EPFL.3

ScienceWISE (phase 2)

Long Title: Web-based Interactive Semantic Environment for e-Science - Phase 2
Leading
Organization:
EPF Lausanne
Domain: ELS
Status: finished
Start Date: 01.01.2012
End Date: 31.12.2012
Project Leader: K. Aberer
Deputy Project Leader: A. Boyarsky
Website: http://sciencewise.info

ScienceWISE - platform for creation of virtual organizations of scientists, working together on a dynamical generation of professional field-specific ontologies - has been further extended with regard to functionalities, to Arxiv.org user support and to other relevant scientific communities.

(See also Phase 1 and Phase 2 continuation )

Results

see Phase 2 continuation (extension)


Goals

During phase 2 ScienceWISE will extend its services into other fields in science (mathematics, biophysics, environmental and computer science) and add depth to analysis of texts (context and similarity analysis, equation search, clustering of articles that contribute to the same topic).
Support for the Arxiv.org users will be provided and functionalities of the ScienceWISE system and reaction for the users' feedback further extended.
Computer Science techniques, related to "information extraction", "reputation and trust management", "ranking" and "modeling of user experience" will be implemented.

Development

  • ScienceWISE will become standard part of a preprint submission procedure at Arxiv.org.
  • Dissemination activities like systematic efforts to inform scientific communities about the project, preparing and submitting papers and conference proceedings and advertising the system at scientific conferences are foreseen.
  • Extend the ScienceWISE system to other relevant scientific communities is an important step. For that purpose a number of technical problems should be solved (e.g. using formats other then TeX) and ontology be significantly expanded, using import of the existing field specific resources as a starting point.
  • To allow users effectively find "articles on the same topic", techniques developed in computer science and statistical physics will be used to realize automatic clustering of scientific articles and thus group similar articles. Based on this clustering follows a fine-tuning of the algorithms of determining the most relevant concepts for each paper.
  • The main approach today for finding articles relevant to a given article - manually searching databases for relevant titles, keywords, terms and references - does not work satisfactory and often crucial information does not reach the interested party. It is, therefore, important to develop methods of automatic context analysis for the calculation of mutual relevance of scientific articles.
  • Automated analysis of mathematical context is an attractive possibility since mathematical formulae are much more rigid syntactically and semantically than the sentences of a natural language. There also exists a set of formal transformation rules establishing equivalence of two equations having different syntax. Based on this development "Equation search" will be realized (finding papers that analyze or use the same mathematical equations, even though the underlying meaning of the variables in these equations can be different).
  • "Matching" of articles that use similar equations (or sets of equations) will be implemented and
  • techniques, existing in computer science to provide correct reputation and trust management of information and resources, available for the ScienceWISE users. Also ranking of concepts, resources and users, based on these techniques and on user's feedback will be developed.
  • "Probabilistic models of user behavior" will allow to optimize the presentation part of the User Interface, based on the models of user behavior.

Back