Information processing device, information processing method, and program

Abstract

Efficiently collect information from multiple perspectives related to the content. [Solution] In the information processing device, the extraction means extracts key information, which is the main information of the content, from the content. The conversion means converts the key information into key information from a different perspective than that of the original key information. The search means searches for and outputs related information related to the converted key information.

Description

[Technical Field]

[0001] The present disclosure relates to information analysis. [Background Art]

[0002] Various contents are published every day through media and SNS (Social Networking Service), and the published information includes misinformation (unintentionally incorrect information) and disinformation (intentionally fabricated false information). As a countermeasure against such misinformation and disinformation, reliability assessment (fact-checking) by experts is effective.

[0003] In the news distribution system described in Patent Document 1, a predetermined expert is requested to perform fact-checking on a news article, and if the expert's evaluation is positive, a fairness score of the article is added. This allows a user to determine whether to trust the news article with reference to the score. [Prior Art Documents] [Patent Documents]

[0004] [Patent Document 1] Japanese Unexamined Patent Application Publication No. 2024-015904 [Summary of the Invention] [Problem to be Solved by the Invention]

[0005] However, fact-checking by experts requires time, labor, and cost for manual content analysis, which poses the problem of a large workload for content reliability assessment.

[0006] One object of the present disclosure is to efficiently collect information from multifaceted perspectives related to content. [Means for Solving the Problem]

[0007] From one perspective of this disclosure, the information processing device is An extraction means for extracting key information, which is the main information of the content, from the content, A conversion means that converts the aforementioned key information into key information from a perspective different from that of the aforementioned key information, A search method that searches for and outputs related information associated with the converted key information, It is equipped with.

[0008] From another perspective of this disclosure, computer-based information processing methods are: Extract key information from the content, which is the main information of that content. The aforementioned key information is converted into key information from a perspective different from that of the aforementioned key information, Search for and output related information associated with the converted key information.

[0009] In yet another aspect of this disclosure, the program is Extract key information from the content, which is the main information of that content. The aforementioned key information is converted into key information from a perspective different from that of the aforementioned key information, The computer is instructed to perform a process of searching for and outputting related information associated with the converted key information. [Effects of the Invention]

[0010] This disclosure makes it possible to efficiently collect information from multiple perspectives related to the content. [Brief explanation of the drawing]

[0011] [Figure 1] This document outlines the overall configuration of the information analysis support system related to this disclosure. [Figure 2] This is a block diagram showing the hardware configuration of an information analysis support device. [Figure 3] This is a diagram showing the functional configuration of an information analysis support device. [Figure 4] An example of related information output is shown. [Figure 5] It is a flowchart of information analysis support processing. [Figure 6] It shows the functional configuration of an information processing apparatus according to the present disclosure. [Figure 7] It is a flowchart of processing performed by the information processing apparatus. MODE FOR CARRYING OUT THE INVENTION

[0012] Hereinafter, preferred embodiments of the present disclosure will be described with reference to the drawings. <First Embodiment> [Basic Idea] As described above, fact-checking by experts requires time, labor, and cost for human content analysis. For this reason, it is desired to reduce the work load of content reliability evaluation. In the present disclosure, in fact-checking, collection of evaluation information for determining the reliability of content of an evaluation target is performed by automated search or the like. This makes it possible to reduce the work load of reliability evaluation.

[0013] When searching for evaluation information for determining reliability, basically, main information (claims and main points) is extracted from content to be evaluated, and search related to the extracted information is performed. However, if the search is performed based only on the main information extracted from the content to be evaluated, information conforming to the main points and claims of the content to be evaluated tends to be easily collected. For this reason, information from positions different from the main points and claims of the content to be evaluated becomes less likely to be collected, and there is a risk that the evaluation becomes unfair in terms of reliability evaluation of information.

[0014] Accordingly, in the present disclosure, first, main information (hereinafter referred to as "key information") is extracted from content to be evaluated. Then, the extracted key information is converted into information from a different standpoint, a search related to the information from the different standpoint is performed, and a search result is output. This makes it possible to collect information from standpoints different from the gist and assertions of the content to be evaluated, and to fairly evaluate the reliability of the information. Herein, the different standpoints include differences in standpoint due to status and role. Examples thereof include a difference between a superior and a subordinate in a company, and a difference between an ordering party and a contractor in business. Further, the different standpoints include differences in standpoint resulting from cultural differences, regional differences, religious differences, and the like.

