Optimizing Natural Language Generation for Conversational Interfaces

Verena RieserVerena Rieser

Heriot-Watt University and Nuance Research Lab

Friday, June 14, 2013
2:00 PM, ICSI Lecture Hall

 

 

Abstract:

Verena Rieser, "Optimizing Natural Language Generation for Conversational Interfaces," available at http://youtu.be/5cvZJ3AwZ54
Watch this talk on YouTube

I will present a novel approach to Natural Language Generation (NLG) in statistical Conversational Systems, e.g. [Levin & Pieraccini, 1997; Williams & Young, 2007], using a data-driven optimization framework for incremental Information Presentation (IP), where there is a trade-off to be solved between presenting “enough” information to the user while keeping utterances short and understandable. In a case study on recommending restaurants, we show that an optimized IP strategy outperforms a baseline mimicking human behavior in terms of total reward gained. The policy is then also tested with real users, and improves on a conventional hand-coded IP strategy with an up to 9.7% increase in task success. This methodology provides new insights into the nature of the NLG problem, which has previously been treated as a module following dialogue management with limited access to context features.

This type of model is now increasingly used in research applications. For example, optimizing information presentation in recommender systems [Rieser et al., 2010], NLG for direction-giving [Dethlefs and Cuaya ́huitl 2011], personalisation [Janarthanam and Lemon 2010], efficient incremental search [Dethlefs et al. 2012b; 2012a] and non-cooperative negotiation [Rieser et al. 2012].

Bio:

Verena has been an assistant professor in Computer Science at Heriot-Watt University, Edinburgh since 2011. She previously undertook post-doctoral research at the School of Informatics at the University of Edinburgh. She holds a PhD (summa cum laude) from Saarland University (2008) and an MSc from the University of Edinburgh (2004). Her research is at the intersection of Machine Learning and Natural Language Processing, with applications ranging over Multi-Modal Interaction, Spoken Dialogue Systems, and Computational Sustainability. She serves on two boards for the ACL special interest groups in generation and dialogue (SIGgen, SIGdial). She is currently a visiting researcher at Nuance's Research Lab, Sunnyvale.

View slides from this talk.