IJCAI-05

10th EUROPEAN WORKSHOP ON NATURAL LANGUAGE GENERATION

8-10 August 2005
Aberdeen, Scotland
(following IJCAI-2005 in Edinburgh)

Sponsored by SIGGEN

Workshop Programme

NEW: Workshop Proceedings (8.9Mb)


Natural language generation (NLG) is a subfield of natural language processing that focuses on the generation of written texts in natural languages from some underlying non-linguistic representation of information, generally from databases or knowledge sources. Accomplishing this goal may be envisioned for a number of different purposes, including standardized and/or multi-lingual reports, summaries, machine translation, dialogue applications, and embedding in multi-media and hypertext environments. Consequently, the automated production of language is associated with a large number of highly diverse tasks whose appropriate orchestration in high quality poses a variety of theoretical and practical problems. Relevant issues include content selection, text organization, production of referring expressions, aggregation, lexicalization, and surface realization, as well as coordination with other media.

The workshop continues a biennial series of workshops on natural language generation that has been running since 1987. Previous European workshops have been held at Royaumont, Edinburgh, Judenstein, Pisa, Leiden, Duisburg, Toulouse (2001) and Budapest (2003). The series provides a regular forum for presentation of research in this area, both for NLG specialists and for researchers who may not think of themselves as part of the NLG community.

The 2005 workshop will span the interest areas of natural language generation and Artificial Intelligence, with a special focus on research that integrates NLG with AI, including vision, robotics, intelligent agents, and knowledge discovery. We also encourage papers that investigate the use of state-of-the-art generation technology in real world applications to handle both spoken and text output, and apply language generation techniques to interactive AI systems like communicating robots, to allow the user to enter into short conversations with the system in search for information. There will be demonstrations of working NLG systems, and special sessions for posters describing real-world applications and advanced language technology systems.

Topics of Interest

We welcome papers on formal, corpus-based, implementational and analytical work on conventional NLG topics (realisation, microplanning, etc), and especially papers with a focus on the following themes:

We also welcome discussion on the challenges that these viewpoints pose for generation systems and applications, as well as new ideas and solutions for architectures and general frameworks. Especially, we invite research papers on applying natural language generation in interactive robotics and other AI systems.

Invited Speaker

Kevin Knight (Information Sciences Institute, University of Southern California) will give an invited talk on

Tree Transducers for Machine Translation and Generation

Abstract: Probabilistic finite-state methods have been very successful for natural language processing (NLP) problems like tagging, entity identification, and transliteration. These methods have also been packaged in very useful software toolkits. However, they are not so good for attacking problems with large-scale reordering (translation, generation, paraphrasing, question answering, etc.) and sensitivity to syntax. Over the past three years, new probabilistic tree-based models have been built and tested for a variety of NLP applications. Many of these models turn out to be instances of tree transducers, a formal automata model first described by W. Rounds and J. Thatcher in 1970. This opens up new opportunities for us to marry deeper representations, automata theory, and machine learning, and to create general-purpose tools that can be applied to many NLP problems. This talk will cover new learning algorithms for tree automata, and large-scale natural language experiments.

In addition to the invited talk at ENLG-05, Kevin Knight will give a tutorial on Statistical Machine Translation and Generation the day after the workshop.

Submissions (now closed)

Papers are divided into two types: long papers describing theoretical contributions, and short papers describing ideas or project implementations. Long papers are presented orally and the short papers are presented as posters (not orally). Demos can be associated with short papers, if the presenters bring their own laptop.

Proceedings

All accepted papers (long and short) will be published in the workshop proceedings. If you will need extra copies of the proceedings, please contact Ehud Reiter in advance.

Important dates

Registration

Registration is now open - please follow the registration procedure.

Optional Events

On Sunday 7 August there will be an optional excursion to Royal Deeside and the Scottish Highlands.

On Thursday 11 August Kevin Knight will give a tutorial on Statistical Machine Translation and Generation.

Workshop Organizers

Programme Committee

We thank EPSRC and the University of Aberdeen for their support.