Corpus Synthesis for Zero-shot ASR domain Adaptation using Large Language Models.
Hsuan SuTing-Yao HuHema Swetha KoppulaRaviteja VemulapalliJen-Hao Rick ChangKarren YangGautam Varma MantenaOncel TuzelPublished in: CoRR (2023)
Keyphrases
- language model
- domain adaptation
- speech recognition
- word error rate
- automatic speech recognition
- document level
- language modeling
- cross domain
- probabilistic model
- n gram
- multiple sources
- document retrieval
- labeled data
- semi supervised
- information retrieval
- test collection
- sentiment classification
- query expansion
- transfer learning
- retrieval model
- speech signal
- target domain
- semi supervised learning
- sentence level
- relevance model
- test data
- pseudo relevance feedback
- co training
- document classification
- test set
- unlabeled data
- text classification
- decision trees
- k nearest neighbor
- information retrieval systems
- knn