[ Overview | Details | Collaborations | Participants | Presentations | Funding | Publications ]
Kindle: encourage, stimulate, promote, inspire
Supporting Interactive Question Answering (IQA) requires the use of deeper structural, relational and semantic properties of the text. There needs to be a unified knowledge representation of the text, that (1) provides an hierarchical encoding of the structural, relational and semantic properties of the given text, (2) is integrated with learning mechanisms that can be used to induce such information from newly observed raw text, and (3) that is equipped with an inferential mechanism that can be used to support inferences with respect to such representations.
The goal of the Kindle Project is to investigate a unified knowledge representation of this sort based on dependency models augmented with sense tags and semantic role labels, that can be utilized in the automatic analysis of raw text. This will require developing improved tools for automatically inducing these representations, sense tagging and semantic role labeling and, crucially, the integration of this approach with the ability to make inferential links within sentences and between sentences. The inference mechanism must be able to support both learning of and reasoning with the representation. This effort represents a unique combination of Knowledge-based, Statistical and Linguistic Approaches to Question Answering.
The focus will be on:
In the course of this project we will develop an annotated corpus of complex questions and answers of varying levels of difficulty, based on our Gold Standard annotated data, develop theoretical understanding as well as basic tools that will allow the preliminary investigation of these topics, and build the infrastructure for a larger scale study of all the above topics in future stages of this research.