[Corpora-List] Corpora annotated with coreference relations and word senses?

From: Josef Meyer (jmeyer@ics.mq.edu.au)
Date: Sat Dec 14 2002 - 05:42:40 MET

  • Next message: Tony Berber Sardinha: "[Corpora-List] sentence aligner script"

    Dear List,

    I am a PhD student currently doing research into resolving associative
    anaphors. My approach to filtering antecedents involves learning
    semantic relationships between concepts from a (automatically) parsed
    corpus. Until now, my evaluations have made use of about 1000
    instances of associative anaphors drawn from a set of encyclopedia
    entries; however, this is clearly inadequate for the sort of
    evaluation that I would like to include in my thesis.

    I was wondering about the availability of corpora of written
    English annotated with the following types of information:

    (1) Coreference relationships between NPs

    (2) Bridging relationships between NPs, and

    (3) Word sense information (WordNet synsets) for NPs

    Ideally I would like to find a number of domain-restricted corpora
    annotated with all three types of information; however, I am well
    aware that it is unlikely that any such corpus currently exists.
    What I am hoping for is a pointer to a corpus annotated with (1) and
    (3).

    Currently I am aware of the following publically-available corpora
    that are annotated with WordNet senses:

    - SemCor, which contains a subset of the brown corpus annotated with
      WordNet 1.6 senses (http://www.cogsci.princeton.edu/~wn/wn1.6.shtml)

    - The Senseval-1 and Senseval-2 corpora, the latter of which contains
      many instances of a limited set of words annotated with WordNet 1.7
      senses, and is mainly drawn from the WSJ entries in the Penn
      Treebank (http://www.senseval.org/)

    I have also come across the following corpora that are annotated with
    coreference information:

    - A corpus of 7 texts (mainly instructional) from the University of
      Wolverhampton (http://clg.wlv.ac.uk/resources/corpus.html)

    Some time ago I thought that I had seen a set of WSJ entries annotated
    with coreference information, but I wasn't able to locate this in the
    quick search that I performed yesterday.

    Regards,

    - jo

    -- 
    +-------------------------------------------------------------------+
    | Josef Meyer                    | http://www.mri.mq.edu.au/~jmeyer |
    | Information and Communication  | jmeyer@ics.mq.edu.au             |
    | Sciences, Macquarie University | Phone: +61 2 9850 9571           |
    | NSW,  Australia  2109          | Fax:   +61 2 9850 9542           |
    +-------------------------------------------------------------------+
    



    This archive was generated by hypermail 2b29 : Sat Dec 14 2002 - 05:45:07 MET