Information processing program, information processing device, and information processing method

Abstract

To present a result of analyzing a plurality of documents, in an easily understandable manner to a user.SOLUTION: An information processing device for conducting processing to analyze the contents of a plurality of documents comprises: a document acquisition section for acquiring a plurality of documents as text data respectively; an appearance frequency calculation section for calculating an appearance frequency of a word in the plurality of documents; an emotion value calculating section for calculating an emotion value of a word in the plurality of documents; a word map generation section for generating and outputting a word map in which each word is plotted on a two-dimensional map by using the appearance frequency and the emotion value respectively as axes, on the basis of the appearance frequency and the emotion value of each word; and a display section for displaying a display screen containing the generated word map in a display for displaying.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to an information processing program, an information processing apparatus, and an information processing method for presenting the results of analyzing a plurality of documents to a user in an easily understandable manner. [Background technology]

[0002] Conventionally, there are technologies that automatically analyze and classify multiple documents and provide them to users. There is a strong need to know the analysis and classification results for a large number of documents, and various analysis and classification methods have been proposed to meet such needs.

[0003] For example, Patent Document 1 discloses a technique relating to a document classification model based on the frequency of occurrence of a group of words that takes into account similarities between words, rather than just the frequency of occurrence of individual words. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2021-140497 A Summary of the Invention [Problem to be solved by the invention]

[0005] By applying the technology described in Patent Document 1 to multiple documents that need to be analyzed, such as multiple review documents for each product on a product sales site, multiple word-of-mouth documents for restaurants, and multiple documents that are text versions of inquiries received at a call center, it is possible to extract frequently occurring words and classify documents based on their occurrence frequency, but no means is disclosed for providing such analysis and classification results to users in an easy-to-understand manner.

[0006] There was a need for a method to provide users with an easy-to-understand view of overall trends across multiple documents and the presence of noteworthy words.

[0007] The present invention has been made in consideration of the above problems, and aims to provide an information processing program, an information processing device, and an information processing method that are capable of presenting the results of analyzing multiple documents to a user in an easy-to-understand manner. [Means for solving the problem]

[0008] The information processing program of the present invention is an information processing program for causing a computer to perform a process of analyzing the contents of multiple documents, and is characterized in that the computer is made to perform a document acquisition function for acquiring each of the multiple documents as text data, an occurrence frequency calculation function for calculating the occurrence frequency of words in the multiple documents, an attention calculation function for calculating the attention level of words in the multiple documents, a word map generation function for generating and outputting a word map in which each word is plotted on a two-dimensional map with the occurrence frequency and the attention level as axes based on the occurrence frequency and the attention level of each word, and a display function for displaying a display screen including the generated word map on a display.

[0009] In addition, the information processing program of the present invention further causes the computer to realize a related word identification function that identifies other words that are highly related to each word in the multiple documents, and the display function is characterized in that when a specified operation is performed on any word in the word map on the display screen, related words of the word are derived from the display position of the word and displayed.

[0010] In addition, the information processing program of the present invention further realizes a document extraction function that, when a specified operation is performed on any of the related words derived and displayed from the display position of the word, extracts documents that contain both the word and its related words from the multiple documents, and the display function is characterized in that it displays the documents that contain both the word and its related words on a display screen.

[0011] In addition, the information processing program of the present invention is further characterized in that the display function displays words whose occurrence frequency and / or attention level are above a predetermined threshold in a different display manner than words whose occurrence frequency and / or attention level are below the threshold.

[0012] In addition, the information processing program of the present invention is further characterized in that the attention calculation function classifies the multiple documents into multiple clusters based on predetermined classification rules, and calculates the attention of words by applying the TF-IDF method on a cluster basis.

