Complex answer retrieval for question answering systems

2018 COE Engineering Design Project (EB02)

---

Faculty Lab Coordinator

Ebrahim Bagheri

Topic Category

Software / Data Engineering

Preamble

While current retrieval systems provide good results for phrase-level retrieval of simple facts, they are less effective in assembling facts into an information package. This project will focus on answering more complex information needs with longer answers through assembling of different information sources. Much like Wikipedia pages that synthesize knowledge that is globally distributed, we are interested in building a retrieval system that collects relevant information from an entire corpus, creating synthetically structured documents by collating retrieved results.

Objective

The objective of this project will be to assemble an information package or the so called complex answer in response to an information need in question answering systems.

Partial Specifications

Students will learn about cutting edge techniques in Natural Language Processing and Information Retrieval.

Suggested Approach

The details of the task are presented in http://trec-car.cs.unh.edu/

Group Responsibilities

The group will be responsible to replicate the work presented in the baseline paper, work with the FLC to further enhance the algorithm, implement the enhancements, perform experiments and report on the findings. All students will be responsible for understanding the baseline paper and participating in group brainstorming sessions.

Student A Responsibilities

Preparation of the dataset for the experiments and assist with the implementation of the enhancements to the baseline.

Student B Responsibilities

Implementation and Replication of the baseline code.

Student C Responsibilities

Leading the implementation of the enhanced code as well as running the experiments.

Course Co-requisites

N/A