Question Answering Handler

A custom question answering handler for 📣 stentor, a framework for building conversational applications. It requires a knowledge base to be setup as part of your conversational application

What is 📣 stentor?

An Apache 2.0 licensed open source framework for building conversational applications. It supports a variety of channels, NLU, and is extensible by creating custom handlers.

What is Question Answering?

Question Answering is a Natural Language Processing (NLP) field that focuses on answering questions from a corpus of documents.

How it works?

Stentor will store all knowledgebase results on the session storage on the key knowledge_base_result. QuestionAnsweringHandler will then generate additional keys on the session from these results:

No Answer

A no answer is used when zero results exist from the knowledgebase.

Tags: KB_NO_ANSWER

AI Generated response for No Knowing

GENERATED_NO_ANSWER - We leverage generative AI to produce a no answer response.

Tags: KB_GENERATED_NO_ANSWER

List of Search Results

SEARCH_RESULTS - A list of search results typically based on keyword search

Tags: KB_LIST_OF_RESULTS

Possible Answer from Own Content

SUGGESTED_ANSWER - A possible answer from source text.

Tags: KB_SUGGESTED_ANSWER

Best Answer from the LLM's General Knowledge

GENERAL_KNOWLEDGE - An answer from general knowledge of the generative' AIs large language model.

Tags: KB_GENERAL_KNOWLEDGE

AI Answer from Owned Content

RAG_RESULT - AI Generated from search results TOP_ANSWER - AI determined, these are pulled directly from the source material and are not generated.

Tags: KB_RAG & KB_TOP_ANSWER

Best Curated Answer

TOP_FAQ - A top match within FAQs that have a human written, curated answer. Determined by match confidence score or fuzzy matching.

Tags: KB_TOP_FAQ