This website uses cookies and similar technologies to understand visitors' experiences. By continuing to use this website, you accept our use of cookies and similar technologies,Terms of Use, and Privacy Policy.

Oct 05 2009 - 02:33 PM
Extracting Information from Conversation using NLP techniques
NLP Stands for Natural Language Processing and is concerned with the interactions between computers and human (natural) languages. TALK on Usage of interviews to establish social networks Social Networks: nodes are people and links are relations Relational Extraction: Features - Words, entity type, etc. all words in between - distance one metnin in between two words aapart in same NP - dependency per<-of<-org 32 52000 - parse tree person-np-pp-organization person-np-pp:of-organization amount of annotation is limited for machine learning all the mentions. malach corpus (schnidlers list) ------------------- videotaped, digitized oral interviews with holocaust survivors. in languages survivors, liberators, rescuers and witnesses collected by usc shoah used many research activities multilingual access to large spoken archives(malach) e.g. (gustman et al., 2002) coreference resolution critical lot more pronouns. words error rate are around 35%. issues typical ie systems: lacking featurs related conversational speech. poor pronoun resolution. improving>
  • Standard Approach - Train classiier to predict whether two mentions are co-referent or not - clustering algorithm to partition mentions into clusters, based on the pair-wise probabilities
  • Improvements - Features focused on conversational speech - Improved clustering End to end system - Leverage ACE System(FLorian et. al 2004) Evaluation of Social Network Extraction - Precision, recall, F-measure of match of nodes and ties - Nodes match if they have the same canonical match use of ACE system is an important network
  • |By: Pranav Garg|17693 Reads