[0015] [Overall Configuration] Figure 1 shows the overall configuration of an information analysis support system according to the present disclosure. The information analysis support system 1 is a system for supporting a user's reliability evaluation of information, and includes a user terminal 2 and an information analysis support apparatus 100. The information analysis support apparatus 100 is an example of the information processing apparatus of the present disclosure, and is configured by a server apparatus or the like. The user terminal 2 is a terminal apparatus such as a personal computer used by a user who evaluates the reliability of information. The user terminal 2 and the information analysis support apparatus 100 are communicably connected via a network or the like.

[0016] A user inputs content to be subjected to reliability evaluation to the information analysis support apparatus 100 via the user terminal 2. The information analysis support apparatus 100 searches for information related to the content input from the user terminal, and transmits the obtained related information to the user terminal 2. At this time, as described above, the information analysis support apparatus 100 performs a search based on information from a standpoint different from that of the key information in the content to be evaluated, and acquires related information. This allows the user to obtain multifaceted information related to the content to be evaluated, and fairly evaluate the reliability of the content to be evaluated based on such information.

[0017] [Hardware Configuration] Figure 2 is a block diagram showing the hardware configuration of the information analysis support device 100. As shown in the figure, the information analysis support device 100 comprises a processor 11, an interface (IF) 12, a ROM (Read Only Memory) 13, a RAM (Random Access Memory) 14, a database (DB) 15, and a storage medium 16. Each component is connected to the others, for example, via a bus 18.

[0018] The processor 11 is a computer such as a CPU (Central Processing Unit) and controls the entire information analysis support device 100 by executing a pre-prepared program. Specifically, the processor 11 can be a CPU, GPU (Graphics Processing Unit), DSP (Digital Signal Processor), MPU (Micro Processing Unit), FPU (Floating Point Number Processing Unit), PPU (Physics Processing Unit), TPU (Tensor Processing Unit), quantum processor, microcontroller, or a combination thereof.

[0019] Furthermore, the processor 11 loads programs stored in the ROM 13 and storage medium 16 into the RAM 14 and executes each process coded in the program. The processor 11 functions as part or all of the information analysis support device 100. Specifically, the processor 11 performs the information analysis support processes described later.

[0020] IF12 transmits and receives data to and from external devices. Specifically, the information analysis support device 100 receives the content to be evaluated from the user terminal 2 via IF12. The information analysis support device 100 also transmits related information obtained through searching to the user terminal 2 via IF12.

[0021] ROM 13 stores various programs executed by processor 11. RAM 14 is used as working memory while processor 11 is executing various processes.

[0022] DB15 stores various algorithms, data, machine learning models, language models, etc., that the information analysis support device 100 uses when it performs the information analysis support processing described later.

[0023] The storage medium 16 is a non-volatile, non-temporary storage medium such as a disk-shaped recording medium or semiconductor memory. The storage medium 16 may be configured to be detachable from the information analysis support device 100. The storage medium 16 records various programs executed by the processor 11.

[0024] [Functional Configuration] Figure 3 is a block diagram showing the functional configuration of the information analysis support device 100. The information analysis support device 100 comprises a key information extraction unit 21, a key information conversion unit 22, and a search unit 23.

[0025] The information analysis support device 100 receives input of the content to be evaluated. The content to be evaluated may be text, images, videos, audio, or any other format.

[0026] The key information extraction unit 21 extracts key information from the input content. "Key information" refers to the main information contained in the content, specifically including the main points, summary, gist, argument, or opinion of the content. The key information extraction unit 21 extracts key information from the content using a language model. A "language model" is a model that outputs a response in language to the input language. The input language and the output language do not necessarily have to match. The language model may also be a model that outputs a response in a format other than language, such as an image or audio. An example of a language model is an LLM (Large Language Model).

