1 / 3

Two-Stage P-R Maximization - Error Correction and MinErr Classifier

Explore the concept of Two-Stage P-R Maximization and Error Correction in the context of MinErr Classifier for identifying new words or compounds. Understand the decision rules and advantages of this approach in achieving precision and recall targets.

karendean
Download Presentation

Two-Stage P-R Maximization - Error Correction and MinErr Classifier

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. Two-Stage P-R Maximization(Errata- Lexicon+MaxPR -10 @2017/04/26) • Why two stage? • No simple analytical decision rules that are capable of achieving any user-specified criterion function of precision and recall • … • Error: • If mimimum error => • Correction: • If minimum error => • or Jing-Shin Chang, EE, National Tsing-Hua University

  2. MinErr Classifier: Two-Class Classifier forIdentifying New Words or Compound Words • Input: n-grams (n-word compounds, n-character words) in the text corpus • Output: assign a class label ("word" or "non-word") to each n-gram • Classifier: a log-likelihood ratio (LLR) tester (minimum error classifier) • Decision Rules: • Advantage: ensure minimum classification error (with 0 =0) if the distributions are known. Jing-Shin Chang, EE, National Tsing-Hua University

  3. Filter: Two-Class Classifier (Log-Likelihood Ratio Ranking Module) • Input: n-grams in the unsegmented text corpus • Output: assign a class label ("word" or "non-word") to each n-gram • Classifier: a log-likelihood ratio (LLR) tester (minimum error classifier) • Decision Rules: • Advantage: ensure minimum classification error (with 0 =0) if the distributions are known. Jing-Shin Chang, EE, National Tsing-Hua University

More Related