Temporal Expressions Processor for Complex NLP
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current natural language processing systems struggle to accurately process complex temporal expressions that require inference and reasoning, such as 'set my alarm to an hour and a half before sunrise', due to the lack of time references and indirect references within commands.
Innovation Solution
A system that employs a temporal expressions processor with a semantic parser and temporal value resolver to convert natural language temporal expressions into concrete dates and times by using a temporal ontology, lexicon, and grammar rules, enabling inference and handling indirect references through a combination of parsing and value resolution functions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If shallow parsing and keyword identification are used to extract text fragments, then processing speed is improved, but the ability to handle complex temporal expressions requiring inference and reasoning deteriorates
Solution Approach 1:
The system segments the temporal expression processing into distinct components: a semantic parser that generates multiple possible interpretations (semantic parses) of the temporal expression, and a temporal value resolver that selects the optimal parse and resolves the temporal value. This segmentation allows the system to handle complex expressions requiring inference while maintaining processing efficiency through specialized sub-components.
2Adaptability or versatility
If a comprehensive approach with semantic parsing and value resolution is implemented, then the ability to resolve indirect temporal references is improved, but system complexity increases
Solution Approach 1:
The semantic parser acts as an intermediary component that bridges the gap between raw temporal expressions and their resolved values. It generates multiple semantic parses that represent different possible interpretations of the expression, allowing the system to handle indirect references and inference requirements without requiring the entire system to be overly complex. The temporal value resolver then selects from these intermediate parses to determine the final meaning.
3Measurement precision
If multiple semantic parses are generated and evaluated, then accuracy in resolving temporal expressions is improved, but processing time increases
Solution Approach 1:
The system performs preliminary action by generating multiple semantic parses before final value resolution. This allows the temporal value resolver to evaluate different interpretations and select the optimal one, improving accuracy. The preliminary generation of parses enables the system to consider multiple possibilities without significantly increasing overall processing time, as the semantic parsing phase is separated from the value resolution phase.
Data Source
AI summary
Embodiments of the present disclosure resolve natural language temporal expressions into temporal values via a temporal expressions processor that parses a given temporal expression into one or more semantic parses, selects an optimal semantic parse and resolves a value for the optimal semantic parse. The temporal values may pertain to one or more calendar dates, one or more calendar periods, one or more clock times, a clock time period and/or a combination of the above.