Section: Application Domains
As we have said, much of the work of the LED team revolves around the development of wide coverage grammars that yield semantic representations, and linking such grammars with inference tools. That is, most of our current work is building both the theory and the resources and software needed to support detailed work on semantics. Until such resources are developed, it is somewhat premature to speak of the application domains — though (judging by international experience for better explored languages such as English) it is likely that our current foundational work will lead to applications in such domains as question answering, intelligent document retrieval, and so on.
In the meantime, LED is exploring the possibilities of using lighter methods (that is, methods making less use of linguistic knowledge and inference) for applications. Application domains we are currently exploring or developing include:
Handling multilingual information in multimedia settings.
Extraction of timelines and textual summaries from online encyclopedias.
Applications of natural language processing to language education.
Application of natural language processing to computing for the humanities.
Applications of natural language processing in the publishing industry.
The applications in multilinguality are well established (notably because of the work of LED member Samuel Cruz-Lara in the setting of the Passepartout and MLIF projects). The other four topics named are newer, and we are currently looking for industrial partners to collaborate with. It should be remarked that in previous years, LED was heavily involved with industry because of its focus on standardisation. However, with the departure of Laurent Romary, standardisation is no longer a core activity, hence our interest in finding new partners and new applications. We are currently discussing with Conitiv Synchron (a Nancy-based firm specialising in the development of software for accelerated language learning; see http://www.conitiv-synchron.com/ ) the possibility of using natural language processing techniques in this domain. A Masters student (supervised by Patrick Blackburn) will start working with Conitiv Synchron on this topic early in 2007.