Richard Beaufort

Train&align: A new online tool for automatic phonetic alignment

Several automatic phonetic alignment tools have been proposed in the literature. They usually rely on pre-trained speaker-independent models to align new corpora. Their drawback is that they cover a very limited number of languages and might not perform properly for different speaking styles. This paper presents a new tool for automatic phonetic alignment available online. Its specificity is that it trains the model directly on the corpus to align, which makes it applicable to any language and speaking style. Experiments on three corpora show that it provides results comparable to other existing tools. It also allows the tuning of some training parameters. The use of tied-state triphones, for example, shows further improvement of about 1.5% for a 20 ms threshold. A manually-aligned part of the corpus can also be used as bootstrap to improve the model quality. Alignment rates were found to significantly increase, up to 20%, using only 30 seconds of bootstrapping data.

Automatic Phone Alignment

Several automatic phonetic alignment tools have been proposed in the literature. They generally use speaker-independent acoustic models of the language to align new corpora. The problem is that the range of provided models is limited. It does not cover all languages and speaking styles (spontaneous, expressive, etc.). This study investigates the possibility of directly training the statistical model on the corpus to align. The main advantage is that it is applicable to any language and speaking style. Moreover, comparisons indicate that it provides as good or better results than using speaker-independent models of the language. It shows that about 2% are gained, with a 20 ms threshold, by using our method. Experiments were carried out on neutral and expressive corpora in French and English. The study also points out that even a small neutral corpus of a few minutes can be exploited to train a model that will provide high-quality alignment.

Automatic detection and correction of syntax-based prosody annotation errors

Both unit-selection and HMM-based speech synthesis require large annotated speech corpora. To generate more natural speech, considering the prosodic nature of each phoneme of the corpus is crucial. Generally, phonemes are assigned labels which should reflect their suprasegmental characteristics. Labels often result from an automatic syntactic analysis, without checking the acoustic realization of the phoneme in the corpus. This leads to numerous errors because syntax and prosody do not always coincide. This paper proposes a method to reduce the amount of labeling errors, using acoustic information. It is applicable as a post-process to any syntax-driven prosody labeling. Acoustic features are considered, to check the syntax-based labels and suggest potential modifications. The proposed technique has the advantage of not requiring a manually prosody-labelled corpus. The evaluation on a corpus in French shows that more than 75% of the errors detected by the method are effective errors which must be corrected.