Imed Zitouni

woman and skyscrape computer

I have been a member of the Relevance and Measurement team of Microsoft since October 2012. I work in improving Bing’s quality by providing better metrics. My current research interest is in the area of information retrieval (IR) focusing on the use of statistics and machine learning techniques to develop web scale offline and online metrics for search engines. I am also interested in using Natural Language Processing (NLP) technologies to add a layer of semantics and understanding to search engines. I am a believer that next generation search engines will be based on dialog and language understanding.

Prior to joining Microsoft, I worked as a research member for IBM Research in the Multilingual NLP group for almost a decade, where I served as team-lead in several NLP projects. During this time, I was again focusing on NLP, IR, machine translation, speech-recognition, language modeling and machine learning. I was a key member of several government projects including the GALE (Global Autonomous Language Exploitation) program. Prior to IBM I spent six years as a research member at Bell Laboratories, Lucent Technologies, working on language modeling, speech recognition, spoken dialog systems and speech understanding. During the last few years at Bell Labs I was also in charge of the speech and natural language call routing activities leading a small team of very talented researchers. Prior to Bell Labs, I was involved in the startup experience at DIALOCA in Paris, France, working on e-mail steering and language modeling. I also served as a temporary assistant professor at the University of Nancy 1, France. I received my M.Sc. and Ph.D. with the highest-honors from the University-of-Nancy1 France. In 1995, I obtained a MEng degree in computer science from ENSI in Tunisia.

I am a senior member of IEEE, serving as a member of the IEEE Speech and Language Processing Technical Committee (99-11), the Information Officer of the ACL SIG on Semitic-Languages, associate editor of TALIP ACM journal and a member of ISCA and ACL. I served as chair and reviewing-committee-member of several conferences and journals. I am the author/co-author of more than 80 papers in international conferences and journals.

My recent book is “Multilingual Natural Language Processing Application: from Theory to Practice”, by Prentice Hall.

Abstract: Voice-controlled intelligent personal assistants, such as Cortana, Google Now, Siri and Alexa are increasingly becoming a part of users’ daily lives, especially on mobile devices. They allow for a radical change in information access, not only in voice control and touch gestures but also in longer sessions and dialogues preserving context, necessitating to evaluate their effectiveness at the task or session level. In this talk, I will present an approach that can evaluate different tasks in voice-activated intelligent assistants. The proposed approach uses implicit feedback from users to predict whether users are satisfied with the intelligent assistant as well as its components, i.e., speech recognition and intent classification. Using this approach, we can potentially evaluate and compare different tasks within and across intelligent assistants according to the predicted user satisfaction rates. The approach is characterized by an automatic scheme of categorizing user-system interaction into task-independent dialog actions, e.g., the user is commanding, selecting, or confirming an action. We use the action sequence in a session to predict user satisfaction and the quality of speech recognition and intent classification.