Improving Information Extraction by Modeling Errors in Speech Recognizer Output

12724

Knowledge Triples