Bypass Temporal Classification: Weakly Supervised Automatic Speech Recognition with Imperfect Transcripts

This paper presents a novel algorithm for building an automatic speech recognition (ASR) model with imperfect training data. Imperfectly transcribed speech is a prevalent issue in humanannotated (a) Deletion (partial transcript) (b) Substitution speech corpora, which degrades the performance of ASR models. To address this problem, we propose Bypass Temporal Classification (BTC) as an expansion of the Connectionist Temporal Classification (CTC) criterion. BTC explicitly encodes the uncertainties associated with transcripts during (c) Insertion (d) Substitution and insertion training. This is accomplished by enhancing the flexibility of the training graph, which is implemented as a weighted finitestate Figure 1: Examples of error in the transcript. The grey box is transducer (WFST) composition. The proposed algorithm the exact text and the red box is the imperfect text. Inaccurate improves the robustness and accuracy of ASR systems, particularly words are marked in bold.