In many studies of language production, the dependent variables are the latency with which speakers produce a spoken response to a stimulus or the temporal structure of a spoken utterance (specifically onset- and offset-times of words). Measuring these variables automatically often yields partially incorrect results. However, exact measurements through the visual inspection of the recordings are extremely time-consuming and resourceful. We present AlignTool, an open source alignment tool based on Praat and the automatic speech recognition system MAUS. AlignTool establishes preliminarily the onset and offset times of words and phonemes in spoken utterances by means of a voice-activitiy detection device implemented in Praat and a forced alignment of the spoken utterances and their orthographic transcriptions in MAUS. AlignTool creates a Praat text grid file for inspection and manual correction by the user, if necessary. AlignTool functions as a voice-key but also automatically aligns word onsets and offsets in longer utterances, for instance in semi-spontaneous dialogues. AlignTool can deal with files with a poor signal-to-noise ratio but may require more manual corrections with decreasing audio quality. However, AlignTool cannot deal with continuous background noise (e.g., from an MRT scanner) or dialogue sequences that include longer intervals of simultaneous speech by more than one speaker. We hope that by rendering the temporal analyses of complex utterances feasible, AlignTool will pave the way for new avenues in language production research.
We have evaluated AlignTool's accuracy extensively in a variety of recording contexts and experimental settings. AlignTool and the results of its evaluation are documented in a paper in Behavior Research Methods:
- Schillingmann, L., Ernst, J., Keite, V., Wrede, B., Meyer, A.S., & Belke, E. (2018). AlignTool: The automatic temporal alignment of spoken utterances in German, Dutch and British English for psycholinguistic purposes. Behavior Research Methods, 50, 466-489.
AlignTool was developped as part of a collaboration between the members of the Applied Informatics Group at the University of Bielefeld (Lars Schillingmann, Britta Wrede) and psycholinguists at Ruhr-University Bochum (Eva Belke, Jessica Ernst, Verena Keite) and the Max Planck Institute for Psycholinguistics (Antje S. Meyer). This work was funded by the German Research Foundation (DFG, BE3176/4-1). The source code for AlignTool is available here.
You find a detailed documentation of the installation and use of AlignTool here. It includes a documentation of how to align automatically data from the Map Task corpus. The TextGrid example files for this are available here.
Please cite AlignTool and its documentation as follows:
Schillingmann, L., Wrede, B., & Belke, E. (2017). AlignTool (Version 0c2855a) [Computer software]. Retreived from https://ruhr-uni-bochum.sciebo.de/s/zCsV7rrb4A8g3U0.
Belke, E., Keite, V., & Schillingmann, L. (2017). AlignTool Documentation (Version 0c2855a). Retrieved from https://www.linguistics.rub.de/~belke/aligntool.shtml.