Target-Text Rewriting with Document-Linked Machine Learning Responses
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing question response apparatus does not verify the correctness of responses from external search engines, leading to potential inaccuracies in language processing results.
Innovation Solution
An information processing apparatus that acquires a target text, extracts relevant documents, rewrites the text using these documents, generates a corresponding text with a machine learning model, and outputs the result with document identification information.
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 reliable before finalizing it, thus resolving the contradiction between fast response generation and reliable verification.
Solution Approach 2:
The patent performs preliminary actions by calculating perplexity values during the response generation process. Before finalizing the response, the system pre-calculates perplexity metrics to assess potential reliability issues. This preliminary verification step allows the system to identify unreliable responses early, preventing incorrect information from being output while maintaining efficient processing.
2Reliability
If the question response apparatus verifies response correctness, then the reliability is improved, but the processing time increases
Solution Approach 1:
The patent applies partial verification by calculating perplexity values selectively during the response generation process. Rather than performing exhaustive verification on all responses, the system uses perplexity as a partial check that can quickly identify obviously unreliable responses. This partial action approach provides reasonable assurance of correctness without requiring complete verification of every response detail, thus balancing reliability improvement with time constraints.
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.