Semantic Vector Splicing for Question Answering Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing question and answer systems face inefficiencies and low accuracy in automatically providing answers, with retrieval-based systems requiring costly model optimization and database-based systems experiencing low answer coverage and poor performance.
Innovation Solution
A method involving semantic representation models to acquire and process questions, where a first model is trained on question and answer samples, and a second model on question and article titles, with spliced semantic vectors used for matching against a pre-established vector index library of answers, eliminating the need for confidence calculations and improving relevance-based matching.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If retrieval-based question-answer systems are used, then answer accuracy can be improved, but model optimization costs and computational complexity increase significantly
Solution Approach 1:
The patent segments the question-answer matching process into two independent semantic representation models: one for questions (trained on question-answer pairs) and one for answers (trained on question-article title pairs). This segmentation allows each model to be optimized independently with simpler training objectives, reducing overall system complexity while maintaining high matching accuracy through their combination.
2Ease of manufacture
If database-based question-answer systems are used, then implementation simplicity is improved, but answer coverage and performance deteriorate
Solution Approach 1:
The patent transforms the traditional database-based approach by changing the representation parameters from raw text to semantic representation vectors. This allows the system to maintain implementation simplicity while significantly improving answer coverage and performance, as the semantic vectors enable more effective matching and retrieval of relevant answers from the database.
3Measurement precision
If semantic vectorization is performed on all candidate answers in real-time, then matching accuracy is improved, but computational load and processing time increase
Solution Approach 1:
The patent performs preliminary action by pre-computing and storing the semantic representation vectors for all candidate answers in advance, building a pre-established vector index library. During online processing, only the question needs to be vectorized and matched against the pre-computed answer vectors, dramatically reducing real-time computational load while maintaining high matching accuracy.
Data Source
AI summary
A method for processing questions and answers includes: in a process of determining an answer to a question to be answered, determining the semantic representation on the question to be answered respectively with a first semantic representation model of question and a second semantic representation model of question. Semantic representation vectors obtained through the first semantic representation model of question and the second semantic representation model of question are spliced. A spliced semantic vector is determined as a semantic representation vector of the question to be answered. An answer semantic vector matching the semantic representation vector of the question to be answered is acquired from a vector index library of answer, and an answer corresponding to the answer semantic vector is determined as a target answer to the question to be answered.