the analysis method, it is possible to extract important points in the text. Therefore, two types of tags, "event" and "analysis", are prepared as tags indicating unique expressions. In addition, as a subclass of "event", for example, there are "case" and "site", and as a subclass of "analysis", there are "test item" and "test result", "event: case", Prepare tags for "event: site", "event: state", "analysis: inspection item", and "analysis: inspection result", and explain the following.

The text information acquisition unit 1031 acquires, for example, an article in the field of science and technology as text data. And, as an example sentence in it, as shown in FIG. 6, the example sentence "Yellow stain occurred in the conjunctiva of the eyeball. The possibility of jaundice symptom was inferred, and the abnormality of bilirubin was confirmed by the blood test." Specific processing is shown as such.

The named entity extraction unit 1032 acquires the named entity and tags it. For example, in the sentence shown in FIG. 6, as shown in FIG. 7, "eyeball" = event: site, "conjunctival" = event: site, "yellow dyeing" = event: state, "occurrence" = event: state, " "Yellow" = event: case, "symptom" = event: condition, "possibility" = event: condition, "guess" = event: condition, "blood" = event, site, "examination" = analysis: examination item, " Bilirubin ”= analysis: inspection item,“ abnormality ”= analysis: inspection result,“ confirmation ”= event: unique expression is extracted and tagged as a state.

Furthermore, the named entity extraction unit 1032 also tags compound words as "eyeball conjunctiva" = event: site, "jaundice symptom" = event: case, "blood test" = analysis: test item. .. FIG. 8 shows a specific example in which a compound word is also tagged.

The dependency extraction unit 1033 extracts the dependency. When the dependency including the named entity is analyzed for the sentence from which the named entity is extracted as shown in FIG. 7, the left side of FIG. 9 is obtained. That is, "eyeball-conjunctiva", "eyeball-yellow dyeing", "conjunctiva-yellow dyeing", "conjunctiva-development", "yellow dyeing-development", "occurrence-", "jaundice-symptom", "jaundice-" "Possibility", "Symptom-Possibility", "Symptom-Guess", "Possibility-Guess", "Blood-Test", "Blood-Abnormal", "Test-Abnormal", "Test-Confirmation", "Bilirubin" -Extract "abnormal", "bilirubin-confirmation", "abnormal-confirmation", and "confirmation-confirmed" as dependencies.

When the dependency extraction unit 1033 has a noun phrase, it searches for and extracts the dependency beyond it, so that "eyeball-conjunctiva", "jaundice-symptom", and "blood-test" are not extracted. Extract the dependencies as shown on the right side of FIG.

Further, the dependency extraction unit 1033 extracts only the dependencies tagged as the named entity by the named entity extraction unit 1032 by both the dependency source and the dependent recipient. That is, "eyeball-conjunctiva", "eyeball-yellow dyeing", "conjunctiva-yellow dyeing", "conjunctiva-development", "yellow dyeing-development", "jaundice-symptom" circled on the right side of FIG. , "Jaundice-possibility", "symptom-possibility", "symptom-guess", "possibility-guess", "blood-test", "blood-abnormal", "test-abnormal", "test-confirmation" , "Bilirubin-abnormal", "bilirubin-confirmation", "abnormal-confirmation" are extracted as dependencies.

The dependency extraction unit 1033 has a relationship of W1-W3 and W2-W4, and when W1W2 is a named entity, it extends to the dependency of W1W2-W3 and W1W2-W4. For example, "eyeball-yellow dyeing", "conjunctiva-yellow dyeing", and "conjunctiva-development" become "eyeball conjunctiva-yellow dyeing" and "eyeball conjunctiva-development", and "jaundice-possibility", "symptom-possibility", and "symptom". -Inference" becomes "jaundice symptom-possibility" and "jaundice symptom-inference", and "blood-abnormality" "test-abnormality" "test-confirmation" becomes "blood test-abnormality" "blood test-confirmation". That is, the dependency extraction unit 1033 extracts the dependency shown on the lower right side of FIG. 7.

