Legal Answer Retrieval With On-the-Fly Semantic Ranking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for identifying relevant legal responses to queries require significant computational resources and time, especially when dealing with large response corpora, making them inefficient and slow.
Innovation Solution
A computer-implemented method using a ranking function to reduce answer passage candidates, followed by on-the-fly vectorization with a semantic language model like SBERT, which embeds text passages in high-dimensional vector spaces, reducing hardware requirements and processing time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a semantic language model is used to embed text passages in high-dimensional vector spaces, then the relevance identification accuracy is improved, but the computational resources and processing time required increase significantly
Solution Approach 1:
The patent segments the legal response corpus into individual text passages and processes them in batches through the semantic language model. By dividing the large corpus into manageable units and processing them sequentially in batches, the system reduces the computational burden on hardware while maintaining accurate relevance identification for each passage.
Solution Approach 2:
The patent performs preliminary actions by pre-processing the legal documents into standardized text passages and preparing them for batch processing. The system pre-organizes the corpus structure and prepares processing parameters before executing the semantic embedding, which reduces the computational overhead during the actual relevance identification process.
2Quantity of substance
If a large response corpus is processed to ensure comprehensive answer coverage, then the completeness of legal information is improved, but the processing time and system response speed decrease
Solution Approach 1:
The patent segments the large response corpus into individual text passages and processes them in manageable batches. This segmentation allows the system to handle comprehensive corpora while maintaining reasonable processing speeds by processing passages in parallel batches rather than sequentially through the entire corpus.
Solution Approach 2:
The patent changes processing parameters by using batch processing with configurable batch sizes and adjusting the semantic language model processing parameters. This allows the system to optimize the balance between processing comprehensive corpora and maintaining acceptable response times through parameter adjustment.
3Ease of operation
If existing methods are used to identify relevant legal responses, then the approach is straightforward, but the system becomes slow and inefficient with large corpora
Solution Approach 1:
The patent replaces traditional mechanical search methods with a semantic language model-based system that uses contextual embedding and similarity computation. This substitution maintains the simplicity of the interface while dramatically improving processing speed through efficient vector space operations and batch processing mechanisms.
Data Source
AI summary
The invention relates to a computer-implemented method and a computer program product for determining relevant answer passages from an answer corpus for a question relating to a legal matter, which method can determine relevant answer passages with relatively low hardware resources and with high speed despite the use of a semantic language model which embeds text passages in high-dimensional vector spaces in a context-sensitive manner. Despite the use of the semantic language model, the method according to the invention can be carried out with relatively low computing power and hardware resources, because the answer passages of an answer corpus for a question are only vectorised in a sorting step on-the-fly, i.e. in real time, when the fixed number of possible candidates among the answer passages has already been significantly reduced by means of a ranking list function.