Introduction
Efficient means for capturing and representing computational semantics of data are critical for coping with current limitations of information systems, especially if one wants to make sense of large amounts of information coming from heterogeneous and/or poorly structured resources. Efforts aimed at representing meaning of data in a machine-readable way (i.e., Semantic Web or deductive databases) have achieved some level of success. There are alternatives to the top-down, assertional approaches to semantics, which can work even without (too much) expensive human involvement. One of the most widely and successfully used are distributional semantics models that have been researched within the field of computational linguistics. These models, based on the distributional hypothesis, provide a bottom-up approach to the computational representation of meaning, where the statistical co-occurrence of words in unstructured corpora can provide a basis for the construction of simplified but comprehensive and extensible models of semantic content.
Most of the research activity on distributional semantics has been targeting theoretical and empirical aspects of distributional semantic models with the bulk of the progress been made to date by the natural language processing community. However, a high demand for robust and comprehensive computational models of meaning is present in different areas such as databases, information retrieval, semantic web, artificial intelligence, human-computer interaction, among other areas. This demand, meeting with the availability of mature distributional models and with large-scale unstructured and structured data resources, brings the opportunity of leveraging robust semantic models in all these fields.
Most of the research activity on distributional semantics has been targeting theoretical and empirical aspects of distributional semantic models with the bulk of the progress been made to date by the natural language processing community. However, a high demand for robust and comprehensive computational models of meaning is present in different areas such as databases, information retrieval, semantic web, artificial intelligence, human-computer interaction, among other areas. This demand, meeting with the availability of mature distributional models and with large-scale unstructured and structured data resources, brings the opportunity of leveraging robust semantic models in all these fields.
Objective
DiDaS 2012 aims at connecting distributional semantics with areas that could benefit from a distributional model of meaning. The workshop targets the exploration of both applied and theoretical aspects of these cross-disciplinary interactions. One of the main objectives of the workshop is to bridge the existing gap between distributional semantics and research areas which can strongly benefit from distributional (i.e., bottom-up) models of meaning that would complement the traditional top-down approaches to semantics (e.g., ontologies or database schemata). Additionally, the workshop encourages the participation of domain experts and industry practitioners focused on domain-oriented applications of distributional semantics.
DiDaS 2012 is collocated with the Sixth International Conference on Semantic Computing (ICSC 2012).