Domain-Centric Indexed Corpus for Context-Aware Question Answering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing question-answering systems lack sufficient contextual understanding, leading to inefficient retrieval processes due to limited domain-specific knowledge organization and trustworthiness.
Innovation Solution
A method and system for generating an indexed corpus by transforming raw corpus data into plain text, normalizing it, performing chunking based on document type, and using a domain model graph and LLM to generate domain-centric chunks, which are then indexed for efficient query processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If existing QA systems use simple retrieval processes, then the system complexity is low, but the contextual understanding and retrieval efficiency are insufficient
Solution Approach 1:
The patent segments the raw corpus into multiple hierarchical levels: document level, section level, paragraph level, and sentence level. This segmentation allows the system to retrieve information at the most appropriate granularity, improving contextual understanding while managing complexity through structured organization.
Solution Approach 2:
The patent introduces a new dimension of domain-specific metadata tagging alongside traditional text retrieval. By embedding domain knowledge, entity relationships, and contextual metadata as additional dimensions, the system enhances retrieval accuracy without proportionally increasing operational complexity.
2Measurement precision
If domain-specific knowledge is organized in a structured manner, then the trustworthiness and accuracy improve, but the processing time and computational resources increase
Solution Approach 1:
The patent performs preliminary actions during the indexing phase by pre-processing the corpus, extracting entities, establishing relationships, and creating metadata tags before queries are submitted. This advance organization reduces processing time during actual query execution while maintaining high accuracy through pre-structured domain knowledge.
Solution Approach 2:
The patent extracts only the essential domain-specific elements (key entities, relationships, and contextual metadata) from the full corpus during indexing, rather than processing entire documents during query execution. This extraction approach maintains response accuracy by focusing on relevant information while reducing computational overhead.
3Productivity
If the corpus is processed with detailed domain-specific chunking, then the retrieval relevance improves, but the device complexity and processing overhead increase
Solution Approach 1:
The patent implements multi-level segmentation of the corpus into documents, sections, paragraphs, and sentences, with each level serving specific retrieval needs. This hierarchical segmentation improves retrieval efficiency by allowing queries to target appropriate granularity levels while maintaining manageable system complexity through systematic organization.
Solution Approach 2:
The patent applies different processing qualities to different parts of the corpus based on their relevance and structure. Domain-specific metadata, entity relationships, and contextual tags are applied selectively to relevant portions rather than uniformly across all text, improving retrieval relevance while controlling processing complexity.
Data Source
Figure 1
Figure 2A
Figure 2B
AI summary
Existing question answering approaches have the disadvantages that they possess limited contextual understanding due to which the retrieval process they use is inefficient in nature. Embodiments disclosed herein provide a method and system for domain-driven knowledge augmented question answering. The system receives a raw corpus data as input, wherein the raw corpus data is a domain specific data. Further, an indexed corpus is generated from the raw corpus data, during which a document chunking approach is used. The indexed corpus is then used for processing received queries received, in order to generate response to the received user queries.