[0027] Specifically, if the input content is text, the key information extraction unit 21 extracts key information from the content using a machine learning-trained language model. As the machine learning-trained language model, for example, a model that has been trained on the arrangement of its constituent elements (such as words) in a sentence, or the arrangement of sentences in a document, may be applied. In terms of obtaining highly accurate output, it is particularly preferable to use an LLM generated by machine learning using a large language corpus. For example, as an LLM used to extract key information such as claims in the content, a Generative Pre-Trained Transformer (GPT) can be used, which outputs a sentence containing the input string by predicting a string with a high probability of following the input string. In addition, other LLMs that can be used to extract key information include, for example, T5 (Text-to-Text Transfer Transformer), BERT (Bidirectional Encoder Representations from Transformers), RoBERTa (Robustly optimized BERT approach), and ELECTRA (Efficiently Learning an Encoder that Classifies Token Replacements Accurately).

[0028] On the other hand, if the input content is an image or video, the key information extraction unit 21 converts the content of the image or video into text using a machine learning-trained image language model, and extracts key information from the text using a machine learning-trained language model. Also, if the input content is audio, the key information extraction unit 21 converts the content of the audio into text using a machine learning-trained transcription model, and extracts key information from the text using a machine learning-trained language model.

[0029] Examples of models for converting image content to text include BLIP (Bootstrap Language Image Pre-Training). Examples of models for converting video content to text include Video-LLaVa. Examples of models for converting audio content to text include Whisper. Furthermore, optical character recognition (OCR) technologies such as ViTSTR (Vision Transformer for Fast and Efficient Scene Text Recognition) can be used to extract text from videos. Alternatively, text may be generated from non-text elements using, for example, a Vision Language Model (VLM) that accepts multiple modals as input and generates text.

[0030] When using a language model to extract key information from the text of content, you should input instructions (prompts) along with the text into the language model, such as the following: You will be given a text whose truth or falsity you want to determine. Your job is to comprehensively analyze the given text and accurately identify and extract the claims contained within it. Here, a claim is an opinion whose truth or falsity can be determined. If the text contains multiple claims, please extract all of them.

[0031] The key information extraction unit 21 outputs the key information extracted in the manner described above to the key information conversion unit 22 and the search unit 23.

[0032] The key information conversion unit 22 converts the assertions and opinions contained in the key information, as well as the validity of the assertions, to generate key information that differs from the assertions and main points contained in the input content (hereinafter also referred to as "converted key information"). Specifically, the key information conversion unit 22 converts the key information by using a machine learning-based language model to replace the subject of the assertions and main points contained in the key information with another subject related to that subject, to change the intensity of modifying parts, or to reverse the validity of predicates. The "predicate" is one of the components of a sentence and refers to the part that describes "what state it is in" or "what is being done" regarding the content expressed by the subject (or subject part).

[0033] For example, if the input content is a politician's argument for legal reform, the key information conversion unit 22 may generate opinions from legal experts or opinions from those opposed to the legal reform based on the key information. Also, if the input content is a statement from management at a company, the key information conversion unit 22 may generate claims and opinions from employees at that company based on that key information.

[0034] Furthermore, if the input content is an assertion that a tax increase is necessary, the key information conversion unit 22 may generate assertions such as assertions that a tax increase is unnecessary or assertions that a tax reduction is necessary, based on the key information. The key information conversion unit 22 may also generate an assertion that a tax increase is necessary, but with a different tax rate. For example, if the input content asserts a 10% tax increase, the conversion unit 22 may generate converted key information asserting a 5% tax increase.

[0035] When converting key information using a language model, the key information conversion unit 22 should input an instruction (prompt) along with the text to the language model, such as the following: "You will be given a text. Your job is to comprehensively analyze the given text and transform its claims into possible alternative opinions, different positions, inverted versions of right and wrong, and opposing arguments."

[0036] The key information conversion unit 22 outputs the converted key information generated in this way to the search unit 23.

[0037] The search unit 23 receives the key information output from the key information extraction unit 21 and the converted key information output from the key information conversion unit 22. The search unit 23 then uses the key information and converted key information as search keywords to search for related information and outputs the obtained related information in association with the key information and converted key information. The search unit 23 can use internet search engines or internally built information retrieval systems as search methods.

[0038] The search unit 23 may output links to each related piece of information instead of the related information itself obtained through the search, or it may output summaries, claims, key points, etc., generated for each related piece of information using a language model, image language model, transcription model, etc., or it may output a combination of these.