[0013] The information processing device of the present invention is an information processing device for performing processing to analyze the contents of multiple documents, and is characterized in that it comprises a document acquisition unit that acquires each of the multiple documents as text data, an occurrence frequency calculation unit that calculates the occurrence frequency of words in the multiple documents, an attention calculation unit that calculates the attention level of words in the multiple documents, a word map generation unit that generates and outputs a word map in which each word is plotted on a two-dimensional map with the occurrence frequency and the attention level as axes based on the occurrence frequency and the attention level of each word, and a display unit that displays a display screen including the generated word map on a display for display.

[0014] The information processing method of the present invention is an information processing method for performing processing for analyzing the contents of multiple documents by a computer, and is characterized in that it includes a document acquisition step of acquiring each of the multiple documents as text data, an occurrence frequency calculation step of calculating the occurrence frequency of words in the multiple documents, an attention calculation step of calculating the attention level of words in the multiple documents, a word map generation step of generating and outputting a word map in which each word is plotted on a two-dimensional map with the occurrence frequency and the attention level as axes based on the occurrence frequency and the attention level of each word, and a display step of displaying a display screen including the generated word map on a display for display. Effect of the Invention

[0015] According to the present invention, a word map is generated by plotting each word on a two-dimensional map with the frequency of occurrence and the attention level as axes, and a display screen including the generated word map is displayed on a display for display, so that the results of analyzing a plurality of documents can be presented in a manner that makes it easy for the user to visually understand them. In addition, when a predetermined operation is performed on any word in the word map on the display screen, related words of the word are derived from the display position of the word and displayed, so that it is possible to easily understand what words each word on the word map is used with from the display of the related words. In addition, when a predetermined operation is performed on any of the related words derived from the display position of a word and displayed, documents including both the word and the related words are extracted from multiple documents and displayed on the display screen, so that it is possible to easily check how the derived word and the related words were used in the actual document. [Brief description of the drawings]

[0016] [Figure 1] 1 is a block diagram showing the configuration of an entire system for realizing an information processing device according to the present invention; [Diagram 2] FIG. 1 is a block diagram showing an example of a configuration of an information processing device (server device) 10 according to the present invention. [Diagram 3] FIG. 2 is a flowchart showing the flow of a word map generation process in the information processing device 10 according to the present invention. [Figure 4] 1 is a flowchart showing the flow of related word display processing in an information processing device 10 according to the present invention. [Diagram 5] FIG. 2 is a flowchart showing the flow of document extraction processing in the information processing device 10 according to the present invention. [Figure 6] FIG. 2 is an explanatory diagram showing an example of how frequently occurring words and highly attention-grabbing words are extracted in the information processing device 10 according to the present invention. [Figure 7] FIG. 2 is an explanatory diagram showing an example of a word map generated based on appearance frequency and attention level in the information processing device 10 according to the present invention. [Figure 8] FIG. 2 is an explanatory diagram showing an example of how related words are derived from an original word and displayed in the information processing device 10 according to the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0017] [First embodiment] Hereinafter, an example of an information processing device according to a first embodiment will be described with reference to the drawings. FIG. 1 is a block diagram showing the configuration of an entire system for realizing an information processing device according to the present invention. The information processing device 10 may be a device designed as a dedicated machine, but is assumed to be realizable by a general computer or server device. In this case, the information processing device 10 is assumed to be equipped with a CPU (Central Processing Unit), memory, hard disk drive, and other storage devices that would normally be equipped in a general computer or server device (not shown). It goes without saying that various processes are executed by a program to make these general computers and server devices function as the information processing device 10 of this example.

[0018] 1, the overall system configuration for realizing the information processing device includes a server device 10 and user terminals 201 to 20n (hereinafter, including cases where these are collectively referred to as user terminals 20) that are mutually connectable via a communication network 30. Of these, the server device 10 functions as the information processing device 10 of this example, and a case will be described in which the information processing device (server device) 10 is accessed from each of the user terminals 201 to 20n to use the system.

[0019] Fig. 2 is a block diagram showing an example of the configuration of an information processing device (server device) 10 according to the present invention. As shown in Fig. 2, the information processing device 10 includes at least a document acquisition unit 11, an appearance frequency calculation unit 12, an attention degree calculation unit 13, a related word specification unit 14, a document extraction unit 15, a word map generation unit 16, a display unit 17, and a storage unit 18.

