Kallirroi Georgila

Reducing working memory load in spoken dialogue systems

We evaluated two strategies for alleviating working memory load for users of voice interfaces: presenting fewer options per turn and providing confirmations. Forty-eight users booked appointments using nine different dialogue systems, which varied in the number of options presented and the confirmation strategy used. Participants also performed four cognitive tests and rated the usability of each dialogue system on a standardised questionnaire. When systems presented more options per turn and avoided explicit confirmation subdialogues, both older and younger users booked appointments more quickly without compromising task success. Users with lower information processing speed were less likely to remember all relevant aspects of the appointment. Working memory span did not affect appointment recall. Older users were slightly less satisfied with the dialogue systems than younger users. We conclude that the number of options is less important than an accurate assessment of the actual cognitive demands of the task at hand.

Being Old Doesn’t Mean Acting Old: How Older Users Interact with Spoken Dialog Systems

Most studies on adapting voice interfaces to older users work top-down by comparing the interaction behavior of older and younger users. In contrast, we present a bottom-up approach. A statistical cluster analysis of 447 appointment scheduling dialogs between 50 older and younger users and 9 simulated spoken dialog systems revealed two main user groups, a “social” group and a “factual” group. “Factual” users adapted quickly to the systems and interacted efficiently with them. “Social” users, on the other hand, were more likely to treat the system like a human, and did not adapt their interaction style. While almost all “social” users were older, over a third of all older users belonged in the “factual” group. Cognitive abilities and gender did not predict group membership. We conclude that spoken dialog systems should adapt to users based on observed behavior, not on age.

Using Integer Linear Programming for Detecting Speech Disfluencies

We present a novel two-stage technique for detecting speech disfluencies based on Integer Linear Programming (ILP). In the first stage we use state-of-the-art models for speech disfluency detection, in particular, hidden-event language models, maximum entropy models and conditional random fields. During testing each model proposes possible disfluency labels which are then assessed in the presence of local and global constraints using ILP. Our experimental results show that by using ILP we can improve the performance of our models with negligible cost in processing time. The less training data is available the larger the improvement due to ILP.

Automatic annotation of context and speech acts for dialogue corpora

Richly annotated dialogue corpora are essential for new research directions in statistical learning approaches to dialogue management, context-sensitive interpretation, and context-sensitive speech recognition. In particular, large dialogue corpora annotated with contextual information and speech acts are urgently required. We explore how existing dialogue corpora (usually consisting of utterance transcriptions) can be automatically processed to yield new corpora where dialogue context and speech acts are accurately represented. We present a conceptual and computational framework for generating such corpora. As an example, we present and evaluate an automatic annotation system which builds ‘Information State Update’ (ISU) representations of dialogue context for the Communicator (2000 and 2001) corpora of human–machine dialogues (2,331 dialogues). The purposes of this annotation are to generate corpora for reinforcement learning of dialogue policies, for building user simulations, for evaluating different dialogue strategies against a baseline, and for training models for context-dependent interpretation and speech recognition. The automatic annotation system parses system and user utterances into speech acts and builds up sequences of dialogue context representations using an ISU dialogue manager. We present the architecture of the automatic annotation system and a detailed example to illustrate how the system components interact to produce the annotations. We also evaluate the annotations, with respect to the task completion metrics of the original corpus and in comparison to hand-annotated data and annotations produced by a baseline automatic system. The automatic annotations perform well and largely outperform the baseline automatic annotations in all measures. The resulting annotated corpus has been used to train high-quality user simulations and to learn successful dialogue strategies. The final corpus will be made publicly available.

Single-Agent vs. Multi-Agent Techniques for Concurrent Reinforcement Learning of Negotiation Dialogue Policies

We use single-agent and multi-agent Reinforcement Learning (RL) for learning dialogue policies in a resource allocation negotiation scenario. Two agents learn concurrently by interacting with each other without any need for simulated users (SUs) to train against or corpora to learn from. In particular, we compare the Qlearning, Policy Hill-Climbing (PHC) and Win or Learn Fast Policy Hill-Climbing (PHC-WoLF) algorithms, varying the scenario complexity (state space size), the number of training episodes, the learning rate, and the exploration rate. Our results show that generally Q-learning fails to converge whereas PHC and PHC-WoLF always converge and perform similarly. We also show that very high gradually decreasing exploration rates are required for convergence. We conclude that multiagent RL of dialogue policies is a promising alternative to using single-agent RL and SUs or learning directly from corpora.

