Brussels, December 4, 1996.
Present at the meeting: Bill Black, Walter Daelemans, Laurence Danlos, Joakim Nivre, Koenraad de Smedt (chair), Hans Uszkoreit, Felisa Verdejo.
ACO*HUM is a SOCRATES TNP (Thematic Network Project) which started on September 1, 1996, and is expected to be active for three years. The theme of ACO*HUM is an investigation of the way in which advanced computing is changing both the content and the process of scholarship in the humanities.
The working group (WG for short; earlier called area committee) on Computational Linguistics and Language Engineering will generate ideas on how this theme can be addressed on a European scale in its competence area.
The WG acknowledges that valuable experience with curriculum work on a European scale has been gained in ICP-95-NL-1022/09 on Natural Language Processing, presently coordinated by Andy Way in Dublin. Several of the WG members were involved in this network project. A summary of the work was written up by Walter Daelemans in T.A.L. Vol. 37(1), pp. 87-91, "Enseignement du TAL". This article contains a table sketching possible core courses and their contents.
The WG recognizes that the identification of core courses is important in curriculum work and may ease the recognition of courses in other countries. At the same time, the identification of local specializations should not be neglected.
The WG finds it desirable for the present TNP to build on this work but would like to place this work in a wider perspective by exploring possibilities for broadening the European scale, increasing the innovation potential, working towards criteria and guidelines, and launching tests.
As part of its initial activities, ACO*HUM will make an agenda of curricula, people, places, resources, problems and ongoing activities. Among the data to be gathered, the WG would like to see a list of courses with their length and course material used, resources and tools for teaching, national summer schools, and links with research. Overviews of programmes in computational linguistics will be gathered. It would also be interesting to gather information on the incorporation of work experience (placement) in curricula and on the accreditation of courses and curricula by professional bodies.
This material will in a later stage be helpful for the identification of common modules, the development and use of course materials on a European scale, and other measures towards an innovative approach to CL teaching in Europe on a wider scale. In this process, the applicability of the preliminary guidelines for core courses by the network on Natural Language Processing will have to be evaluated and examples of best practice will be identified.
ACO*HUM intends to write country profiles describing and comparing the situation in the various participating countries. From a management perspective, it may be difficult to get equally detailed analyses from the different countries. The involvement of a national coordinator for each country seems the best way to take on this task (Note: ELSNET reports good experience with this approach). An effort will be made to involve various national associations, including:
If possible, use will be made of earlier work by the European Commission on equivalences of degrees.
Among its partners, ACO*HUM already includes the ALLC, EACL and FOLLI. The EACL has not been significantly active in education, but has expressed an interest. The ALLC has some activities targeted at students and wishes to expand. FOLLI organizes successful courses at the graduate and postgraduate level. The WG advises to obtain participation from other groups such as ELRA and ELSNET (which has a training and mobility task group).
The possible future link with Central European countries must be investigated, as several of these countries have a serious interest and present activities in CL.
The WG is aware of the fact that the beginning integration of advanced computing in linguistics curricula has led to great diversity and lack of coordination at a European scale. In order to strengthen the European impact of many local initiatives, it is therefore regarded as an important goal to develop Europe-wide recommendations for common elements in curricula, without denying the importance of local specializations. It is not deemed necessary to unify all basic requirements at every place, but those requirements need to become more transparent on a transnational basis.
The difference between using tools (e.g. consulting corpora, using grammar-writing workbenches) and making tools (building computer programs) is identified as an important guiding perspective. The increasing use of tools can be stimulated among general linguistics students, while tool making, in the form of a computational methodology and technology, needs to be strengthened in specialized programmes for computational linguistics.
It is agreed that the areas of grammars and parsing would be good starting points for an inventory of what is taught in Europe.
The delivery of courses using distance education technology (including web technology) may offer advantages due to distribution of effort, lower cost of maintenance, and flexibility by extension. These factors are important for CL, which has a growing student basis but is faced with a shortage of teaching staff and teaching materials.
In how far teaching materials need to be localized with respect to the language of delivery is a problem which needs to be investigated. Different languages not only present problems, but also have added value since they prepare for student exchange. Useful contributions may consist of the construction of multi- and monolingual glossaries of linguistic and CL terminologies.
The construction of a European textbook on CL is a challenging and interesting task but would require great effort concentrated in time and place. As an alternative, an incremental, flexible and distributed approach needs to be considered. Web technology can be used to gather materials in a loose and piecemeal way with various coexisting superstructures. The didactic approach would incorporate connections to various sites for hands-on experience with computational resources (corpora, lexicons, texts in different languages, etc.). Guidelines for the development of modules within this approach need to be worked out.
One topic or module could serve in a feasibility study. Computational morphology (with the use of e.g. PC-Kimmo) might be a suitable candidate.
Despite the fact that ACO*HUM is seriously underfinanced, the WG does not lack ideas for working with the theme. Among the interesting topics for future activities in the TNP are the following:
The joint call for proposals, open to several European programmes including SOCRATES, could be used to support future activites by the ACO*HUM network. Even though ACO*HUM does not have sufficient resources to be a coordinator, it could be interested in several types of projects, provided sufficient funding.
The next meeting will be in Madrid, during or adjacent to the ALC/EACL conference. Felisa Verdejo will host the meeting.
Minutes written by Koenraad de Smedt
"Le sort des nations qui négligeront
la science et les savants est marqué pour la décadence."
(Albert I, roy des Belges, Oct. 1, 1927).