Query Rephrasing Using Encoder-Decoder Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Computer-based query-answering systems often fail to provide satisfactory answers due to unclear or improperly phrased queries, especially when queries are submitted as speech, which may contain informal language, repetition, or filler words, leading to inefficiencies and reduced accuracy in answering.
Innovation Solution
A neural network-based system that uses an encoder-decoder architecture with a copy mechanism to rephrase queries by selecting words from both the original query and a set of common words, enhancing the vocabulary for generating more accurate and diverse rephrased queries without increasing processing time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If queries are submitted as speech with informal language, repetition, or filler words, then the system can accept more natural user input, but the accuracy of query understanding and answer retrieval deteriorates
Solution Approach 1:
The patent introduces an intermediary processing system that includes speech-to-text conversion, query rephrasing mechanisms, and synonym expansion. This intermediary layer transforms informal speech input into standardized, accurate queries before they reach the answer retrieval system, thereby maintaining both ease of operation and measurement precision.
Solution Approach 2:
The patent replaces direct mechanical matching of speech queries with answer databases by substituting it with an intelligent processing system using natural language understanding, query rephrasing, and semantic analysis. This substitution enables accurate interpretation of informal speech without requiring precise grammatical input.
2Productivity
If the system uses a limited vocabulary for query rephrasing, then the processing speed is maintained, but the accuracy and diversity of rephrased queries deteriorates
Solution Approach 1:
The patent applies preliminary action by pre-processing and expanding the vocabulary database with synonyms, related terms, and contextually relevant words before query processing. This pre-prepared expanded vocabulary enables accurate and diverse query rephrasing without increasing real-time processing complexity, thus maintaining productivity while improving measurement precision.
3Reliability
If the system rephrases all queries, then the answer retrieval accuracy improves, but the processing time increases
Solution Approach 1:
The patent applies partial action by selectively rephrasing only those queries that require improvement, rather than uniformly processing all queries. The system identifies queries with syntax errors, ambiguity, or poor structure and applies rephrasing only to these cases, maintaining high answer retrieval accuracy while minimizing unnecessary processing time for already well-formed queries.
4Measurement precision
If the system uses complex syntax and grammar rules for query validation, then the query accuracy improves, but the ease of user input deteriorates
Solution Approach 1:
The patent applies inversion by reversing the traditional approach: instead of requiring users to input queries that conform to complex syntax and grammar rules, the system accepts flexible, natural speech input and automatically transforms it into grammatically correct, syntactically accurate queries through rephrasing and normalization processes.
Data Source
AI summary
A method comprising receiving first data representative of a query. A representation of the query is generated using an encoder neural network and the first data. Words for a rephrased version of the query are selected from a set of words comprising a first subset of words comprising words of the query and a second subset of words comprising words absent from the query. Second data representative of the rephrased version of the query is generated.