ALICE is a tool for estimating the number of adult-spoken linguistic units from child-centered audio recordings, as captured by microphones worn by children. It is meant as an open-source alternative for LENA adult word count (AWC) estimator [1].
ALICE uses SylNet [2] for feature extraction and voice type classifier [3] for broad-class speaker diarization. The used model for linguistic unit counts has been optimized across four languages: Argentinian Spanish, Tseltal, Yélî Dnye, and American and UK variants of English. SylNet uses a model that has been adapted for daylong child-centered audio, starting from the baseline model available in standard SylNet.
ALICE outputs an estimate for the number of phonemes, syllables, and words in the input. Only speech detected as spoken by adult male or female talkers is considered towards the counts.
Unit counts from ALICE are not (and are not meant to be) accurate at short time-scales, but optimized for counting across several minutes of audio. Also note that ALICE is NOT designed for "typical" high-quality audio recordings, and may not operate on such data properly.
- Installation
- Usage
- Demo & installation verification
- Common problems
- Patch notes
- License
- Related tools
If you use ALICE or its derivatives, please cite the following paper:
Räsänen, O., Seshadri, S., Lavechin, M., Cristia, A. & Casillas, M. (in press): ALICE: An open-source tool
for automatic linguistic unit count estimation from child-centered daylong recordings. Behavior Research Methods.
Online open acccess: https://link.springer.com/article/10.3758/s13428-020-01460-x.
If you use the speaker diarization output (e.g., to compute conversational turns), please cite the following paper:
Lavechin, M., Bousbib, R., Bredin, H., Dupoux, E., & Cristia, A. (2020).
An open-source voice type classifier for child-centered daylong recordings. Interspeech.
Online open access: https://www.isca-archive.org/interspeech_2020/lavechin20_interspeech.pdf
[1] Xu, D., Yapanel, U. Gray, S., Gilkerson, J., Richards, J. Hansen, J. (2008).
Signal processing for young child speech language development
Proceedings of the 1st Workshop on Child Computer and Interaction (WOCCI-2008), Chania, Crete, Greece.
(https://www.lena.org/)
[2] Seshadri S. & Räsänen O. (2019). SylNet: An Adaptable End-to-End Syllable Count Estimator for Speech.
IEEE Signal Processing Letters, vol 26, pp. 1359--1363 (https://github.com/shreyas253/SylNet)
[3] Lavechin, M., Bousbib, R., Bredin, H., Dupoux, E., & Cristia, A. (2020).
An open-source voice type classifier for child-centered daylong recordings. Interspeech.
(https://github.com/MarvinLvn/voice-type-classifier)