MoTT: A Speech Dataset for Modular Composition of Turn-Taking Conversations

MoTT This is MoTT (Modular Turn-Taking), an English speech dataset specifically designed to isolate key components of human conversation. The dataset includes audio files (Speech.zip) of questions and answers for 8 participants across 10 different topics (music preferences, cooking, hobbies, travel, job, preferred school subject, languages, entertainment, animals, superpowers).Participants were asked to record other speech elements typically occurring in a conversation, including reciprocal questions, i.e., short generic queries used to send back the question, and backchannel responses. The collected speeches are intended to be interchangeable items that can be arbitrarily assembled to simulate fictional conversations that closely mimic the dynamics of natural human interaction.The dataset is described in the following paper: Salada, G., Fantini, D., Avanzini, F., & Presti, G. (2025, September). MoTT: A speech dataset for modular composition of turn-taking conversations. In 2025 Immersive and 3D Audio: from Architecture to Automotive (I3DA) (pp. 1-8). IEEE. The MoTT dataset also includes room impulse response (RIR) measurements performed in the room where the speech dataset has been recorded (RIRs.zip). Further, a use case of the MoTT dataset for the modular composition of turn-taking conversations is provided (UseCase.zip). In particular, 9 different conversations are included, each encompassing from 2 to 4 speakers, which were generated by assembling the elements of the MoTT dataset. How to cite If you use the MoTT dataset, please cite the following paper: @inproceedings{salada2025mott, title = {{MoTT}: A Speech Dataset for Modular Composition of Turn-Taking Conversations}, author = {Salada, Giulio and Fantini, Davide and Avanzini, Federico and Presti, Giorgio}, year = {2025}, month = {September}, pages = {1-8}, booktitle = {2025 Immersive and 3D Audio: from Architecture to Automotive (I3DA)}, location = {Bologna, Italy}, publisher = {IEEE}, doi = {10.1109/I3DA65421.2025.11202114}} Acknowledgments This work is part of SONICOM, a project that has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 101017743.