Objective
The current implementation uses a list of regex patterns, word count, and domain-specific keywords to determine if a user's input needs clarification. This methodology should be improved to avoid using specific words/phrases to infer user intent and meaning.
Initial Implementation Requirements
- Evaluate some methods for identifying phrases that "vague_patterns" is trying to match
- Identify intent behind "specific_keywords" to handle those cases in a more scalable/generic way
Other Considerations
vague_patterns appear to be identifying user follow-up questions where a user is unsure how to address the panel of advisorsspecific_keywords appears to be a list of words related to the application that is used to essentially patch out matched vague patterns- A lightweight sentiment analysis model or similar is probably enough to address the intended functionality here
- In general, lists of patterns or words will not be easily extensible and will be a problem if non-English inputs are ever to be supported
Objective
The current implementation uses a list of regex patterns, word count, and domain-specific keywords to determine if a user's input needs clarification. This methodology should be improved to avoid using specific words/phrases to infer user intent and meaning.
Initial Implementation Requirements
Other Considerations
vague_patternsappear to be identifying user follow-up questions where a user is unsure how to address the panel of advisorsspecific_keywordsappears to be a list of words related to the application that is used to essentially patch out matched vague patterns