Home About Publications Dataset Platform



On this page you can find the several datasets that are based on the speech features.

Speech Features

Our speech feature extraction code is implemented in matlab. To run the speech feature extraction you need first to install the Bayes Net Toolbox from Kevin Murphy.

Reference Paper:

Negotiation DataSet

Because of privacy issue we could not put the original wave files online. Instead, you can access the extracted audio features for the 46 sessions (download [350MB])

For the social signaling measurement analysis we used first five minutes of each session. (data_five_min, Readme)

This scripts uses the audio features to generate the social signaling measurements and combines them in a table with the objective outcomes. (download)

Reference Paper:

Group Polarization DataSet

This dataset contains four group conversation with four people each containing twelve sessions. In each session the group had to discuss the standard questions for the risky shift as used by Stoner (Wikipedia). Each person expressed his opinion before and after the group conversation which was approximately 3 minutes.

All extracted audio features for each session are combined in an zip file (download [290MB])

The audio features are provided here (download) together with the matlab script that created them (download)

The original file with the subjects opinion are stored in an excel file. This file contains the raw data of personality questionnaire that each participant had to fill out several days after the session (download)

The same data can be downloaded in a processed matlab format (download). The script that generated this file is available as well (download)

Speed Dating DataSet

This dataset consists of extracted speech features from 52 5-minute conversations.

Conversational Interest DataSet