WordNet-based Semantic Relatedness Measures in Automatic Speech Recognition for Meetings

This paper presents the application of WordNet-based semantic relatedness measures to Automatic Speech Recognition (ASR) in multi-party meetings. Different word-utterance context relatedness measures and utterance-coherence measures are defined and applied to the rescoring of Nbest lists. No significant improvements in terms of Word-Error-Rate (WER) are achieved compared to a large word-based ngram baseline model. We discuss our results and the relation to other work that achieved an improvement with such models for simpler tasks.

• Type: Conference
• Year: 2007
• Where: ACL
• Authors: Michael Pucher
• Keywords: ACL 2007, Computational Linguistics, Context Relatedness Measures, Relatedness Measures, Semantic Relatedness Measures