Grounded Language Modeling for Automatic Speech Recognition of Sports Video

Grounded language models represent the relationship between words and the non-linguistic context in which they are said. This paper describes how they are learned from large corpora of unlabeled video, and are applied to the task of automatic speech recognition of sports video. Results show that grounded language models improve perplexity and word error rate over text based language models, and further, support video information retrieval better than human generated speech transcriptions.

• Type: Conference
• Year: 2008
• Where: ACL
• Authors: Michael Fleischman, Deb Roy
• Keywords: ACL 2008, Automatic Speech Recognition, Computational Linguistics, Language Models, Models Improve Perplexity