Developing INOTS to support interpersonal skills practice

The Immersive Naval Officer Training System (INOTS) is a blended learning environment that merges traditional classroom instruction with a mixed reality training setting. INOTS supports the instruction, practice and assessment of interpersonal communication skills. The goal of INOTS is to provide a consistent training experience to supplement interpersonal skills instruction for Naval officer candidates without sacrificing trainee throughput and instructor control. We developed an instructional design from cognitive task analysis interviews with experts to serve as a framework for system development. We also leveraged commercial student response technology and research technologies including natural language recognition, virtual humans, realistic graphics, intelligent tutoring and automated instructor support tools. In this paper, we describe our methodologies for developing a blended learning environment, and our challenges adding mixed reality and virtual human technologies to a traditional classroom to support interpersonal skills training.1 2

Time-offset interaction with a holocaust survivor

Time-offset interaction is a new technology that allows for two-way communication with a person who is not available for conversation in real time: a large set of statements are prepared in advance, and users access these statements through natural conversation that mimics face-to-face interaction. Conversational reactions to user questions are retrieved through a statistical classifier, using technology that is similar to previous interactive systems with synthetic characters; however, all of the retrieved utterances are genuine statements by a real person. Recordings of answers, listening and idle behaviors, and blending techniques are used to create a persistent visual image of the person throughout the interaction. A proof-of-concept has been implemented using the likeness of Pinchas Gutter, a Holocaust survivor, enabling short conversations about his family, his religious views, and resistance. This proof-of-concept has been shown to dozens of people, from school children to Holocaust scholars, with many commenting on the impact of the experience and potential for this kind of interface.

The MATCH corpus: a corpus of older and younger users’ interactions with spoken dialogue systems

We present the MATCH corpus, a unique data set of 447 dialogues in which 26 older and 24 younger adults interact with nine different spoken dialogue systems. The systems varied in the number of options presented and the confirmation strategy used. The corpus also contains information about the users’ cognitive abilities and detailed usability assessments of each dialogue system. The corpus, which was collected using a Wizard-of-Oz methodology, has been fully transcribed and annotated with dialogue acts and “Information State Update” (ISU) representations of dialogue context. Dialogue act and ISU annotations were performed semi-automatically. In addition to describing the corpus collection and annotation, we present a quantitative analysis of the interaction behaviour of older and younger users and discuss further applications of the corpus. We expect that the corpus will provide a key resource for modelling older people’s interaction with spoken dialogue systems.

New Dimensions in Testimony: Digitally Preserving a Holocaust Survivor’s Interactive Storytelling

We describe a digital system that allows people to have an interactive conversation with a human storyteller (a Holocaust survivor) who has recorded a number of dialogue contributions, including many compelling narratives of his experiences and thoughts. The goal is to preserve as much as possible of the experience of face-to-face interaction. The survivor’s stories, answers to common questions, and testimony are recorded in high fidelity, and then delivered interactively to an audience as responses to spoken questions. People can ask questions and receive answers on a broad range of topics including the survivor’s experiences before, after and during the war, his attitudes and philosophy. Evaluation results show that most user questions can be addressed by the system, and that audiences are highly engaged with the resulting interaction.

Simulating the Behaviour of Older versus Younger Users when Interacting with Spoken Dialogue Systems

In this paper we build user simulations of older and younger adults using a corpus of interactions with a Wizard-of-Oz appointment scheduling system. We measure the quality of these models with standard metrics proposed in the literature. Our results agree with predictions based on statistical analysis of the corpus and previous findings about the diversity of older peoples behaviour. Furthermore, our results show that these metrics can be a good predictor of the behaviour of different types of users, which provides evidence for the validity of current user simulation evaluation metrics.