JUSThink Dialogue and Actions Corpus

1. Description

The information contained in JUSThink Dialogue and Actions Corpus dataset includes dialogue transcripts, event logs, and test responses of children aged 9 through 12, as they participate in a robot-mediated human-human collaborative learning activity named JUSThink [1, 2], where children in teams of two solve a problem on graphs together.

As a use case, it is processed to analyse how the children align their use of task-specific referents in their dialogue and actions (JUSThink Alignment Analysis) that is available from the Zenodo Repository, DOI: 10.5281/zenodo.4675070, and to produce the results and figures in [3].

2. Publications

If you use this work in an academic context, please cite the following publications:

• Norman*, U., Dinkar*, T., Bruno, B., & Clavel, C. (2022). Studying Alignment in a Collaborative Learning Activity via Automatic Methods: The Link Between What We Say and Do. Dialogue & Discourse, 13(2), 1–48. *Contributed equally to this work. https://doi.org/10.5210/dad.2022.201

• Nasir*, J., Norman*, U., Bruno, B., & Dillenbourg, P. (2020). When Positive Perception of the Robot Has No Effect on Learning. 2020 29th IEEE International Conference on Robot and Human Interactive Communication (RO-MAN), 313–320. *Contributed equally to this work. https://doi.org/10.1109/RO-MAN47096.2020.9223343

• Norman, Utku, Dinkar, Tanvi, Nasir, Jauwairia, Bruno, Barbara, Clavel, Chloé, & Dillenbourg, Pierre. (2021). JUSThink Dialogue and Actions Corpus [Data set]. In Dialogue & Discourse (v1.0.0, Vol. 13, Number 2, pp. 1–48). Zenodo. https://doi.org/10.5281/zenodo.4627104.

3. Content

The dataset is consisted of three parts:

1. transcripts: anonymised dialogue transcripts for 10 teams of two children
2. logs: anonymised event logs for 39 teams of two children
3. test responses: pre-test and post-test responses for 39 teams, and the key i.e. the correct response

In addition, there is metadata that contains information on the network that the children have worked on: 
It is a JSON file in a node-link format, providing the node labels (e.g. "Mount Luzern"), node ids, x, y position of a node, edges between the nodes, and edge costs (metadata/network.json).  It can be read into a NetworkX graph.

2.1. Transcripts

This part of the dataset contains the anonymised dialogue transcripts for 10 teams (out of 39 teams).
It consists of 10 files, with one tab-separated text file per team (transcripts/justhink19_transcript_<team_no>.csv).
In particular, the columns are:

• team_no: The number of the team that the dialogue belongs to
• utterance_no: The number of the utterance, starting from 0
• start: The start timestamp of the utterance (in seconds), from the beginning of the activity
• end: The end timestamp of the utterance (in seconds)
• interlocutor: The person (or the robot) that is speaking (A, B: the participants; R: the robot; I: an experimenter)
• utterance: The content of the utterance

Note that the start and end times are synchronised with the log times.

3.2. Logs

This part of the dataset contains anonymised event log data for 39 teams.
It consists of 39 files, with one tab-separated text file per team (logs/justhink19_log_<team_no>.csv).
In particular, the columns are:

• team_no: The number of the team that the event belongs to
• attempt_no: The attempt number that the event belongs to, starting from 1. An attempt is the duration of the team constructing a solution and submitting it together.
• turn_no: The turn number of the event, starting from 1. A turn is the duration where one of the participants is in figurative view, and the other is in abstract view (see [2] for a description of the views)
• event_no: The event number of the event, starting from 1
• time: The logging timestamp of the event from the beginning of the activity (in seconds)
• subject: The person (or the robot, or the team) that the event is executed by (A, B: the participants; R: the robot; T: the team)
• verb: The verb that describes the event (e.g. "presses", "adds", "removes")
• object: The object that is acted on by the subject performing the verb (e.g. "submit (enabled)" for subject: A, verb: "presses") 

3.3. Test Responses

This part of the dataset contains the responses of each participant in each team to the pre-test and post-test for 39 teams. 
Each test contains 10 multiple-choice (single answer) questions (i.e. items) with 3 options (recorded as 0, 1 or 2), and assesses a concept on spanning trees (see [2]).
It consists of two files: 

1. one comma-separated text file for the pre-test responses (test_responses/justhink19_pretest.csv)
2. one comma-separated text file for the post-test responses (test_responses/justhink19_posttest.csv)

In particular, the columns are:

• team_no: The number of the team, or "key" for the correct responses
• q?_A: The response of participant A to a particular item (among 10 items indexed from 1 to 10)
• q?_B: The response of participant B to a particular item

See README.md for further details.
 

Acknowledgements

This project has received funding from the European Union's Horizon 2020 research and innovation programme under grant agreement No 765955. Namely, the ANIMATAS Project.