Corpora: testing association strength between elements of trigrams

From: John Colby (colby@cse.ucsc.edu)
Date: Mon Jan 10 2000 - 05:19:45 MET

  • Next message: Priscilla Rasmussen: "Corpora: FOUR ANLP/NAACL-2000 Workshop Calls for Papers"

    I would like to use likelihood ratios, as has been done in Dunning[1993]
    for bigrams, to test the amount of association between the elements of
    trigrams. Dunning did this for a bigram AB by determining if the distribution
    of A given that B is present is the same as A given that B is not present.

    To do something similar for trigrams, is it sufficient to determine for
    a trigram ABC if the distribution of A given the presence of B and C is
    the same as the distribution of A given that both B and C are not present?

    This archive was generated by hypermail 2b29 : Wed Jan 12 2000 - 23:12:57 MET