Torbjörn Lager

Swi-prolog

SWI-Prolog is neither a commercial Prolog system nor a purely academic enterprise, but increasingly a community project. The core system has been shaped to its current form while being used as a tool for building research prototypes, primarily for knowledge-intensive and interactive systems. Community contributions have added several interfaces and the constraint (CLP) libraries. Commercial involvement has created the initial garbage collector, added several interfaces and two development tools: PlDoc (a literate programming documentation system) and PlUnit (a unit testing environment). In this article, we present SWI-Prolog as an integrating tool, supporting a wide range of ideas developed in the Prolog community and acting as glue between foreign resources. This article itself is the glue between technical articles on SWI-Prolog, providing context and experience in applying them over a longer period.

DEAL: dialogue management in SCXML for believable game characters

In order for game characters to be believable, they must appear to possess qualities such as emotions, the ability to learn and adapt as well as being able to communicate in natural language. With this paper we aim to contribute to the development of believable non-player characters (NPCs) in games, by presenting a method for managing NPC dialogues. We have selected the trade scenario as an example setting since it offers a well-known and limited domain common in games that support ownership, such as role-playing games. We have developed a dialogue manager in State Chart XML, a newly introduced W3C standard, as part of DEAL --- a research platform for exploring the challenges and potential benefits of combining elements from computer games, dialogue systems and language learning.

Pengines: Web Logic Programming Made Easy

When developing a (web) interface for a deductive database, functionality required by the client is provided by means of HTTP handlers that wrap the logical data access predicates. These handlers are responsible for converting between client and server data representations and typically include options for paginating results. Designing the web accessible API is difficult because it is hard to predict the exact requirements of clients. Pengines changes this picture. The client provides a Prolog program that selects the required data by accessing the logical API of the server. The pengine infrastructure provides general mechanisms for converting Prolog data and handling Prolog non-determinism. The Pengines library is small (2000 lines Prolog, 150 lines JavaScript). It greatly simplifies defining an AJAX based client for a Prolog program and provides non-deterministic RPC between Prolog processes as well as interaction with Prolog engines similar to Paul Taraus engines. Pengines are available as a standard package for SWI-Prolog 7. 1

Multilingual Generation and Summarization of Job Adverts: the TREE Project

A multilingual Internet-based employment advertisement system is described. Job ads are submitted as e-mail texts, analysed by an example-based pattern matcher and stored in language-independent schemas in an object-oriented database. Users can search the database in their own language and get customized summaries of the job ads. The query engine uses symbolic case-based reasoning techniques, while the generation module integrates canned text, templates, and grammar rules to produce texts and hypertexts in a simple way.

Part of Speech Tagging from a Logical Point of View

This paper presents logical reconstructions of four different methods for part of speech tagging: Finite State Intersection Grammar, HMM tagging, Brill tagging, and Constraint Grammar. Each reconstruction consists of a first-order logical theory and an inference relation that can be applied to the theory, in conjunction with a description of data, in order to solve the tagging problem. The reconstructed methods are compared along a number of dimensions including ontology, expressive power, mode of reasoning, uncertainty, underspecification, and robustness. It is argued that logical reconstruction of NLP methods in general can lead to a deeper understanding of the knowledge and reasoning involved, and of the ways in which different methods are related.

μ-TBL lite: a small, extendible Transformation-Based Learner

This short paper describes - and in fact gives the complete source for - a tiny Prolog program implementing a flexible and fairly efficient Transformation-Based Learning (TBL) system.

Tagging spoken language using written language statistics

This paper reports on two experiments with a probabilistic part-of-speech tagger, trained on a tagged corpus of written Swedish, being used to tag a corpus of (transcribed) spoken Swedish. The results indicate that with very little adaptations an accuracy rate of 85% can be achieved, with an accuracy rate for known words of 90%. In addition, two different treatments of pauses were explored but with no significant gain in accuracy under either condition.

Statecharts and SCXML for Dialogue Management

The World Wide Web Consortium (W3C) has selected Harel Statecharts, under the name of State Chart XML (SCXML), as the basis for future standards in the area of (multimodal) dialog systems. In this talk, I give a brief introduction to Statecharts and to SCXML, show what it can do (and not do) for someone wanting to use it for dialogue management, describe its relation to other dialogue system components, explain its possible use as a meta-dialogue manager (i.e. as a manager of dialogue managers), as well as some ideas for how to compile other dialogue management languages (such as VoiceXML) into SCXML. If time permits, I will also give some pointers to existing SCXML implementations.

Shallow Processing and Cautious Incrementality in a Dialogue System Front End: Two Steps towards Robustness and Reactivity

This paper presents the design and implementation of a simple and robust dialogue system front end which performs rule-driven, incremental processing of user contributions. We describe how a particular instantiation of the front end can be made to perform a variety of tasks such as part of speech disambiguation, word sense disambiguation, noun phrase detection, and dialogue act recognition, always in an incremental manner. Still, incrementality is never allowed to compromise the accuracy of the disambiguation decisions taken by the system. Incrementality is cautious; when correctness is at stake, decisions are delayed, until more information becomes available. Furthermore, the format of the necessary rules is very simple and uniform across different tasks, and rules can be learned automatically from tagged dialogue corpora, using transformation-based learning.