WERd: Using Social Text Spelling Variants for Evaluating Dialectal Speech Recognition

TL;DR

This paper introduces WERd, a new evaluation metric for dialectal speech recognition that accounts for non-standard orthography by using social media spelling variants, improving upon traditional WER.

Contribution

The paper proposes WERd, a dialect-specific evaluation metric based on TERp, leveraging social media spelling variants to better assess dialectal ASR systems.

Findings

WERd outperforms traditional WER in dialectal ASR evaluation

Social media spelling variants improve evaluation accuracy

Manual analysis confirms WERd's effectiveness

Abstract

We study the problem of evaluating automatic speech recognition (ASR) systems that target dialectal speech input. A major challenge in this case is that the orthography of dialects is typically not standardized. From an ASR evaluation perspective, this means that there is no clear gold standard for the expected output, and several possible outputs could be considered correct according to different human annotators, which makes standard word error rate (WER) inadequate as an evaluation metric. Such a situation is typical for machine translation (MT), and thus we borrow ideas from an MT evaluation metric, namely TERp, an extension of translation error rate which is closely-related to WER. In particular, in the process of comparing a hypothesis to a reference, we make use of spelling variants for words and phrases, which we mine from Twitter in an unsupervised fashion. Our experiments with evaluating ASR output for Egyptian Arabic, and further manual analysis, show that the resulting WERd (i.e., WER for dialects) metric, a variant of TERp, is more adequate than WER for evaluating dialectal ASR.