Minimally Supervised Written-to-Spoken Text Normalization
This work addresses the challenge of reducing manual effort in developing text normalization engines for speech applications like TTS and ASR, but it is incremental as it builds on existing methods with variations in data and grammar usage.
The study tackled the problem of text normalization for speech applications by examining trade-offs between using extensive language-specific domain knowledge versus more automated approaches, finding that a language-universal covering grammar with hallucinated written data achieved competitive accuracy, with results showing 95.2% accuracy in English and 93.7% in Russian.
In speech-applications such as text-to-speech (TTS) or automatic speech recognition (ASR), \emph{text normalization} refers to the task of converting from a \emph{written} representation into a representation of how the text is to be \emph{spoken}. In all real-world speech applications, the text normalization engine is developed---in large part---by hand. For example, a hand-built grammar may be used to enumerate the possible ways of saying a given token in a given language, and a statistical model used to select the most appropriate pronunciation in context. In this study we examine the tradeoffs associated with using more or less language-specific domain knowledge in a text normalization engine. In the most data-rich scenario, we have access to a carefully constructed hand-built normalization grammar that for any given token will produce a set of all possible verbalizations for that token. We also assume a corpus of aligned written-spoken utterances, from which we can train a ranking model that selects the appropriate verbalization for the given context. As a substitute for the carefully constructed grammar, we also consider a scenario with a language-universal normalization \emph{covering grammar}, where the developer merely needs to provide a set of lexical items particular to the language. As a substitute for the aligned corpus, we also consider a scenario where one only has the spoken side, and the corresponding written side is "hallucinated" by composing the spoken side with the inverted normalization grammar. We investigate the accuracy of a text normalization engine under each of these scenarios. We report the results of experiments on English and Russian.