Open University Pair-Programming Dialogues with Audio, Video, and Code, 2023-2024

23 multimodal dialogues from Open University students recorded during Python pair-programming sessions between May-June 2023 and March-April 2024. The dataset was collected by Cecilia Domingo for a PhD project on automated processing of referring acts. It includes anonymized audio, screen-capture videos, code files, keylogs, transcripts, and reference annotations.

Use Cases

• Train models for referring expression resolution based on annotated dialogue and code context.
• Analyze collaborative problem-solving patterns based on synchronized audio, video, and keylog data.
• Develop multimodal anonymization techniques based on the provided voice-converted and filtered audio samples.
• Study the relationship between verbal dialogue and code evolution in educational pair programming.

Strengths

• Contains 23 complete multimodal sessions with synchronized audio, video, code, and interaction logs.
• Includes multiple anonymization versions (voice conversion and filters) for audio, enhancing privacy while preserving intelligibility.
• Data size of 11.6 GB indicates substantial, detailed recordings per session.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Participant profiles are described as 'very varied', which may introduce uncontrolled variability for some analyses.
• Row count is unknown, which may limit suitability assessment for certain statistical tasks.

Provenance

Source

Cecilia Domingo Merino, Open University.

Collection Method

Recorded from Open University students during pair-programming sessions using Python.

Time Range

May-June 2023 (pilot) and March-April 2024 (main).

Freshness

Last updated 2026-03-17 15:21:38.

Geography

null