Domain-Specific Speech Intent Inference with Phonetic Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Digital assistants face challenges in accurately inferring user intent from speech inputs due to errors in speech-to-text conversion, leading to unsatisfactory responses, as they struggle to recognize words and phrases with imperfect transcriptions.
Innovation Solution
The implementation of a natural language processing system that uses an ontology organized into domains, with a two-stage word-matching process to determine user intent, including a first exact match and a second more relaxed phonetic or approximate match within the identified domain, and considers multiple candidate text strings to select the most accurate representation of user input.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If speech-to-text conversion is used to process user input, then the digital assistant can handle natural language speech inputs, but errors in transcription propagate to the natural language processor causing incorrect intent inference
Solution Approach 1:
The system performs preliminary domain identification before detailed intent analysis. By first determining which domain the speech input belongs to (e.g., weather, news, sports), the system can then apply domain-specific processing to correct transcription errors and improve intent inference accuracy.
Solution Approach 2:
The system applies different processing strategies to different parts of the input based on domain. Once the domain is identified, the natural language processor uses domain-specific vocabularies, error correction rules, and intent models tailored to that domain, rather than applying a single generic processing approach to the entire input.
2Productivity
If a single exact word-matching process is used to determine user intent, then the processing is simple and fast, but it fails to recognize words with transcription errors
Solution Approach 1:
The word-matching process is segmented into multiple stages: first an exact match attempt for quick processing, then phonetic matching for potential corrections, and finally approximate matching as a fallback. This segmented approach maintains speed for clear inputs while providing accurate correction for erroneous transcriptions.
Solution Approach 2:
The matching process is dynamic and adaptive. The system adjusts the strictness of matching criteria based on the domain identified and the confidence level of initial matching. If exact matching fails, the system dynamically transitions to more flexible phonetic and approximate matching strategies within the context of the identified domain.
Data Source
AI summary
A text string with a first and a second portion is provided. A domain of the text string is determined by applying a first word-matching process to the first portion of the text string. It is then determined whether the second portion of the text string matches a word of a set of words associated with the domain by applying a second word-matching process to the second portion of the text string. Upon determining that the second portion of the text string matches the word of the set of words, it is determined whether a user intent from the text string based at least in part on the domain and the word of the set of words.