Improving Statistical Parsers Using Cross-Corpus Data - 2002
Save Big on Coursera Plus. 7,000+ courses at $160 off. Limited Time Only! Grab it Explore techniques for improving statistical parsers using cross-corpus data in this 55-minute lecture by Xiaoqiang Luo from IBM. Learn how to leverage data annotated for other purposes to enhance parser performance without the need for expensive, labor-intensive labeling. Discover how the EM algorithm can be employed to infer missing information from partial constraints provided by label information from other domains or corpora. Examine results demonstrating improvements to a maximum entropy parser using cross-domain and cross-corpus data. Gain insights into statistical modeling for natural language processing, language modeling, speech recognition, and spoken dialog systems from Luo's experience developing semantic parsers and interpreters for IBM's DARPA Communicator project.