Decision Support Question Answering - overview
The Natural Language Analytics team pursues basic research in Natural Language Processing. The goal of the research is to create a new paradigm for Question Answering systems as active collaborators in information gathering for complex decision making.
Existing question answering (QA) systems:
- assume that the decision problem can be directly formulated into a question and that the answer can directly inform the decision
- are one-shot systems that can’t build on prior interactions
- are black boxes that do not expose their interpretation and process
- don’t learn/adapt from interactions with an end user
To address the limitations of existing QA systems, the team is conducting research into a new kind of exploratory question answering system with the following properties:
- Contextual: takes into consideration arbitrary context representing background information that will inform question answering behavior
- Transparent: exposes elements of interpretation so it is clear why the system behaves a certain way
- Guidable: allows intervention and guidance to improve question answering relevance
- Iterative: builds on a persistent context, taking into consideration previous statements, questions and their results
- Adaptable: learns from user behavior, allowing improvement on interpretations of user input and the question answering results produced
Research Areas
Artificial Intelligence