News

I have joined the EmpowerMD within Microsoft Healthcare to work on problems related to NLP + Healthcare. EmpowerMD is a tool that produces clinical notes from doctor-patient conversations.

Our paper on Zero Shot Intern Classification using LSTMs has been accepted for publication at INTERSPEECH 2019. In this paper we propose an LSTM-based encoder that can be used to map user utterances to dialog intents. We show how the LSTM-based method outperforms a DSSM on this task, both for zero shot and few shot intent classification.

Our paper on Automatic Tasks Completion Flows from Web APIs was accepted as a short paper at SIGIR 2019. This paper, co-authored with Seyyed Hadi Hashemi and Imed Zitouni proposed a method for automatically creating task completion flows by chaining together Web APIs. Given a user query, it automatically constructs a flow by chaining together APIs and allowing data to flow from one API to the next. The order in which the APIs are chained is learned by a machine learning model, which generates a set of plans and then scores them by the likelihood of them leading to user satisfaction.