Decision Support Text Rewriting With Source-Linked Response Verification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The question response apparatus disclosed in Patent Literature 1 does not verify the correctness of responses from external search engines, potentially leading to incorrect answers.
Innovation Solution
An information processing apparatus that acquires a target text, extracts a related document, rewrites the text using the document, generates a text with a machine learning model, and outputs the result with document information to enhance reliability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If a question response apparatus uses an external search engine to generate responses, then the response generation speed is improved, but the reliability of the response is worsened due to lack of verification
Solution Approach 1:
The patent implements a feedback mechanism where the generated response is fed back into the language model to calculate perplexity. This perplexity value serves as a verification metric to assess response reliability. The system uses this feedback loop to evaluate whether the response from the external search engine is trustworthy before finalizing it.
Solution Approach 2:
The patent performs preliminary verification by calculating the perplexity of the generated response before outputting it. This preliminary action involves dividing the response into morphemes, extracting keywords, searching for evidence documents, and computing perplexity values to ensure the response is reliable before it is presented to the user.
2Device complexity
If a language model generates text based on input text without verification, then the processing simplicity is improved, but the accuracy of the generated text is worsened
Solution Approach 1:
The patent replaces simple text generation with a verification mechanism based on perplexity calculation. Instead of relying solely on the language model's generation capability, the system substitutes in a computational verification process that measures how surprising or improbable the generated text is, thereby assessing its accuracy.
Solution Approach 2:
The patent introduces perplexity as an intermediary metric between the language model's generated text and the final output. This intermediary serves as a mediator that evaluates the quality of the generated text by comparing it against the input and evidence, ensuring accuracy before the text is finalized.
3Reliability
If the system extracts and verifies documents related to target text, then the reliability of language processing results is improved, but the processing time is worsened
Solution Approach 1:
The patent applies partial verification by focusing on key aspects of document validation rather than exhaustive checking. It extracts keywords from the target text, performs targeted searches for evidence documents, and calculates perplexity based on these partial results, achieving reliable verification without complete processing of all possible data.
Data Source
AI summary
To improve reliability of a result of language processing carried out with use of a machine learning model, an information processing apparatus includes at least one processor that carries out: an acquisition process of acquiring a target text; an extraction process of extracting a document related to the target text; a rewriting process of rewriting the target text with use of the document; a generation process of generating a text corresponding to the rewritten target text with use of a machine learning model trained to generate a text based on an input text; and an output process of outputting a result obtained by adding information identifying the document to the generated text.