Clione

P2PSTORY is a dataset of young children (5-6 years old) engaging in natural peer-to-peer storytelling interactions with fellow classmates. The dataset consists of rich social behaviors of children without adult supervision, with each participant demonstrating being a storyteller and a listener. The dataset contains 58 recorded storytelling sessions along with a diverse set of behavioral annotations, as well as developmental and demographic profiles of each child participant.

The data in P2PSTORY includes: 1. Audio and video: Recordings were collected for each session from three cameras (capturing the frontal-view of the storyteller's face, the frontal-view of the listener's face, and a bird's-eye view of both participants) and a high quality microphone (the MXL AC404 USB conference microphone), all time-synchronized. The video recordings have a resolution of 720x1280 at 30 Hz and are encoded as JPEG files and the audio recordings are 16bit at 44.1 kHz. 2. Behavioral features: Video recordings were coded for a wide range of behaviors, selected as they were either found in prior works or commonly observed in the storytelling interaction: gaze, posture, nods, smiles & frowns, eyebrow movement, and backchannel utterances. Interaction-level features were also annotated, including listener’s attention and whether the pair of participants were on or off task. 3. Prosodic features: The storyteller’s use of prosodic cues (speech modulations) including pitch, energy, pauses, filled pauses, and long utterances were also annotated. 4. Personal features: Demographic and socio-emotional development profiles were collected for each participant, including each participant's gender, age, birth date, household income, ethnicity, mother’s highest education, siblings’ ages, and both raw questionnaire answers and cumulative scores from the Ages and Stages Questionnaire (ASQ). ASQ is a standardized measure of a child’s social-emotional development (lower scores indicate that a child’s development is as expected while higher values are indicative of potential need for assessment by a professional). 5. Child perceptions: To better understand how children perceived the effectiveness of their interaction partner, participants were asked to rate their partner on measures relating to attention and understanding.

Over a span of five weeks, each participant completed at least three rounds of storytelling with different partners and text-less storybooks. The storybooks were a series of colored pictures (story scenes) with illustrated characters and scenes that children could use to craft their own narratives.

Each round varied in the level of instructions (heavy, light, or none), partner type (friend, ability lo/hi, mixed-gender, or misc), and number of story scenes (1, 2, or 3).

In a given round, the pair of students took turns narrating a story to their partner with each turn generating a storytelling episode. The dataset contains a total of 58 storytelling episodes, where the average length of a child’s story was 1 minute and 17 seconds, totaling to about 75 minutes of content.

The core value of the dataset is to inform the design of child-computer and child-robot interfaces.

Katharine Brush, Raina Hall, Emily Holding, Yiyu Li, and Matthew Sears annotated videos and helped conduct user studies.