The co-occurrence degree calculation unit 1034 calculates the co-occurrence degree by using a commonly used method such as a Jaccard index. In this modification, the degree of co-occurrence is calculated from the viewpoints related to events and the viewpoints related to analysis.

First, from the viewpoint of an event, the degree of co-occurrence is extracted for the dependency in which the dependency source is an event and the dependency destination is also an event. That is, when there is a dependency of W1-W2 and a dependency of W3-W4 regarding an event, the number of sentences including both W1-W2 and W3-W4 is determined by the dependency of W1-W2 and / or W3-W4. The degree of co-occurrence is calculated by dividing (dividing) by the number of sentences.

Next, from the viewpoint of analysis, both the dependency source and the dependency destination may calculate the co-occurrence degree for the combination with the word whose tag is the analysis, in addition to the dependency whose tag is the event. That is, when there is a word of the dependency with the event tag of W1-W2 and the analysis tag of B1, the number of sentences including both the dependency of W1-W2 and the sentence including B1 is the dependency of W1-W2. The degree of co-occurrence is calculated by dividing (dividing) by the number of sentences containing the word of receiving and / or B1.

For example, when there is a word in the text with a dependency on the event of W1-W2, W3-W4 and an analysis tag of B1, B2, (W1-W2, B1) (W1-W2, B2) ( The degree of co-occurrence is calculated for any combination of W3-W4, B1) and (W3-W4, B2).

The co-occurrence extraction unit 1035 extracts combinations in which the degree of co-occurrence exceeds the threshold value from the viewpoint of the event and the viewpoint of analysis.

When the text has a tag having a hierarchical structure, the co-occurrence extraction unit 1035 may extract by paying attention to the degree of co-occurrence with a specific tag including the hierarchical relationship.

(Process flow)
FIG. 10 shows a flow of processing of the information analysis device according to the first embodiment of the present disclosure.

The text information acquisition unit 1031 acquires the text data to be analyzed (step S1031).

The named entity extraction unit 1032 extracts the named entity and attaches a tag to the text data acquired by the text information acquisition unit 1031 (step S1032). The type of tag may be determined in advance.

The dependency extraction unit 1033 analyzes the dependency on the text data from which the named entity is extracted (step S1033).

In extracting the dependency, when the noun phrase is extracted as the dependency, the dependency may be extracted by further searching the tip of the sentence. Further, only the dependency of the relationship in which the word tagged as the named entity by the named entity extraction unit 1032 is the dependency source and the dependency destination may be extracted.

The co-occurrence degree calculation unit 1034 calculates the co-occurrence degree for the extracted dependent words (step S1034).

The co-occurrence extraction unit 1035 extracts a combination of dependencies having a co-occurrence degree exceeding the threshold value (step S1035).

(effect)
According to the present embodiment, as a premise for the dependency analysis, by extracting the named entity and extracting only the named entity, the more important words, especially the words that are keywords in the field of science and technology, are extracted. It is possible to extract dependencies.

In addition, by calculating the dependency and the degree of co-occurrence and extracting the dependency with a high degree of co-occurrence, a causal relationship can be obtained from the acquired literature even for contents including technical terms. And / or it becomes possible to analyze the relationship between the event and the analysis method.

<Embodiment 2>
The second embodiment further includes a function of visually displaying the analysis performed on the acquired text information to the user.

<Configuration of information analysis device>
FIG. 11 shows the configuration of the server 20 in this embodiment. The control unit 203 of the server 20 includes a text information acquisition unit 1031, a named entity extraction unit 1032, a dependency extraction unit 1033, a co-occurrence degree calculation unit 1034, a co-occurrence extraction unit 1035, and a display unit 2036. The text information acquisition unit 1031, the named entity extraction unit 1032, the dependency extraction unit 1033, the co-occurrence degree calculation unit 1034, and the co-occurrence extraction unit 1035 are the same as the control unit 103 of the server 10.