[0039] [Example of related information output] Figure 4 shows an example of related information output obtained by the information analysis support device 100. Figure 4 shows an example of related information output by the information analysis support device 100 when a user inputs content arguing for a "tax increase on XX" using the user terminal 2 as the content to be evaluated.

[0040] In this example, the key information extraction unit 21 extracts key information 51 from the content to be evaluated, which includes the assertion, "The XX tax needs to be increased." The key information conversion unit 22 then generates converted key information 52 from key information 51, which includes the opposing assertion, "The XX tax does not need to be increased." The search unit 23 searches for related information using key information 51 and converted key information 52, and obtains related information 53 corresponding to key information 51 and related information 54 corresponding to converted key information 52. The information analysis support device 100 then associates key information 51 with its corresponding related information 53 and associates converted key information 52 with its corresponding related information 54, and displays them.

[0041] Thus, in the information analysis support device 100, the key information conversion unit 22 converts the claims and main points of the input content into information from different perspectives, which is then used as search keywords in the search unit 23. This makes it possible to collect information related to the input content from multiple viewpoints.

[0042] [Information analysis support processing] Next, the processing performed by the information analysis support device 100 will be described. Figure 5 is a flowchart of the information analysis support processing. This processing can be realized by having the processor execute a pre-prepared program and operate as each element shown in Figure 3. The entity executing each step in this information analysis support method may be the processor 11 of the information analysis support device 100 shown in Figure 2, or it may be a processor in another device, or the entity executing each step may be a processor in a different device.

[0043] First, when the user inputs the content to be evaluated, the key information extraction unit 21 acquires the content to be evaluated (step S11). Next, the key information extraction unit 21 extracts key information from the content to be evaluated (step S12). If the input content is not text, the key information extraction unit 21 converts it to text. Then, the key information extraction unit 21 inputs the obtained text into the language model and extracts key information including the main points and main arguments of the input content.

[0044] Next, the key information conversion unit 22 converts the extracted key information into information from a different perspective (step S13). Specifically, the key information conversion unit 22 uses a language model to convert the key information by replacing the subject of the claims and main points contained in the key information with another subject, changing the intensity of the modifying parts, and reversing the validity of the predicates.

[0045] Next, the search unit 23 performs a search using the key information and the converted key information to obtain related information (step S14). Then, the search unit 23 outputs the obtained related information to the user terminal 2, associating it with the key information and the converted key information (step S15). The information analysis support process then ends.

[0046] [Differentiation] In the example above, the user inputs the content to be evaluated, but the application of this disclosure is not limited to this. For example, the information analysis support device 100 may periodically acquire news or articles from a designated website, perform information analysis support processing, and then acquire, output, or display related information.

[0047] [Examples of application] The information analysis support disclosed here can be applied to information analysis support and related information acquisition in various fields. For example, this information analysis support can be applied to medical diagnostic support in the medical field. Specifically, it can be used to evaluate a doctor's diagnosis results regarding a patient's symptoms and medical history, and to provide diverse diagnostic results such as medical literature and second opinion results as related information. Furthermore, this information analysis support can be applied to news analysis systems. This makes it possible to build a system that generates and aggregates various opposing viewpoints, such as agreement and disagreement, regarding a particular news item or article. In addition, this information analysis support can be applied to research systems in the educational field. This makes it possible to generate various theories and perspectives on a given research topic and to collect and summarize papers and articles from multiple angles.

[0048] <Second Embodiment> Figure 6 is a block diagram showing the functional configuration of the information processing device of the second embodiment. The information processing device 70 includes an extraction means 71, a conversion means 72, and a search means 73.

[0049] Figure 7 is a flowchart of the processing performed by the information processing device of the second embodiment. The extraction means 71 extracts key information, which is the main information of the content, from the content (step S71). The conversion means 72 converts the key information into key information from a different perspective (step S72). The search means 73 searches for and outputs related information related to the converted key information (step S73).

[0050] According to the information processing device 70 of the second embodiment, it becomes possible to efficiently collect information from multiple perspectives related to the input content.

[0051] Some or all of the above embodiments may also be described as follows, but are not limited to the following:

[0052] (Note 1) An extraction means for extracting key information, which is the main information of the content, from the content, A conversion means that converts the aforementioned key information into key information from a perspective different from that of the aforementioned key information, A search method that searches for and outputs related information associated with the converted key information, An information processing device equipped with the following features.

