participants

• Amit
• Oleg 
• Artem
• Aliona

applications

• More meaningful clustering of concepts, papers, across SW and within personal collections

• Map of EPFL

  • Create a representation of EPFL laboratories on a map (using e.g. a “paperscape”) clustering them by “similar concepts” (ideally, one should be able to see

    1. list of topics on which EPFL labs are working

    2. topics where EPFL is well represented/poorly represented

    3. different laboratories working on the same subject

G​iven, for example, physics labs, mark laboratories according to the “influential papers” their members have written (in a given period of time). This allows to see “trend setters”​

All these maps should have “logical zoom” feature enabled in them. This means, that more specific details appear on the map as one zooms in (otherwise showing only the large scale information)

Experiments done

• clustering of papers (represented with found ontological concepts)
  • LDA - giving it no. clusters, cut-off
  • Newman algorithm - probabilistic, similar in spirit to LDA
  • modularity - yielded much better precision

• InfoMap alignment to the others

• initial experiments (LDA, Newman, modularity) - have data
• stability of the methods - did not do experiments, to be explored