The display unit 2036 visually expresses and displays the combination of dependencies analyzed by the co-occurrence extraction unit 1035 to the user. Specifically, after displaying the word of the dependency source and the word of the dependency destination, the dependency may be connected by a line and expressed in the form of a network diagram.

The display unit 2036 may visually represent and display the combination of the word dependency relationship tagged with the event tag and the word tagged with the analysis tag to the user.

When a tag having a text hierarchical structure is attached, the display unit 2036 may display a combination of dependencies including the hierarchical structure.

The display unit 2036 may be expressed by changing the thickness of the line or changing the color of the network diagram according to the magnitude of the co-occurrence degree. Further, the display unit 2036 may also indicate the degree of co-occurrence.

(Concrete example)
FIG. 12 shows a specific example of the visual expression of the dependency displayed by the display unit 2036. The display unit 2036 visually expresses and displays the dependencies analyzed by the dependency extraction unit 1033, for example, by using a representation of a network diagram. In the example of FIG. 12, for the event, the combination of dependencies such as "system-stop", "power supply-possibility", and "power supply-abnormality", and these dependencies, "visual confirmation of parts", "condenser", and "cooling device" For combinations with words of analysis methods such as "voltage measurement" and "tester", those having a co-occurrence degree exceeding the threshold value are connected by a line and displayed. In the example of FIG. 12, the dependency relationship of the event is displayed as a white background, and the word of the analysis method is displayed as a black background.

The display unit 2036 may display the co-occurrence extraction unit 1035 in consideration of the hierarchical structure of the tags when the co-occurrence extraction unit 1035 extracts the co-occurrence relationship in consideration of the hierarchical structure of the tags.

For example, regarding the analysis method tag, there is a tag of "analysis: cause identification method", which is attached to "visual confirmation of parts" and "voltage measurement", and there is a tag of "analysis: object", and "capacitor" and "condenser". It is attached to "cooling device", has a tag of "analysis: measuring instrument", and is attached to "tester". Here, when the co-occurrence relationship is extracted by focusing on "analysis: cause identification method", the relationship between the co-occurrence relationship in the event and "analysis: cause identification method" is extracted and illustrated.

FIG. 13 shows a specific example of the visual expression of the dependency displayed by the display unit 2036 when the co-occurrence relationship is extracted by focusing on “analysis: method for identifying the cause”. Compared with FIG. 12, since only the co-occurrence relationship of interest is extracted, it becomes easier to grasp the relationship.

Since FIG. 13 focuses on "analysis: method for identifying the cause", only the co-occurrence relationship between the dependency and the word with such a tag is extracted. On the other hand, since "capacitor", "cooling device", and "tester" have a co-occurrence relationship with the words "visual confirmation of parts" or "voltage measurement", the co-occurrence relationship is extracted in connection with them. , Illustrated.

The display unit 2036 may display the co-occurrence degree calculated by the co-occurrence degree calculation unit 1034 together with the network diagram.

In this way, by using the visual expression, it is possible to make the user understand not only the dependency but also the relationship of important words in a visually easy-to-understand manner.

(Process flow)
FIG. 14 shows a flow of processing of the information analysis apparatus according to the second embodiment of the present disclosure.

The text information acquisition unit 1031 acquires the text data to be analyzed (step S1031).

The named entity extraction unit 1032 extracts the named entity and attaches a tag to the text data acquired by the text information acquisition unit 1031 (step S1032). The type of tag may be determined in advance.

The dependency extraction unit 1033 extracts the dependency from the text data from which the named entity is extracted (step S1033).

In extracting the dependency, when the noun phrase is extracted as the dependency, the dependency may be extracted by further searching the tip of the sentence. Further, only the dependency of the relationship in which the word tagged as the named entity by the named entity extraction unit 1032 is the dependency source and the dependency destination may be extracted.

