Language Model Score Regularization for Speech Recognition

• Published in: Chinese Journal of Electronics Vol. 28; no. 3; pp. 604 - 609
• Main Authors: Zhang, Yike, Zhang, Pengyuan, Yan, Yonghong
• Format: Journal Article
• Language: English
• Published: Published by the IET on behalf of the CIE 01.05.2019
• Subjects: Bidirectional neural network; bidirectional neural networks; composite model; consistent word error rate reduction; count‐based LMs; different order models; fault‐tolerance; feedforward neural nets; Feedforward neural network LMs; fixed order Markov assumption; interpolated smoothing algorithms; Interpolation; Language model score regularization; Markov processes; neural nets; N‐best lists re‐scoring show; probability; recognition errors; recurrent neural nets; Recurrent neural network LMs; regularization algorithm; sentence‐level Language model; skip‐gram features; Speech recognition; statistical language modeling; recognition errors; fault-tolerance; sentence-level Language model; count-based LMs; skip-gram features; Bidirectional neural network; Recurrent neural network LMs; Language model score regularization; speech recognition; consistent word error rate reduction; fixed order Markov assumption; N-best lists re-scoring show; different order models; probability; recurrent neural nets; regularization algorithm; Feedforward neural network LMs; Interpolation; interpolated smoothing algorithms; Markov processes; bidirectional neural networks; composite model; neural nets; statistical language modeling; feedforward neural nets
• ISSN: 1022-4653; 2075-5597
• DOI: 10.1049/cje.2019.03.015

Inspired by the fact that back-off and interpolated smoothing algorithms have significant effect on statistical language modeling, this paper proposes a sentence-level Language model (LM) score regularization algorithm to improve the fault-tolerance of LMs for recognition errors. The proposed algorithm is applicable to both count-based LMs and neural network LMs. Instead of predicting the occurrence of a sequence of words under a fixed order Markov assumption, we use a composite model consisting of different order models with either n-gram or skip-gram features to estimate the probability of the sequence of words. In order to simplify implementations, we derive a connection between bidirectional neural networks and the proposed algorithm. Experiments were carried out on the Switchboard corpus. Results on N-best lists re-scoring show that the proposed algorithm achieves consistent word error rate reduction when it is applied to count-based LMs, Feedforward neural network (FNN) LMs, and Recurrent neural network (RNN) LMs.