Skip navigation

Deep Learning and Community Question Answering

Event Date: March 24, 2017
Speaker: Mugizi Robert Rwebangira
Speaker Affiliation: Assistant Professor, Howard University, Washington, DC
Type: CNSIP Research Area Seminar
Time: 10:30am
Location: MSEE 239
Contact Name: Professor David J. Love
Contact Phone: 765-49-66797
Contact Email:
Priority: No
School or Program: Electrical and Computer Engineering
College Calendar: Show


 One of the major problems with community question answering systems such as Yahoo! Answers is that a significant percentage of questions are left unanswered. One way to reduce this percentage of unanswered questions is to make use answers to past resolved questions to resolve similar questions. Most of the proposed methods that use this approach rely on extracted features and typically feature engineering requires external knowledge sources that might be difficult and computationally expensive to obtain. Recently, deep learning approaches have been shown to be able to automatically learn optimal feature representations in given tasks in speech recognition, computer vision, and natural language processing. In this work we propose a convolutional neural network architecture that compares a given question with past resolved questions in entity-rich categories such as sports and entertainment. In these categories, there is a high usage of entities and entity variations; hence, semantically similar questions are often worded differently. Our approach identifies entity name variations in questions and learns the optimal representations of question pairs and how to relate them using a similarity function. We test our model on a dataset of questions from Yahoo! Answers.


Robert Rwebangira completed his PhD in computer science from Carnegie Mellon University in 2008 under the supervision of Avrim Blum and John Lafferty. His dissertation was on techniques for semi-supervised learning. He is currently an Assistant Professor of Computer Science at Howard University with a research focus on applications of machine learning to natural language processing and computational biology.