De-identified dataset underlying the paper 'Remembering Past Emotions". The dataset consists of 17 participants discussing happy and sad memories and rated each memory on a past and present valence scale ranging from -1 (negative) to 1 (positive). Each memory was segmented based on meaningful topics. The lexical, acoustic and facial features were then extracted based on the segments transcripts and the video and audio recordings. Due to privacy reasons and consent of participants, only de-identified data is shared. The dataset does not contain transcripts, topic labelling, video and audio-recordings or any other personal identifiable data.

For more information on the dataset, data collection, extraction of the features, please refer to the paper.