Audio-based Distributional Semantic Models (ADSM) toolkit. ADSM compute distributional word representations based on their acoustic properties. To achieve this, acoustic features are extracted from audio clips and clip-tag mappings are used to compute tag representations. For more info see:
Karamanolakis, Giannis, et al. "Audio-based Distributional Representations of Meaning Using a Fusion of Feature Encodings." Interspeech 2016 (2016): 3658-3662.