[0053] (Note 2) The aforementioned key information is an information processing device as described in Appendix 1, which includes the main points of the content or the claims made in the content.

[0054] (Note 3) The conversion means is an information processing device as described in Appendix 2, which changes the position of the main point or the assertion to a different position.

[0055] (Note 4) The conversion means is an information processing device as described in Appendix 2, which reverses the validity of the main points or the assertion.

[0056] (Note 5) The conversion means is an information processing device according to Appendix 2 that changes the intensity of the main points or the claims.

[0057] (Note 6) The conversion means is an information processing device according to any one of the appendices 1 to 5, which inputs the key information and an instruction statement instructing to convert the key information into key information from a different perspective into a language model, and generates the converted key information.

[0058] (Note 7) The extraction means is an information processing device according to any one of the appendices 1 to 6, which inputs the content and an instruction statement instructing the extraction of the key information from the content into a language model and extracts the key information from the content.

[0059] (Note 8) The aforementioned content may be text, images, videos, or audio. The information processing device described in Appendix 7, wherein, if the content is not text, the extraction means converts the content to text and then inputs it into the language model.

[0060] (Note 9) A method of information processing performed by a computer, Extract key information from the content, which is the main information of that content. The aforementioned key information is converted into key information from a perspective different from that of the aforementioned key information, An information processing device that searches for and outputs related information associated with the converted key information.

[0061] (Note 10) Extract key information from the content, which is the main information of that content. The aforementioned key information is converted into key information from a perspective different from that of the aforementioned key information, A program that instructs a computer to search for and output related information associated with the converted key information.

[0062] Furthermore, some or all of the configurations described in Appendices 2 to 8, which are dependent on Appendice 1 above, may also be dependent on Appendices 9 and 10 in the same way as those described in Appendices 2 to 8. Moreover, not limited to Appendices 1, 9, and 10, some or all of the configurations described as appendices may also be dependent on various hardware, software, various recording means for recording software, or systems, without departing from the embodiments described above.

[0063] Although the present disclosure has been described above with reference to embodiments and examples, the present disclosure is not limited to the above embodiments and examples. Various modifications to the structure and details of the present disclosure can be understood by those skilled in the art within the scope of the present disclosure. [Explanation of symbols]

[0064] 1. Information Analysis Support System 2 User terminals 11 processors 21 Key Information Extraction Unit 22 Key Information Conversion Unit 23 Search Section 100 Information analysis support equipment

Claims

1. An extraction means for extracting key information, which is the main information of the content, from the content, A conversion means that converts the aforementioned key information into key information from a perspective different from that of the aforementioned key information, A search method that searches for and outputs related information associated with the converted key information, An information processing device equipped with the following features.

2. The information processing apparatus according to claim 1, wherein the key information includes the main points of the content or the claims made in the content.

3. The information processing apparatus according to claim 2, wherein the conversion means changes the position of the main point or the assertion to a different position.

4. The conversion means is an information processing device according to claim 2, which reverses the validity of the main point or the assertion.

5. The information processing apparatus according to claim 2, wherein the conversion means changes the intensity of the main points or the assertion.

6. The information processing apparatus according to any one of claims 1 to 5, wherein the conversion means inputs the key information and an instruction statement that instructs to convert the key information into key information from a different perspective into a language model, and generates the converted key information.

7. The information processing apparatus according to any one of claims 1 to 6, wherein the extraction means inputs the content and an instruction statement instructing the extraction of the key information from the content into a language model, and extracts the key information from the content.

8. The aforementioned content may be text, images, videos, or audio. The information processing apparatus according to claim 7, wherein, if the content is not text, the extraction means converts the content to text and then inputs it into the language model.

9. A method of information processing performed by a computer, Extract key information from the content, which is the main information of that content. The aforementioned key information is converted into key information from a perspective different from that of the aforementioned key information, An information processing device that searches for and outputs related information associated with the converted key information.

10. Extract key information from the content, which is the main information of that content. The aforementioned key information is converted into key information from a perspective different from that of the aforementioned key information, A program that instructs a computer to search for and output related information associated with the converted key information.

Images