Artwalk Corpus

If you use this data in your research, please refer to and cite: Kris Liu, Jean Fox Tree and Marilyn Walker. "Coordinating Communication in the Wild: The Artwalk Dialogue Corpus of Pedestrian Navigation and Mobile Referential Communication," Language Resources and Evaluation Conference (LREC), Portorož, Slovenia, 2016.

Overview: The Artwalk Corpus is a collection of mobile-to-Skype conversations between friend and stranger dyads performing a real world-situated task. This task is a hybrid of two traditional lab tasks, where participants are directed to a location (map task, e.g. Anderson et al., 1991) and find a target object from a variety of objects (tangram task, e.g. Clark & Wilkes-Gibbs, 1989). Artwalk captures how conversation sounds when one partner has to walk, talk and navigate everyday obstacles in a small city environment, such as cars and pedestrians.

The Data: Corpus consists of transcripts of 24 friend and 24 stranger dyads who did this two-round referential communication task. In total, it contains approximately 185,000 words and 23,000 turns, from conversations that ranged from 24 to 55 minutes. It includes referent negotiation, direction-giving and small talk (non-task talk). Additional information includes selected post-experiment questionnaire responses, participant genders, sunset times, weather conditions, photos of the targets and their coordinates.

Download: Fill out the following form to download the Artwalk Corpus.

GitHub: https://github.com/zhichaohu/artwalk-corpus

Contact: Please direct questions to Zhichao Hu: zhu [at] ucsc [dot] edu

Website last updated June 21, 2024.

Main menu