The Language Technology Group

News and Events

The REMU project - Reliable Multilingual Digital Communication: Methods and Applications.

Third GF Summer School 2013 - Frontiers of Multilingual Technology.

Background

The Language Technology Group at the Department of Computer Science and Engineering was founded in 2001. It built on the Department's earlier competence areas:

• logic and type theory
• functional programming
• compiler construction

With this background, the group was in the beginning very much profiled towards precision-oriented tasks, rather than to wide coverage

• multilingual generation
• technical translation
• natural language interfaces
• spoken dialogue systems

In more recent years, the efforts have been extended to the creation of tools and resources usable in all kinds of language technology tasks:

• grammar formalism implementations
• generation of speech recognizer language models
• lexicon building tools
• extraction of lexica from raw text data
• extraction of inflection rules from raw text

The main characteristics identifying our group are

• multilinguality: theories and tools that apply to many languages simultaneously
• compiler technology: many of our NLP ideas are generalizations from modern compilers; we have also developed compiler tools inspired by NLP

Currently (February 2008), the group has 8 members with a PhD and 4 PhD students. Some of the senior members have their principal affiliations at the Departments of Linguistics and Swedish Language.

People

• Krasimir Angelov, PhD.
• John J. Camilleri, PhD student.
• Ramona Enache, PhD student.
• Thomas Hallgren, PhD, Research Engineer.
• Muhammad Humayoun, PhD student in joint supervision with the University of Savoie.
• Inari Listenmaa, Phd student.
• Peter Ljunglöf, Docent.
• Bengt Nordström, Professor.
• Aarne Ranta, Professor.

Alumni

• Björn Bringert, Former PhD student, now at Google in London.
• Johan Boye, Docent at University of Gothenburg, Associate Professor at KTH in Stockholm.
• Håkan Burden, Former PhD student.
• Olga Caprotti, Former Project Manager of MOLTO.
• Dana Dannélls, PhD student in joint supervision with the Department of Swedish Language.
• Markus Forsberg, Former PhD student, now in at the Department of Swedish Language.
• Harald Hammarström, Former PhD student, now at the University of Nijmegen.
• Kristofer Johannisson, Former PhD student, now at Apptus in Lund.
• Janna Khegai, Former PhD student, Now at Goldman Sachs in London.
• Shafqat Virk, former PhD student.

Multilingual grammars

Grammatical Framework (GF), grammaticalframework.org, is a multilingual grammar formalism based on the idea of a shared abstract syntax and mappings between the abstract syntax and concrete languages. GF has hundreds of users all over the world.

The GF Resource Grammar Library, grammaticalframework.org/lib/doc/synopsis.html, implements the morphology (inflection) and basic syntax (phrase structure) of 16 languages: Bulgarian, Catalan, Danish, Dutch, English, Finnish, French, German, Italian, Norwegian, Polish, Romanian, Russian, Spanish, Swedish, and Urdu. These resources are, freely available as open-source software. More languages are under construction, in both in-house and external projects.

Embedded grammars are parsing and generation programs compiled from GF grammars and usable as parts of programs written in other languages: Haskell, Java, and JavaScript.

Multilingual grammar applications

The Numeral Translator, http://www.cse.chalmers.se/alumni/bringert/gf/translate/, is a demo of embedded grammars in Java. It translates number words between 88 languages.

The Letter Editor http://www.cse.chalmers.se/alumni/markus/gramlets/letter-applet.html is another demo of embedded grammars in Java. It allows the user to write a letter in languages she doesn't know while viewing it in a language she knows.

The Pizza ordering system, http://www.cse.chalmers.se/alumni/bringert/xv/pizza/pizza-movie-large.html, is a demo of integrated speech language model, JavaScript, and VoiceXML generation from GF grammars. In a browser supporting the used WC3 standards (e.g. Opera on Windows), the user can construct an order by using spoken language.

Morphology and lexicon

Lexicon Extraction from Raw Text Data, www.cse.chalmers.se/alumni/markus/extract, a tool for collecting a morphological lexicon by the use of inflection paradigms.

Functional Morphology, www.cse.chalmers.se/alumni/markus/FM, a Haskell library for developing inflection engines and morphological lexica.

Unsupervised Learning of Morphology, www.cs.chalmers.se/~harald2/lic.pdf, a technique usable for languages with scarce resources.

Language identification

A Fine-Grained Model for Language Identification, www.cs.chalmers.se/~harald2/id_inews07.pdf, a technique usable for short passages and language switching.

Compiler technology

The BNF Converter (BNFC) is a high-level multi-backend compiler tool, inspired by GF. It has thousands of users and is included in Linux distributions such as Debian and Ubuntu.