A Comparison of Hybrid and End-to-End Models for Syllable Recognition

TL;DR

This paper compares hybrid and end-to-end models for German syllable recognition, demonstrating that hybrid models with strong language models outperform end-to-end approaches significantly.

Contribution

It provides a detailed comparison showing the continued importance of explicit prior knowledge modeling in speech recognition systems.

Findings

Hybrid models achieved 10.0% WER on syllable recognition.

End-to-end models achieved 27.53% WER.

Structured hybrid approach outperforms end-to-end in syllable recognition.

Abstract

This paper presents a comparison of a traditional hybrid speech recognition system (kaldi using WFST and TDNN with lattice-free MMI) and a lexicon-free end-to-end (TensorFlow implementation of multi-layer LSTM with CTC training) models for German syllable recognition on the Verbmobil corpus. The results show that explicitly modeling prior knowledge is still valuable in building recognition systems. With a strong language model (LM) based on syllables, the structured approach significantly outperforms the end-to-end model. The best word error rate (WER) regarding syllables was achieved using kaldi with a 4-gram LM, modeling all syllables observed in the training set. It achieved 10.0% WER w.r.t. the syllables, compared to the end-to-end approach where the best WER was 27.53%. The work presented here has implications for building future recognition systems that operate independent of a large vocabulary, as typically used in a tasks such as recognition of syllabic or agglutinative languages, out-of-vocabulary techniques, keyword search indexing and medical speech processing.