Events Calendar

MS Final Exam – Purbasha Chatterjee

Answer Selection with Attentive Clustering

Question answering forums have been quite effective in improving social interaction and disseminating useful information. Due to the abundance of bad and irrelevant responses, selecting good answers to questions plays a vital role in these forums. In this work, we design an attentive clustering neural network architecture that learns to discriminate good answers from bad answers using community ratings as supervision. Previous research has focused mainly on evaluating individual question-answer pairs and ignored the similarities between different good answers. Taking advantage of the problem setting where there are usually many answers to the question available at the same time and the good answers are similar to each other, we adapt our attentive clustering approach to bias the learner so that similar answers have similar scores. This enables the model to learn the similarities between the answers as well as the relationship between the question and the answer. We implement a wide convolutional neural network that derives attention vectors for both question-answer and answer-answer pairs and computes a normalized score from the pooled representations between the pairs. Empirical results demonstrate that our model outperforms the baseline models and achieves the state-of-the-art performance in multiple benchmark domains.

Major Advisor: Dr. Prasad Tadepalli
Minor Advisor: Debashis Mondal
Committee: Fuxin Li
GCR: Christine Escher

Tuesday, December 11, 2018 at 10:00am to 12:00pm

Kelley Engineering Center, 1007
110 SW Park Terrace, Corvallis, OR 97331

Event Type

Lecture or Presentation

Event Topic


College of Engineering, Electrical Engineering and Computer Science
Contact Name

Calvin Hughes

Contact Email

Google Calendar iCal Outlook

Recent Activity