Queens College (CUNY)
Office: NSB A330
AuToBI is a tool for the automatic analysis of Standard American English prosody. It is based largely on work described my dissertation -- Automatic Detection and Classification of Prosodic Events.
This page will host code, packages and models as they become available.
AuToBI is free software, released under the GNU GPL v3. AuToBI is delivered as-is.
AuToBI is a java toolkit that hypothesizes pitch accents and phrase boundaries. The toolkit includes an acoustic feature extraction frontend, and a classification backend that is heavily supported by the weka machine learning toolkit and to a lesser degree LIBLINEAR.
The best way to get the most current version of AuToBI is through github, or cloning directly:
git clone firstname.lastname@example.org:AndrewRosenberg/AuToBI.git
git clone git://github.com/AndrewRosenberg/AuToBI.git
In addition to the github repository, milestone releases are available here.
AuToBI 1.4 compiled .jar file [updated 5/30/2014]
AuToBI 1.3 compiled .jar file
AuToBI 1.2 compiled .jar file
AuToBI 1.1 compiled .jar file
AuToBI 1.0 compiled .jar file
AuToBI Manual [Updated 10/2/2012/]
To cite AuToBI, please use:
Andrew Rosenberg. 2010. AuToBI - A tool for Automatic ToBI annotation. Interspeech '10
To post questions, feature requests and receive update notifications join the AuToBI Users Google Group.
Included Utilities -- There are three classes used within AuToBI which can be run independently from the prosodic annotation pipe.
Syllabifier -- Construct acoustic based psedosyllabification of a wave file.
PitchExtractor -- Extract a pitch track (Hz) from a wave file.
IntensityExtractor -- Extract an intensity track (dB) from a wave file.
Trained Models (version 1.4)
Models trained on pseudosyllable regions, i.e. without word boundaries
Models for languages other than English
Trained Models (version 1.3)
Trained Models (version 1.1 & 1.2)
Trained Models (version 1.0)
This work was partially supported by NSF IIS-HLC 0534568.