Conditional Structure versus Conditional Estimation in NLP Models

Abstract This paper separates conditional parameter estima- tion, which consistently raises test set accuracy on statistical NLP tasks, from conditional model struc- tures, such as the conditional Markov model used for maximum-entropy tagging, which tend to lower accuracy. Error analysis on part-of-speech tagging shows that the actual tagging errors made by the conditionally structured model derive not only from label bias, but also from other ways in which the independence assumptions of the conditional model structure are unsuited to linguistic sequences. The paper presents new word-sense disambiguation and POS tagging experiments, and integrates apparently conflicting reports from other recent work.
Keywords No keywords specified (fix it)
Categories No categories specified (fix it)
Options
 Save to my reading list
Follow the author(s)
My bibliography
Export citation
Find it on Scholar
Edit this record
Mark as duplicate
Revision history Request removal from index
 
Download options
PhilPapers Archive


Upload a copy of this paper     Check publisher's policy on self-archival     Papers currently archived: 5,672
External links
  •   Try with proxy.
  • Through your library Only published papers are available at libraries

    Similar books and articles
    Richard Bradley (1999). Conditional Desirability. Theory and Decision 47 (1):23-55.

    Analytics

    Monthly downloads

    Added to index

    2010-12-22

    Total downloads

    2 ( #232,382 of 549,069 )

    Recent downloads (6 months)

    0

    How can I increase my downloads?


    My notes
    Sign in to use this feature


    Discussion
    Start a new thread
    Order:
    There  are no threads in this forum
    Nothing in this forum yet.

    Other forums