Morphosyntactic Parser for Brazilian Portuguese: Methodology for Development and Assessment

In text-to-speech (TTS) systems, an effective morphosyntactic classification is important to improve the prosody of synthesized speech as well as the pronunciation of words subject to vocalic alternation. This research work presents a methodology used for developing and assessing an ad hoc morphosyntactic parser to a TTS system for Brazilian Portuguese. The developed parser is composed of a dictionary and a set of rules structured in four levels. The methodology used for development consisted firstly in the creation of a large annotated dataset and an incremental development of rules for morphosyntactic classification. By using this approach, the achieved accuracy rate of the classification process is of 98.59% for words and 80.66% for sentences in a specific dataset.