Mariët Theune

Real versus Template-Based Natural Language Generation: A False Opposition?

This article challenges the received wisdom that template-based approaches to the generation of language are necessarily inferior to other approaches as regards their maintainability, linguistic well-foundedness, and quality of output. Some recent NLG systems that call themselves template-based will illustrate our claims.

Error Detection in Spoken Human-Machine Interaction

Given the state of the art of current language and speech technology, errors are unavoidable in present-day spoken dialogue systems. Therefore, one of the main concerns in dialogue design is how to decide whether or not the system has understood the user correctly. In human-human communication, dialogue participants are continuously sending and receiving signals on the status of the information being exchanged. We claim that if spoken dialogue systems were able to detect such cues and change their strategy accordingly, the interaction between user and system would improve. The goals of the present study are therefore twofold: (i) to find out which positive and negative cues people actually use in human-machine interaction in response to explicit and implicit verification questions and how informative these signals are, and (ii) to explore the possibilities of spotting errors automatically and on-line. To reach these goals, we first perform a descriptive analysis, followed by experiments with memory-based machine learning techniques. It appears that people systematically use negative/marked cues when there are communication problems. The experiments using memory-based machine learning techniques suggest that it may be possible to spot errors automatically and on-line with high accuracy, in particular when focussing on combinations of cues. This kind of information may turn out to be highly relevant for spoken dialogue systems, e.g., by providing quantitative criteria for changing the dialogue strategy or speech recognition engine.

Emotional characters for automatic plot creation

The Virtual Storyteller is a multi-agent framework for automatic story generation. In this paper we describe how plots emerge from the actions of semi-autonomous character agents, focusing on the influence of the characters’ emotions on plot development.

Let's Pretend I Had a Sword

We describe how autonomous character agents that inhabit a story world can make (out-of-character) decisions about its details, filling in the story world as they go. We describe how we model these kind of late commitment decisions, and discuss how we use them to support action selection and to justify the adoption of character goals. Although a rigorous evaluation remains future work, we have implemented the approach presented here and have performed some exploratory testing.

The virtual storyteller

In this paper we present the Virtual Storyteller, a multi-agent framework for dynamic story creation.

GRAPH: the costs of redundancy in referring expressions

We describe a graph-based generation system that participated in the TUNA attribute selection and realisation task of the REG 2008 Challenge. Using a stochastic cost function (with certain properties for free), and trying attributes from cheapest to more expensive, the system achieves overall .76 DICE and .54 MASI scores for attribute selection on the development set. For realisation, it turns out that in some cases higher attribute selection accuracy leads to larger differences between system-generated and human descriptions.

Generating Embodied Information Presentations

The output modalities available for information presentation by embodied, human-like agents include both language and various nonverbal cues such as pointing and gesturing. These human, nonverbal modalities can be used to emphasize, extend or even replace the language output produced by the agent. To deal with the interdependence between language and nonverbal signals, their production processes should be integrated. In this chapter, we discuss the issues involved in extending a natural language generation system with the generation of nonverbal signals. We sketch a general architecture for embodied language generation, discussing the interaction between the production of nonverbal signals and language generation, and the different factors influencing the choice between the available modalities. As an example we describe the generation of route descriptions by an embodied agent in a 3D environment.

A Tabletop Board Game Interface for Multi-User Interaction with a Storytelling System

The Interactive Storyteller is an interactive storytelling system with a multi-user tabletop interface. Our goal was to design a generic framework combining emergent narrative, where stories emerge from the actions of autonomous intelligent agents, with the social aspects of traditional board games. As a visual representation of the story world, a map is displayed on a multi-touch table. Users can interact with the story by touching an interface on the table surface with their fingers and by moving tangible objects that represent the characters. This type of interface, where multiple users are gathered around a table with equal access to the characters and the story world, offers a more social setting for interaction than most existing interfaces for AI-based interactive storytelling.

Social Behaviour in Police Interviews: Relating Data to Theories

We analysed a corpus of enacted police interviews to get insight into the social behaviour of interviewees and police officers in this setting. We (exhaustively) collected the terms used to describe the interactions in those interviews. Through factor analysis, we showed that the theories interpersonal stance, face, and rapport and the meta-concepts information and strategy are necessary to include in a model that captures the social interaction in a police interview. Subsequent validation and relational analysis of the concepts from these theories showed which concepts from these theories are related. This work will be used to inform the construction of a virtual agent acting as a suspect in a training game for police officers.

An Experiment in Improvised Interactive Drama

To inform the design of interactive drama systems, we investigate the experience of an interactor being part of a story that they can have a fundamental influence on. Improvisational theatre might serve as a model for this experience, where there is no pre-scripted plot; each of its actors shares responsibility for the collaborative emergence of a story. This requires a performer attitude from the interactor. We describe an experiment in which improv actors create a story together with subjects who have no improv experience, to find out how we can characterize this experience, and how it might be achieved. Our results support a recent hypothesis that an interactor in interactive drama might be treated as a collaborative performer rather than an (antagonistic) player.