[0020] The document acquisition unit 11 has a function of acquiring a plurality of documents as text data. Any type of document may be acquired by the document acquisition unit 11, and examples include a plurality of review documents for each product on a product sales site, a plurality of word-of-mouth documents for restaurants, and a plurality of documents in which consultation details received at a call center have been converted into text. Note that the document acquisition unit 11 may acquire a plurality of text data as a plurality of documents by providing the information processing device of this example with a function of acquiring voice data and converting it into text data.

[0021] The occurrence frequency calculation unit 12 has a function of calculating the occurrence frequency of a word in a plurality of documents. Here, the occurrence frequency refers to the count of the number of occurrences of each word in all the documents in a predetermined category, or the score (normalized) of the number of occurrences. The range for calculating the occurrence frequency may be the entire acquired document, or each category after the acquired document is classified into a plurality of categories based on a predetermined classification rule. An example of the predetermined classification rule is LDA (Latent Dirichlet Allocation). The calculated information on the occurrence frequency of each word is stored in the storage unit 18 described later. In addition, when the acquired documents are classified into a plurality of categories based on the predetermined classification rule, the information on the category to which each document belongs is also stored in the storage unit 18 described later.

[0022] In addition, in the occurrence frequency calculation unit 12, when calculating the occurrence frequency of each word, it is necessary to perform a pre-processing to divide the text into predetermined units such as nouns, verbs, and adjectives based on a predetermined rule. In addition, when calculating the occurrence frequency, it is also possible to calculate the occurrence frequency by treating several consecutive words as one group. As an example, a method is conceivable in which a text is divided into subwords (groups of several consecutive words) using byte pair encoding, and then subwords with high occurrence frequency are treated as one word to calculate the occurrence frequency, and subwords with low occurrence frequency are divided into shorter units to calculate the occurrence frequency. It can be said that the document analysis result can be conveyed to the user more easily if the subwords with high occurrence frequency are treated as one word and presented to the user.

[0023] The attention calculation unit 13 has a function of calculating the attention of words in multiple documents. Here, the attention refers to a score that indicates the degree to which a word is likely to characterize the entire document or a category. Any method may be used as long as it is possible to calculate the attention of each word. For example, it is possible to calculate the attention of a word by applying the TF-IDF method to multiple documents. In addition, by applying the TF-IDF method to each category after classifying the acquired documents into multiple categories based on a predetermined classification rule, it is possible to obtain an effect that it becomes easier to extract highly attention-grabbing words that are different from frequently occurring words.

[0024] The related word identification unit 14 has a function of identifying other words (hereinafter referred to as related words) that are highly related to each word in a plurality of documents. Any method may be used as long as it is possible to extract related words, but for example, it is possible to extract related words based on the degree of collocation. Information on the relationship between the identified words and the related words is stored in the storage unit 18, which will be described later.

[0025] The document extraction unit 15 has a function of extracting documents that contain both a word and its related words from multiple documents when a combination of the word and its related words is specified. When a combination of a word and its related words is specified, the document extraction unit 15 executes a process of extracting documents that contain both of the words from multiple documents.

[0026] The word map generating unit 16 has a function of generating and outputting a word map in which each word is plotted on a two-dimensional map with the occurrence frequency and the attention level as axes based on the occurrence frequency and the attention level of each word. The word map is a display of words with high occurrence frequency or high attention level in a plurality of documents arranged on a two-dimensional map, and can be used to present the analysis results of a plurality of documents to a user in a visually easy-to-understand manner. This word map may be generated for all the acquired documents, or may be generated for each category after the acquired documents are classified into a plurality of categories based on a predetermined classification rule. The occurrence frequency and attention level information used on the vertical and horizontal axes may be count numbers such as the number of occurrences, or may be scored (normalized) numerical values. In addition, instead of plotting all the words on the word map, threshold values ​​may be set for the occurrence frequency and attention level, and words with occurrence frequency or attention level equal to or higher than the threshold value may be plotted on the word map. Furthermore, because the word map is a means of visually communicating analysis results to the user in an easily understandable manner, the plot points do not necessarily need to be in exact positions, and may be positioned with some deviation within a specified range, for example to avoid overlapping characters.

