iLFQA
Repository containing all source code and documentation of iLFQA: intelligent Longform Question Answering
We present an efficient and accurate long-form question-answering platform, dubbed iLFQA (intelligent Long-Form Question Answering). The purpose of iLFQA is to function as a plat- form which accepts unscripted, human generated questions and efficiently produces a semanti- cally meaningful, explanatory, and accurate response. iLFQA introduces and integrates a num- ber of natural language processing as well as supervised learning modules to provide answers to questions from a pre-defined and sizable knowledge corpus including numerous long-documents (e.g., books). iLFQA leverages Transformer Architecture, Zero-Shot Classification, and Term Fre- quency{Inverse Document Frequency to rapidly select paragraphs from a series of texts that are most relevant to a given question. Once the paragraphs have been selected iLFQA synthesizes an appropriate textual response for the question and returns it to the user.
All source code for project and the source code used to produce individual results is available in this repository. The results produced by each iteration of development is available in the individual folders.