At today's Interspeech 2021 conference, AppTek’s speech scientists will be presenting papers on a variety of topics, starting with sampling-based criteria for language modeling with large vocabularies, a hot topic when it comes to modern word-based language models, which are becoming increasingly larger. Their research shows that all sampling methods perform equally well when model outputs are corrected for the intended class posterior probabilities, while delivering the expected speedups. Their claims are supported by experimental evidence in language modeling and ASR on LibriSpeech and SwitchBoard.
Y. Gao, D. Thulke, A. Gerstenberger, V. A. K. Tran, R. Schlüter, H. Ney:
"On Sampling-Based Training Criteria for Neural Language Modeling"
https://arxiv.org/abs/2104.10507
Novel end-to-end ASR architectures do not distinguish between acoustic and language models. While these novel architectures enable more efficient search, they can be challenging in case of domain mismatch. The next paper is concerned with methods for language model prior correction.
A. Zeyer, A. Merboldt, W. Michel, R. Schlüter, H. Ney
"Librispeech Transducer Model with Internal Language Model Prior Correction"
http://arxiv.org/abs/2104.03006
AppTek is a global leader in artificial intelligence (AI) and machine learning (ML) technologies for automatic speech recognition (ASR), neural machine translation (NMT), natural language processing/understanding (NLP/U) and text-to-speech (TTS) technologies. The AppTek platform delivers industry-leading solutions for organizations across a breadth of global markets such as media and entertainment, call centers, government, enterprise business, and more. Built by scientists and research engineers who are recognized among the best in the world, AppTek’s solutions cover a wide array of languages/ dialects, channels, domains and demographics.