[0027] The display unit 17 has a function of displaying a display screen including the generated word map on a display for display. Here, in a case where the user terminal 20 is configured to have a display for display, the display screen generated by the display unit 17 is transmitted to the user terminal 20 via the communication network 30 so as to be displayed on the display for display of the user terminal 20.

[0028] Moreover, the display unit 17 has a function of, when a predetermined operation is performed on any word in the word map on the display screen, deriving and displaying related words of the word from the display position of the word. Any predetermined operation on a word may be performed, and an example is an operation of hovering a pointer over a word on the word map. When a word on which a predetermined operation is performed is specified, the related word specification unit 14 refers to information on related words of each word that has been specified and stored in advance, reads out information on related words of the word, and derives and displays the related words from the display position of the word. The derived display may be any display that can show the relationship between the original word and the related words, and an example is a method of displaying related words by connecting them with edges like a tree structure from the display position (node) of the original word. That is, a display process may be performed in which the related words are displayed for the first time when the mouse pointer is hovered over a word, and the display of the related words disappears when the mouse pointer is removed.

[0029] In addition, the display unit 17 has a function of displaying a document including both the word and the related word on the display screen when a predetermined operation is performed on any of the related words derived from the display position of the word. The predetermined operation on the related word may be any type, and an example is a single click operation on the related word on the word map. When the related word on which the predetermined operation is performed is specified, the document extraction unit 15 extracts documents including both the original word and the related word and displays them on the display screen. The method of displaying documents including both the original word and the related word may be any type, and an example is that only two or three examples may be displayed instead of all the relevant documents, or that only parts including both the original word and the related word are excerpted and displayed instead of the entire document.

[0030] The display unit 17 may display words whose frequency of occurrence and / or attention level are equal to or higher than a predetermined threshold in a different display manner from words whose frequency of occurrence and / or attention level are lower than the threshold. In other words, it is possible to allow the user to distinguish between words whose frequency of occurrence and words whose attention level are high. Examples of a method for changing the display manner include changing the font size and the display color.

[0031] The storage unit 18 has a function of storing data required for various processes performed in the information processing device 10 and data obtained as a result of the processes.

[0032] Next, the flow of various processes in the information processing device 10 according to the present invention will be described. FIG. 3 is a flow chart showing the flow of the word map generation process in the information processing device 10 according to the present invention. As shown in FIG. 3, the word map generation process in the information processing device 10 is started by acquiring a plurality of documents in the information processing device 10 (step S101). Next, the information processing device 10 calculates the occurrence frequency of each word in the acquired plurality of documents (step S102). Next, the information processing device 10 calculates the attention degree of each word in the acquired plurality of documents (step S103). Next, the information processing device 10 identifies related words of each word in the acquired plurality of documents (step S104). Next, the information processing device 10 generates a word map by plotting words on a two-dimensional map with the occurrence frequency and the attention degree as axes (step S105). Then, the information processing device 10 causes a display screen including the generated word map to be displayed on the display (transmits the display screen to the user terminal) (step S106), and ends the word map generation process.

[0033] FIG. 4 is a flow chart showing the flow of the related word display process in the information processing device 10 according to the present invention. As shown in FIG. 4, the related word display process in the information processing device 10 is started by receiving operation information of a predetermined operation (e.g., mouse over) on a word on a word map being displayed on the information processing device 10 (step S201). Next, the information processing device 10 reads out related words of the word on which the predetermined operation was performed (step S202). Next, the information processing device 10 generates a display screen on which related words are derived from the word on the word map and displayed (step S203). Then, the information processing device 10 displays the generated display screen on the display (transmits the display screen to the user terminal) (step S204), and ends the related word display process. Note that the flow chart shown in FIG. 4 is an example of the related word display process, and any internal process may be used as long as related words can be derived and displayed when a predetermined operation is performed on a word on a word map being displayed. For example, in step S105 in the flowchart shown in FIG. 3, a word map may be generated in which related words of a word are embedded, and when a specified operation is performed on a word on the displayed word map, the embedded related words may be displayed immediately.