The co-occurrence degree calculation unit 1034 calculates the co-occurrence degree for the extracted dependent words (step S1034).

The co-occurrence extraction unit 1035 extracts a combination of dependencies having a co-occurrence degree exceeding the threshold value (step S1035).

The display unit 2036 causes the user to display the dependency by using a visual expression (step S2036). At this time, the display unit 2036 may also display the co-occurrence degree, and may change the display method according to the co-occurrence degree.

(effect)
According to the present embodiment, displaying the extracted dependencies in consideration of the degree of co-occurrence and the like helps the user to accurately grasp important words.

Especially in specialized fields and literature such as science and technology, there are many technical terms, so it was difficult to extract important or meaningful words with the natural language processing methods that have been commonly used so far, but this is accurate. It is possible to express after extracting to.

This is the end of the description of the embodiment, but the above embodiment is only an example. Therefore, the specific configurations, processing contents, and the like of the servers 10 and 20 are not limited to those described in the above embodiment.

The information analysis device according to the present disclosure is not limited to being realized on a computer that operates standalone, for example, and may be operated as, for example, a server-type computer.

For example, as shown in FIG. 15, the server 10 and the user terminal 30 operated by the user are connected via a network, operated from the user terminal 30, and the server 10 analyzes a document, and the result is analyzed by the user terminal. It may be displayed at 30.

Further, the information analysis device according to the present disclosure may realize its function by, for example, a computer executing a program, regardless of the above device. The program to realize the function of the information analysis device is computer reading of USB (Universal Serial Bus) memory, CD-ROM (Compact Disc-Read Only Memory), DVD (Digital Versatile Disc), HDD (Hard Disc Drive), etc. It may be stored on a possible recording medium or downloaded to a computer via a network.

Although the preferred embodiments of the present disclosure have been described above, the present disclosure is not limited to such specific embodiments, and the present disclosure includes the inventions described in the claims and the equivalent scope thereof. Is done. Further, the configurations of the devices described in the above-described embodiments and modifications can be appropriately combined as long as there is no technical contradiction.

10, 20 ... server, 11 ... processor, 12 ... memory, 13 ... storage, 14 ... communication IF, 15 ... input / output IF, 30 ... user terminal, 101 ... communication unit, 102 ... storage unit, 103 ... control unit, 1031 ... Text information acquisition unit, 1032 ... Unique expression extraction unit, 1033 ... Dependency extraction unit, 1034 ... Co-occurrence extraction unit, 1035 ... Co-occurrence extraction unit, 2036 ... Display unit, 211 ... Specific example to be displayed by the display unit

Claims (13)

