1 / 7

ElkfEd/IDC alias BART

ElkfEd/IDC alias BART. Yannick Versley Simone Ponzetto Jason Smith Vladimir Eidelman Alan Jern Massimo Poesio Xiaofeng Yang. Introduction. Modularized and revamped version of EML-R system Full coreference including mention detection Will be publically available as Open Source

Download Presentation

ElkfEd/IDC alias BART

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. ElkfEd/IDC alias BART Yannick Versley Simone Ponzetto Jason Smith Vladimir Eidelman Alan Jern Massimo Poesio Xiaofeng Yang

  2. Introduction • Modularized and revamped version of EML-R system • Full coreference including mention detection • Will be publically available as Open Source • Useful Java-only subset (e.g. for teaching)‏

  3. Components • Preprocessingaggregate data in MMAX2 annotation layers • Mention extractioncreate markables from chunks/NEs • Extract information about mentionsmention type, semantic class ... • Encode coref into classifier decisions and extract features

  4. Machine Learning Infrastructure • WEKA Machine Learning Toolkit • C4.5, RIPPER, other learning modules • SVMlight-TK • SVMs, different kernels (linear, polynomial, etc.)‏ • Tree kernels • Custom kernels • MaxEnt-based ranking • find best among a set of candidates

  5. Flexibility in Preprocessing • Chunker (YamCha) vs. Parser (Charniak)‏ • Stanford NER vs. Carafe • Mentions from Chunks/NEs vs.Mentions from Mention Tagger (Carafe)‏

  6. Resolution Algorithms • Closest-first decoding (Soon et al., 2001)‏ • Separate classifiers for pronouns/non-pronouns • Ranking-based resolution • Stacking ranking+classifier • Global models

  7. Quantitative / Qualitative Evaluation • MUC scoring (per document, in total)‏ • Link-based scores, by type of anaphora(pronouns, appositions / copula, names, nominals)‏ • Qualitative evaluationinspect results in MMAX2

More Related