Malte Gabsdil

Put My Galakmid Coin into the Dispenser and Kick It: Computational Linguistics and Theorem Proving in a Computer Game

We combine state-of-the-art techniques from computational linguisticsand theorem proving to build an engine for playing text adventures,computer games with which the player interacts purely through naturallanguage. The system employs a parser for dependency grammar and ageneration system based on TAG, and has components for resolving andgenerating referring expressions. Most of these modules make heavy useof inferences offered by a modern theorem prover for descriptionlogic. Our game engine solves some problems inherent in classical textadventures, and is an interesting test case for the interactionbetween natural language processing and inference.

A Tool for Supporting Progressive Refinement of Wizard-of-Oz Experiments in Natural Language

Wizard-of-Oz techniques are an important method for collecting data about the behavior of students in tutorial dialogues with computers, especially when the interaction is done in natural language. Carrying out such experiments requires dedicated tools, but the existing ones have some serious limitations for supporting the development of systems with ambitious natural language capabilities. In order to better meet such demands, we have developed DiaWoZ, a tool that enables the design and execution of Wizard-of-Oz experiments to collect data from dialogues and to evaluate components of dialogue systems. Its architecture is highly modular and allows for the progressive refinement of the experiments by both designing increasingly sophisticated dialogues and successively replacing simulated components by actual implementations. A first series of experiments carried out with DiaWoZ has confirmed the need for elaborate dialogue models and the incorporation of implemented components for subsequent experiments.

Natural language and inference in a computer game

We present an engine for text adventures - computer games with which the player interacts using natural language. The system employs current methods from computational linguistics and an efficient inference system for description logic to make the interaction more natural. The inference system is especially useful in the linguistic modules dealing with reference resolution and generation and we show how we use it to rank different readings in the case of referential and syntactic ambiguities. It turns out that the players utterances are naturally restricted in the game scenario, which simplifies the language processing task.

Classifying Recognition Results for Spoken Dialog Systems

This paper investigates the correlation between acoustic confidence scores as returned by speech recognizers with recognition quality. We report the results of two machine learning experiments that predict the word error rate of recognition hypotheses and the confidence error rate for individual words within them.