[0034] Fig. 5 is a flow chart showing the flow of document extraction processing in the information processing device 10 according to the present invention. As shown in Fig. 5, the document extraction processing in the information processing device 10 is started by receiving operation information of a predetermined operation (e.g., single click) on a related word on a word map being displayed on the information processing device 10 (step S301). Next, the information processing device 10 extracts documents including both the derived word and the related word (step S302). Then, the information processing device 10 displays a display screen showing the extracted document on a display (transmits the display screen to a user terminal) (step S303), and ends the document extraction processing.

[0035] Next, the flow of document analysis will be explained with a concrete example. An example will be explained in which the documents to be analyzed are multiple reviews on a review site about instant cup noodles manufactured by one company. First, as an example, 1000 documents are acquired as text data in a situation where there are a total of 1000 reviews. Next, the 1000 documents are classified into multiple categories by applying LDA (Latent Dirichlet Allocation) as a predetermined classification rule. As a result, it is assumed that the 1000 documents are classified into five categories, category 1 to category 5. Next, the frequency of occurrence of words and the attention degree of words are calculated for each category.

[0036] Fig. 6 is an explanatory diagram showing an example of how frequently occurring words and highly attention-grabbing words are extracted in the information processing device 10 according to the present invention. Fig. 6 shows an example of frequently occurring words and attention-grabbing words extracted for each of five categories, category 1 to category 5. In this way, it is possible to extract frequently occurring words and attention-grabbing words for each category and display them in a list, but the system is also provided with a function to generate and display a word map based on this.

[0037] FIG. 7 is an explanatory diagram showing an example of a word map generated based on the occurrence frequency and attention level in the information processing device 10 according to the present invention. FIG. 7 shows a word map generated based on the occurrence frequency and attention level in category 1 in FIG. 6. In this way, by plotting each word on a two-dimensional map with the occurrence frequency and attention level as axes, rather than displaying them in a list, it becomes easier to visually grasp the analysis results of a document. For example, the analysis results of each word can be immediately grasped based on the display position on the word map, such as that this word has a high occurrence frequency but a low attention level, or that this word has a medium occurrence frequency but a very high attention level.

[0038] FIG. 8 is an explanatory diagram showing an example of how related words are derived from an original word and displayed in the information processing device 10 according to the present invention. Suppose a user who has seen the word map shown in FIG. 7 is interested in the word "smell". In this case, when the user places the mouse pointer over the display area of ​​the word "smell", "open", "plastic", "oil", and "foul smell" are displayed as related words of "smell", as shown in FIG. 8. These related words are words that were used in the same context as "smell", so by simultaneously displaying the derived word and the related words, such as "the smell at the moment of opening the cup noodles", "the smell of plastic", "the smell of oil", and "I felt it was a foul smell", it is possible to instantly guess the intention of use of the original document, and an effect is obtained that the user can easily grasp the tendency of the document analysis result. In addition, by single-clicking on this related word, the original document containing both the derived word and the related word can be displayed, so that the intention of use of the guessed related word can be easily confirmed in the original document.

[0039] As described above, according to the information processing device of this example, a plurality of documents are each acquired as text data, the frequency of occurrence of words in the plurality of documents is calculated, the attention level of words in the plurality of documents is calculated, and based on the frequency of occurrence and attention level of each word, a word map is generated in which each word is plotted on a two-dimensional map with the frequency of occurrence and attention level as axes, and a display screen including the generated word map is displayed on a display for display. This makes it possible to present the results of analysis of a plurality of documents to a user in an easy-to-understand manner.

