Domain QA Retrieval Using Event-Aware Document Ranking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing document question answering systems struggle with low accuracy in retrieving relevant documents and generating accurate answers due to insufficient utilization of semantic vectors, event categories, and argument information in specific fields.
Innovation Solution
An answer information generation method using a large language model that involves obtaining semantic vectors and event information from user questions, retrieving candidate documents based on semantic vectors and event categories, determining quality evaluation information, and selecting target documents based on correlation and quality scores to enhance document recall accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional document retrieval methods are used, then the system is simple to operate, but the accuracy of retrieving relevant documents is low
Solution Approach 1:
The patent transforms traditional single-dimension keyword matching into multi-dimensional retrieval by introducing semantic vectors, event categories, and argument information as additional dimensions. This allows documents to be retrieved based on multiple features simultaneously, significantly improving retrieval accuracy while managing complexity through structured processing of each dimension.
Solution Approach 2:
The patent segments the document retrieval process into distinct components: semantic vector extraction, event category identification, argument information extraction, and quality evaluation. Each component handles a specific aspect of the retrieval task, making the overall complex system manageable through modular design while achieving high accuracy through coordinated operation of all segments.
2Reliability
If multiple retrieval criteria are used to improve document quality, then the relevance of retrieved documents improves, but the complexity of the retrieval process increases
Solution Approach 1:
The patent performs preliminary actions by pre-extracting semantic vectors, event categories, and argument information from both query and document texts before the actual retrieval matching process. This preparation work is done in advance, allowing the retrieval system to focus on comparing pre-computed features rather than processing raw text during retrieval, thus improving relevance while managing complexity through staged processing.
Solution Approach 2:
The patent changes the parameters used for document evaluation by introducing quality evaluation information that combines multiple criteria (semantic similarity, event category matching, argument information alignment). This multi-parameter approach improves document relevance selection while the systematic way of combining parameters keeps the process manageable through weighted scoring mechanisms.
3Measurement precision
If event information and argument information are extracted and utilized, then the precision of answer generation improves, but the processing time and system complexity increase
Solution Approach 1:
The patent extracts event information and argument information in advance during the document retrieval phase, so that when answer generation occurs, these structured elements are already available for immediate use. This preliminary extraction improves answer precision by providing contextual frameworks while reducing processing time during the actual answer generation stage since the work is already done.
Data Source
AI summary
A method is provided. The method includes: obtaining, in response to receiving a question text from a user, a semantic vector of the question text and event information related to a specific field; obtaining a plurality of candidate documents from a document library of the specific field based on at least two of the semantic vector of the question text, the at least one piece of argument information and the event category; determining quality evaluation information for each candidate document in the plurality of candidate documents based on the event category; and determining at least one target document from the plurality of candidate documents based on the quality evaluation information of each candidate document and a correlation between each candidate document and the question text, to obtain, based on the at least one target document, answer information used to answer the question text.