A Maximum Likelihood Approach to Continuous Speech Recognition

By L. Bahl, Frederick Jelinek, Robert L. Mercer. 1983

Subfields: Natural Language Processing, Speech & Audio

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Natural Language Processing

In Speech & Audio

Citations in the corpus

LLaMA: Open and Efficient Foundation Language Models (2023, methodology): “Traditionally, language models were based on n-gram count statistics (Bahl et al., 1983), and various smoothing techniques were proposed to improve the estimation of rare events (Katz, 1987; Kneser and Ney, 1995).”
Augmented Language Models: a Survey (2023, background): “Is a model trained to do intermediate reasoning steps or having access to the internet still purely performing language modeling? Indeed, in NLP, language modeling (Bahl et al., 1983) is generally defined as the task of predicting missing tokens given a context and is relied heavily on for pre-train”
Unbounded cache model for online language modeling with open vocabulary (2017, background): “It is often framed as learning the conditional probability of words, given their history [4].”
StarCoder: may the source be with you! (2023, methodology): “To evaluate the ability of the model to beneﬁt from this larger context, we compare its perplexity (Bahl et al., 1983) when using a full window size of 8K tokens versus a window size of 2K tokens (as used in many prior code models).”
Unsupervised Speech Recognition (2021, background): “The acoustic and the language model are often learned separately with a maximal likelihood objective (Bahl et al., 1983; Juang et al., 1986):”
Twin Networks: Matching the Future for Sequence Generation (2017, unknown): “, 2003; Mikolov, 2010) and speech recognition (Bahl et al., 1983).”
Dialog act modelling for conversational speech (1998, methodology): “The combination of likelihood and prior modeling, HMMs, and Viterbi decoding is fundamentally the same as the standard probabilistic approaches to speech recognition ( Bahl, Jelinek, & Mercer 1983 ) and tagging (Church 1988).”
Probabilistic models with unknown objects (2006, methodology): “8 It interpolates between unigram, bigram and trigram models using weights estimated on held-out data [Bahl et al., 1983].”
Learning Lexicons From Speech Using a Pronunciation Mixture Model (2013, background): “In this way, the lexicon fits nicely into the probabilistic framework other components of an automatic speech recognizer already enjoy [5], and its training is well-motivated through its incorporation into the fundamental equations of recognition described above.”