Terminology has a unique status in language and communication. Theoretically, it is within the tension between flexibility of natural language and the rigidity of artificial sign system. This point is elegantly stated by Dr. Alain Rey, the cerebrated editor of Le Robert:
To the extent that a terminological system, even if it matches a coherent conceptual system, is incapable of reflecting its internal relationships, terminology is autonomous with respect to epistemology. To the extent that a terminological system, even if it is formed from unmotivated and common language words, denotes a conceptual system and exists only for denoting it, terminology is autonomous with respect to linguistics.
Within the context of NLP, we can replace "epistemology" with "ontology" as well.
Reflecting this theoretical status, terms are treated in a specific way in human language practice. In technical translation pipeline, terms are not "translated" but relevant target language terms are looked up and used, as "mistranslation" can cause grave consequences. Many organisations, including such public institutions as EU, WIPO and NLM and private LSPs, construct and maintain terminologies. Translation quality assurance scheme identify terminology related issues as one of the focal checking points. Terminologies also provide an important resources for education and knowledge transfer.
Although substantial ratio of terms are linguistically so-called multiword expressions, the requirements and desiderata for handling terms as well as their status in language practice pipelines are different from most other multiword expressions such as idioms.
Computational terminology, if it is to make an in vivo contribution to the human communication ecosystem, need to take into account this uniqueness of terminology, at every stage of research starting from defining problems to be solved, methods to be adopted and developed to evaluation schemes to be used.
During four years since the 5th Workshop on Computational Terminology (Computerm 2016) was held, advancement in distributional representations and deep learning has changed, at least on surface, the major NLP scene. What about terminology processing? This issue is yet to be fully explored or discussed. For instance, while NMT improved greatly the target language (TL) fluency, it is sometimes reported that the quality of terminology translation is on a par with SMT. Given the unique status of technical terms in communication and language practice, there is much to delve into computational terminology, in the face of recent development of deep learning based NLP technologies. It was time now to share not only the technical development in computational terminology but also how new technical developments fare with the theoretical status of and practical requirements for terminology-related applicational missions.
Research in computational terminology consolidated itself, arguably, in early 1990s, when a series of papers in monolingual and multilingual extraction of terms from textual corpora appeared in main computational linguistic conferences.
Even before that period, during 1980s, need for a common forum to exchange information among researchers in terminology was well recognised. In 1988, the first of an important conference in the field of terminology, i.e. Terminology and Knowledge Engineering (TKE), was held in Trier, Germany. TKE was and is more oriented to concepts organisation and knowledge engineering, reflecting an important feature of terms and terminology.
Separately in 1995, a group of French researchers in artificial intelligence and natural language processing held the first of the conference series Terminologie et Intelligence Artificielle (TIA) in Villetaneuse, France. Started as mostly for francophone researchers, TIA quickly grew out as an international forum. Both conferences continue as an important forum, with slightly different emphasis and geographical balance, for research in terminology that cover computational approach. In 1997, a Japanese group started a shared task of automatic term recognition, the result of which was reported in NTCIR TMREC session held in 1999.
It is within these environment that the First International Workshop on Computational Terminology (Computerm 1998) was held in 1998 at the occation of COLING-ACL 1998, in Motreal, Canada. Computerm is intended to consolidate the tradition of terminological research in relation to computational linguistics and put it explicitly within the context of natural language processing. As such, it is at the same time closely related and complimentary to TKE and TIA. Since the first Computerm workshop, it continued, though irregularly, as a workshop of COLING or LREC. We are pleased to announce that the 6th Workshop on Computational Terminology (COMPUTERM 2020) is to be held as a full day workshop at the occasion of LREC 2020 in Marseille, France.