Invited Speakers

Abstract of Talk 1

Slacker Semantics: Why Superficiality, Dependency and Avoidance of Commitment Can Be the Right Way to Go

Standard linguistic approaches to compositional semantics need adaptation for use in broadcoverage computational processing. I’ll argue that an ideal compositional semantic representation should capture all and only the information available from syntax and productive morphology, allowing for shallow as well as deep syntactic analysis and avoiding over-commitment when ambiguity cannot be resolved. This leads to ‘surfacy’ semantic representations which must nevertheless allow for enrichment by deeper analysis (including lexical semantics and anaphora resolution), support (robust) inference and have a logically-sound interpretation. Other desiderata are compatibility with generation, statistical ranking of analyses/realisations and distributional semantics, ease of use for consumers of the representation and for human annotators, and cross-linguistic adequacy.

Several current approaches to computational compositional semantics meet at least some of
these goals, but in this talk I’ll concentrate on work within the DELPH-IN community using Minimal
Recursion Semantics (MRS) and Robust MRS. I’ll give an overview to illustrate the extent to which I
believe we’re currently meeting these aims. I’ll go on to show that, under certain assumptions about the grammar, an interconversion is possible between MRS and a variable-free semantic dependency notation (DMRS), which removes redundancy, supports additional forms of underspecification and makes the representation easier for humans to process.

Abstract of Talk 2

NLP and the Humanities: The Revival of an Old Liaison

The humanities and the field of natural language processing (NLP) have always had common playgrounds. The liaison was never constrained to linguistics; also philosophical, philological and literary studies have had their impact on NLP, and there have always been dedicated conferences and journals for the humanities and the NLP community. The more recent emergence of the field of Computational Humanities demonstrates that the potential for mutual impact has gained in strength and diversity and that important lessons can be learned for other application areas than the humanities. A renewed liaison with the now computational humanities can help NLP to set up an innovative research agenda which covers a wide range of topics including semantic analysis, integration of multimodal information, language-based interaction, performance evaluation, service models, and usability studies.

As will be discussed, the further and combined exploration of these topics will help to develop
an infrastructure that will allow content and data driven research domains in the humanities to renew their field and to exploit the potential coming from the wide-scale digitization initiative. To
name a few: art history, media studies, oral history, archeology, archiving studies have needs that can be served in novel ways by the matured branches that NLP offers today. Based on an overview of initiatives, an analysis of the characteristics of the interaction between the humanities and NLP, and some telling examples, it will be outlined what could be done to increase the chances for a bright future for the old ties, and how other domains can benefit as well.