Morpho-syntactic post-processing of N-best lists for improved French automatic speech recognition

Abstract

International audienceMany automatic speech recognition (ASR) systems rely on the sole pronunciation dictionaries and language models to take into account information about language. Implicitly, morphology and syntax are to a certain extent embedded in the language models but the richness of such linguistic knowledge is not exploited. This paper studies the use of morpho-syntactic (MS) information in a post-processing stage of an ASR system, by reordering N-best lists. Each sentence hypothesis is first part-of-speech tagged. A morpho-syntactic score is computed over the tag sequence with a long-span language model and combined to the acoustic and word-level language model scores.This new sentence-level score is finally used to rescore N-best lists by reranking or consensus. Experiments on a French broadcast news task show that morpho-syntactic knowledge improves the word error rate and confidence measures. In particular, it was observed that the errors corrected are not only agreement errors and errors on short grammatical words but also other errors on lexical words where the hypothesized lemma was modified

Statistical Pronunciation Adaptation for Spontaneous Speech Synthesis

“…, p m ) of m phonemes p i is assigned a score s(h) mixing the CRF and phonological model (PM) probabilities. This mixture is computed by a log-linear interpolation-which has been successfully used for N -best list reranking in various domains [18,19]-, and is formulated as follows:…”

Manual sorting of numerals in an inflective language for language modelling

“…Proper names are also a topic of interest in pronunciation (Reveil et al 2012;Schlippe et al 2014), which is also important for speech recognition. Huet et al (2010) proposed a post-processing method of speech recognizer hypotheses, which includes morphosyntactic description tagging. In the post processing step consecutive numerals and consecutive proper names are grouped into single cardinal and proper names tags.…”

Enhancing lexical cohesion measure with confidence measures, semantic relations and language model interpolation for multimedia spoken content topic segmentation

“…In the final steps, a 4-gram language model over a vocabulary of 65,000 words is used with context-dependent phone models to generate a list of 1,000 sentence transcription hypotheses. Morphosyntactic tagging, using a tagger specifically designed for ASR transcripts, is used in a post-processing stage to generate a final transcription by consensus from a confusion network, combining the acoustic, language model and morphosyntactic scores [25]. Confusion network posterior probabilities are used directly as confidence measures.…”