1 / 11

Structure Learning for NLP Named-entity recognition using generative models

Master in Artificial Intelligence (MIA). Structure Learning for NLP Named-entity recognition using generative models. An Algorithm that Learns What’s in a Name (IdentiFinder TM ) Dan M. Bikel, Richard Schwartz, Ralph Weischedel Machine Learning Special Issue on Natural Language Learning

blake
Download Presentation

Structure Learning for NLP Named-entity recognition using generative models

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Master in Artificial Intelligence (MIA) • Structure Learning for NLP • Named-entity recognition using generative models

  2. An Algorithm that Learns What’s in a Name (IdentiFinderTM) Dan M. Bikel, Richard Schwartz, Ralph Weischedel Machine Learning Special Issue on Natural Language Learning C. Cardie and R. Mooney eds. 1999

  3. Generalities • HMM-based model for NERC • Evolution from the Nymble system (1997) • Applied in MUC conferences and other corpora with very good results • Competitive (and better in some cases) to the best hand-developed rule-based NERC systems • State-of-the-art for the task until CoNLL conferences (ML)

  4. HMMs

  5. HMMs Decoding: The argmax function can be computed in linear time O(n) using dynamic programming (Viterbi algorithm)

  6. IdentiFinder HMM: Graphical Representation

  7. IdentiFinder HMM • Generative model • Probability distributions needed: • P(NC|NC-1,w-1) • P(<w,f>first |NC,NC-1) • P(<w,f> |<w,f>-1,NC-1)

  8. IdentiFinder HMM • Example: Mr. [John]E-PERSON eats Probability of the previous annotated sequence:

  9. IdentiFinder HMM • Parameter Estimation • Maximum likelihood estimates • Incorporates smoothing and back-off to face sparseness • Decoding • argmaxNC (NC1…NCn|w1…wn) • Viterbi algorithm (dynamic programming)

  10. IdentiFinder: Results

  11. IdentiFinder: Results

More Related