A program that causes a computer equipped with a processor to perform processing, and the processor
The text information acquisition step to acquire text information and
A named entity extraction step that extracts a named entity from the text information and assigns a tag to it.
A dependency extraction step of analyzing a dependency from the text information and extracting a dependency for which one or both of the dependency source and the dependency destination are unique expressions.
In the dependency extracted in the dependency extraction step from the text information, the co-occurrence degree calculation step for calculating the co-occurrence degree for the combination of the dependencies, and the co-occurrence degree calculation step.
A co-occurrence extraction step that extracts a combination of dependencies based on the degree of co-occurrence,
To execute,
In the dependency extraction step, in the extracted dependency, the first word, the second word, the third word, and the fourth word are unique expressions, and the first word is the dependency source. There is a dependency of the third word as a dependency, a dependency of the second word as a dependency source, and a dependency of the fourth word as a dependency, and the first word. When the compound word of the second word and the second word is a proper expression, the third dependency with the compound word as the dependency source and the third word as the dependency destination, and the compound word are involved. A program that adds and extends the fourth dependency, which is the receiver and the dependency destination is the fourth word , to the extracted dependency . The program according to claim 1 , wherein the tag includes a tag having a hierarchical structure. The program according to claim 1 or 2 , wherein the named entity and the tag include at least an event tag representing an event and an analysis tag representing an analysis method for an event. The named entity extraction step is one of claims 1 to 3 , wherein the named entity is extracted from the text information by using machine learning using the model for extracting the named entity, and the tag is attached to the text information. The described program. The program according to any one of claims 1 to 4 , wherein the dependency extraction step analyzes the relationship between the modifier and the modified word as a dependency. The program according to claim 5 , wherein the dependency extraction step analyzes a word analyzed by a predetermined clue expression as a dependency in addition to the relationship between the modifier and the modified word. The program according to any one of claims 1 to 6 , wherein the dependency extraction step searches further ahead of the noun phrase and extracts the dependency when the noun phrase is extracted as the dependency. The program according to any one of claims 1 to 7 , wherein the co-occurrence degree calculation step calculates the co-occurrence degree for the combination of the dependency, the combination of the dependency and the word, and / or the combination of the word and the word. .. The program according to any one of claims 1 to 8 , wherein the co-occurrence extraction step extracts a combination of dependencies whose co-occurrence degree exceeds a threshold value. The program according to any one of claims 1 to 9 , wherein the co-occurrence extraction step extracts a specified number of dependency combinations from the one having the highest degree of co-occurrence with the selected dependency. The tag includes a tag having a hierarchical structure, and the co-occurrence extraction step is one of claims 1 to 10 for extracting a combination of dependencies according to the degree of co-occurrence with the selected tag. The program described. A method for causing a computer equipped with a processor to perform processing, wherein the processor is capable of performing processing.
The text information acquisition step to acquire text information and
A named entity extraction step that extracts a named entity from the text information and assigns a tag to it.
A dependency extraction step of analyzing a dependency from the text information and extracting a dependency for which one or both of the dependency source and the dependency destination are unique expressions.
In the dependency extracted in the dependency extraction step from the text information, the co-occurrence degree calculation step for calculating the co-occurrence degree for the combination of the dependencies, and the co-occurrence degree calculation step.
A co-occurrence extraction step that extracts a combination of dependencies based on the degree of co-occurrence,
And run
In the dependency extraction step, in the extracted dependency, the first word, the second word, the third word, and the fourth word are unique expressions, and the first word is the dependency source. There is a dependency of the third word as a dependency, a dependency of the second word as a dependency source, and a dependency of the fourth word as a dependency, and the first word. When the compound word of the second word and the second word is a proper expression, the third dependency with the compound word as the dependency source and the third word as the dependency destination, and the compound word are involved. The fourth dependency, which is the receiver and the dependency destination is the fourth word, is expanded by adding it to the extracted dependency.
Method. An information analysis device including a control unit, wherein the control unit
The text information acquisition step to acquire text information and
A named entity extraction step that extracts a named entity from the text information and assigns a tag to it.
A dependency extraction step of analyzing a dependency from the text information and extracting a dependency for which one or both of the dependency source and the dependency destination are unique expressions.
In the dependency extracted in the dependency extraction step from the text information, the co-occurrence degree calculation step for calculating the co-occurrence degree for the combination of the dependencies, and the co-occurrence degree calculation step.
A co-occurrence extraction step that extracts a combination of dependencies based on the degree of co-occurrence,
And run
In the dependency extraction step, in the extracted dependency, the first word, the second word, the third word, and the fourth word are unique expressions, and the first word is the dependency source. There is a dependency of the third word as a dependency, a dependency of the second word as a dependency source, and a dependency of the fourth word as a dependency, and the first word. When the compound word of the second word and the second word is a proper expression, the third dependency with the compound word as the dependency source and the third word as the dependency destination, and the compound word are involved. The fourth dependency, which is the receiver and the dependency destination is the fourth word, is expanded by adding it to the extracted dependency.
Information analysis device.

Images