Improving Language Models by Learning from Speech Recognition Errors in a Reading Tutor that Listens Export

@inproceedings{citeulike:309376, abstract = {Lowering the perplexity of a language model does not always translate into higher speech recognition accuracy. Our goal is to improve language models by learning from speech recognition errors. In this paper we present an algorithm that first learns to predict which n–grams are likely to increase recognition errors, and then uses that prediction to improve language models so that the errors are reduced. We show that our algorithm reduces a measure of tracking error by more than 24\% on unseen test data from a Reading Tutor that listens to children read aloud.}, author = {Banerjee, Satanjeev and Mostow, Jack and Beck, Joseph and Tam, Wilson}, citeulike-article-id = {309376}, comment = {uses not only bigram features, but also bigram context and language learner features (e.g. reading level) to generalize bigrams.}, keywords = {asr, call, cmu, language\_model, listen, machine\_learning, pedagogy}, organization = {ICAAI}, posted-at = {2005-08-31 19:22:59}, priority = {0}, title = {Improving Language Models by Learning from Speech Recognition Errors in a Reading Tutor that Listens}, year = {2003} }

TY - CONF ID - citeulike:309376 L3 - citeulike-article-id:309376 N2 - Lowering the perplexity of a language model does not always translate into higher speech recognition accuracy. Our goal is to improve language models by learning from speech recognition errors. In this paper we present an algorithm that first learns to predict which n–grams are likely to increase recognition errors, and then uses that prediction to improve language models so that the errors are reduced. We show that our algorithm reduces a measure of tracking error by more than 24% on unseen test data from a Reading Tutor that listens to children read aloud. OR - ICAAI TI - Improving Language Models by Learning from Speech Recognition Errors in a Reading Tutor that Listens KW - asr KW - call KW - cmu KW - language_model KW - listen KW - machine_learning KW - pedagogy N1 - uses not only bigram features, but also bigram context and language learner features (e.g. reading level) to generalize bigrams. AU - Banerjee, Satanjeev AU - Mostow, Jack AU - Beck, Joseph AU - Tam, Wilson PY - 2003/// ER -