Spoken Language Processing

Research tools and resources for Spoken Language Processing

Spoken Language Processing (SLP) is a portal designed for language technologies that have been developed by the Center for Artificial Intelligence of the Romanian Academy (RACAI) but also to enable collaboration between research institutes across the field. We provide a demo for NLP Processing and a demo for TTS.

Currently the main frameworks developed here are Modular Language Processing for Lightweight Applications (MLPLA) and Speech Synthesis for Lightweight Applications (SSLA)

MLPLA has the following features:

  • Modular and scalable architecture
  • Out-of-the-box standard processors designed for basic NLP and TTS synthesis
  • Plug-and-play processors
  • Configurable processing pipeline
  • Part-of-speech tagging, lemmatization, chunking, syllabification and letter-to-sound conversion for English and Romanian
  • Support for ID3 decision trees and DNN classifiers
  • Configurable output formats: out-of-the-box support for tab separated values and HTS feature output

SSLA has the following features:

  • Zero-configuration feature templates (the features provided as it’s input don’t require any description and no configuration is necessary for the system to use them)
  • Support for Statistical Parametric Speech Synthesis using the MLSA and STRAIGHT filters
  • Hybrid speech synthesis support is currently under development
  • Compatible with HTS models