Highlighting Answer Phrases Using RNN Relevance Scores
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing question and answer systems face challenges in providing accurate and concise answers to user queries due to limitations in extracting relevant features and entities, leading to less accurate and longer answer passages.
Innovation Solution
The system generates relevance scores for words in answer passages by encoding natural language queries and passages into semantic vector representations using Recurring Neural Networks (RNNs) with Long Short-Term Memory (LSTM), allowing for the identification and highlighting of highly relevant words and phrases.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional question and answer systems extract relevant features and entities from search results, then they can provide answers to user queries, but the answer passages become less accurate and longer
Solution Approach 1:
The system segments the answer passage at the word level by generating individual relevance scores for each word. This allows the system to identify and highlight only the most relevant words rather than presenting the entire passage, thereby improving answer accuracy while reducing unnecessary information length.
Solution Approach 2:
The system applies local quality by assigning different relevance scores to different words within the passage. Words with higher relevance scores (above a threshold) are highlighted, while less relevant words are not emphasized. This selective highlighting improves the precision of the answer by making important information stand out without requiring the user to read the entire passage.
2Reliability
If the system provides complete answer passages, then all relevant information is available, but user interaction time increases
Solution Approach 1:
The system extracts and highlights only the most relevant words from the complete answer passage by comparing relevance scores against a threshold. This allows users to quickly grasp the essential information without reading the entire passage, significantly reducing user interaction time while maintaining access to complete information if needed.
Solution Approach 2:
The system performs preliminary analysis by pre-calculating relevance scores for all words in the passage before presentation. This preprocessing step identifies which words are most relevant in advance, allowing the system to present a condensed highlighted version immediately without requiring users to manually search through the complete passage.
3Measurement precision
If the system highlights more words in the passage, then more relevant information is emphasized, but the complexity of the system increases
Solution Approach 1:
The system changes the parameter of word importance by assigning numerical relevance scores to each word based on its relationship to the query. By adjusting the threshold parameter, the system can control how many words are highlighted, balancing between precision of relevance identification and system complexity. This parameter-based approach provides a simple yet effective mechanism for managing highlighting without excessive complexity.
Data Source
AI summary
Examples of the present disclosure describe systems and methods relating to generating relevance scores for one or more words of a passage which is an answer to a natural language query. For instance, a passage extracted from a highly relevant electronic file along with the query may encoded and augmented to generate a multi-dimensional, augmented semantic vectors using recurring neural networks. The augmented semantic vectors along with a multi-dimensional vector that represent words of the passage may be decoded to generate relevance scores for one or more words of the passage, based on levels of relevance to the query.