[0040] Furthermore, when a specified operation is performed on any word in the word map on the display screen, the related words of that word are derived and displayed based on the display position of that word, so that it is possible to easily grasp from the display of the related words what words each word on the word map is used with.

[0041] Furthermore, when a specified operation is performed on any of the related words derived and displayed from the display position of a word, documents containing both the word and its related words are extracted from multiple documents and displayed on the display screen, making it possible to easily check how the derived word and the related words were used in the actual document.

[0042] [Second embodiment] In the first embodiment, the frequency of occurrence and the degree of attention are used as axes for generating the word map, but the present invention is not limited to these, and any axis that can be used as an evaluation index for a word can be used. An example of another axis is an emotional element. It is possible to determine the emotional element of whether each word is used in a good sense or a bad sense. In addition to the emotional element of the word itself, the emotional element of the word in the context in which it is used can be considered. As an example, the emotional value of each word can be extracted by using a trained model that has been trained in advance on the extraction of emotional values. By extracting emotional values ​​for each word, such as a large value for good emotion and a small value for bad emotion, and using the emotional values ​​as the axes of the word map, it is possible to present different analysis results to the user.

[0043] The present invention is not limited to the above-described embodiment, and various configurations or embodiments can be adopted without departing from the gist of the present invention. [Explanation of symbols]

[0044] 10 Server device (information processing device) 11 Document Acquisition Section 12 Frequency of occurrence calculation section 13 Attention Calculation Unit 14 Related Word Identification Section 15 Document Extraction Unit 16 Word Map Generator 17 Display section 18 Memory section 20, 201~20n User terminals 30 Communication Network

Claims

1. An information processing program for causing a computer to realize a process of analyzing the contents of a plurality of documents, wherein the computer is caused to have a document acquisition function of acquiring a plurality of documents as text data respectively, have an appearance frequency calculation function of calculating the appearance frequency of words in the plurality of documents, have an emotional value calculation function of calculating the emotional value of words in the plurality of documents, have a word map generation function of generating and outputting a word map in which each word is plotted on a two-dimensional map with the appearance frequency and the emotional value as axes respectively based on the appearance frequency and the emotional value of each word, and have a display function of causing a display for display to display a display screen including the generated word map is an information processing program for realizing the above.

2. The emotional value calculation function is executed based on a learned model that has been pre-learned about calculating the emotional value of each word in the input document. The information processing program according to Claim 1.

3. The computer is caused to have a related word identification function of identifying other words with a high degree of relevance to each word in the plurality of documents, and when a predetermined operation is performed on any word in the word map on the display screen, the display function causes related words of the word to be derived and displayed from the display position of the word. The information processing program according to Claim 1 or 2.

4. An information processing apparatus for performing a process of analyzing the contents of a plurality of documents, comprising a document acquisition unit that acquires a plurality of documents as text data respectively, an appearance frequency calculation unit that calculates the appearance frequency of words in the plurality of documents, an emotional value calculation unit that calculates the emotional value of words in the plurality of documents, a word map generation unit that generates and outputs a word map in which each word is plotted on a two-dimensional map with the appearance frequency and the emotional value as axes respectively based on the appearance frequency and the emotional value of each word, and a display unit that causes a display for display to display a display screen including the generated word map is an information processing apparatus provided with the above.

5. An information processing method for causing a computer to perform a process of analyzing the contents of a plurality of documents, comprising a document acquisition procedure of acquiring a plurality of documents as text data respectively, an appearance frequency calculation procedure of calculating the appearance frequency of words in the plurality of documents, an emotional value calculation procedure of calculating the emotional value of words in the plurality of documents, A word map generation procedure that generates and outputs a word map in which each word is plotted on a two-dimensional map with the frequency of occurrence and the sentiment value as axes respectively based on the frequency of occurrence and the sentiment value of each word; A display procedure for displaying a display screen including the generated word map on a display for display; An information processing method including the above.

Images