# Data, content, and software notices This repository is a collection of independently licensed materials. The MIT license in `LICENSE` applies only to original website software; it does not relicense corpus content, datasets, model outputs, papers, names, or logos. ## License layers | Material used by the site | Source | Applicable terms | | --- | --- | --- | | Authentic maps and dialogue transcripts | HCRC Map Task Corpus Annotations v2.1, Human Communication Research Centre, University of Edinburgh and University of Glasgow, © 2007 Human Communication Research Centre | [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/), as stated for downloads on the [official HCRC download page](https://groups.inf.ed.ac.uk/maptask/maptasknxt.html) (checked 31 August 2026) | | Perspectivist landmark interpretations, reference-expression annotations, schemas, and derived metadata | Grounded Misunderstandings in MapTask (GMMT), Nan Li, Albert Gatt, and Massimo Poesio | [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/) | | Interpretation-matching records and released result extracts | Seeing Is Not Sharing (SINS), Nan Li, Albert Gatt, and Massimo Poesio | [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/) | | Original HTML, CSS, JavaScript, and repository scripts written for this site | GMMT × SINS project page contributors | MIT; see `LICENSE` | The upstream SINS **code** also has an MIT license, but that upstream license is separate from this repository's license. Third-party libraries or assets, if added later, retain their own notices and must not be assumed to be MIT merely because they appear in this repository. ## Why maps and dialogue can be shown Serving a map image or dialogue excerpt on a public web page is “Share”—copying and redistributing the material in a medium or format—under CC BY 4.0. It is therefore not outside the license merely because a visitor views it in a browser. CC BY 4.0 expressly permits sharing and adaptation for any purpose, provided the project gives appropriate credit, links to the license, indicates whether changes were made, and does not impose additional restrictions. This notice is intended to satisfy those conditions for the included HCRC, GMMT, and SINS material. It is not a substitute for reading the licenses or for legal advice. ## Attribution and changes ### HCRC MapTask Suggested attribution: > HCRC Map Task Corpus Annotations Version 2.1, Human Communication Research > Centre, University of Edinburgh and University of Glasgow, © 2007 Human > Communication Research Centre. Obtained from the official HCRC MapTask > download page and used under CC BY 4.0. Changes made for this site: selected official full-size PostScript maps were converted to PNG and proportionally downscaled so that the longest edge is at most 1024 pixels. They are not cropped and their depicted content is not edited. Clickable landmark hotspots, selection states, and highlights are separate HTML overlays added by the project page; they are not part of the original maps. The homepage presents a moving context window around the selected reference; the corpus browser presents all 128 transcripts. Text is reconstructed from preprocessed timed units (verified against the original XML), with reference spans anchored by the released timed-unit IDs. Original words, chronological line order, speaker roles and utterance IDs are retained; whitespace is normalised. Clickable highlights and reference-ID buttons are added. Text may be reflowed to fit the interface. The four existence/lexical/multiplicity/ identical gallery overlays are explanatory, not model outputs. No semantic changes to the quoted dialogue are intended. The page must not state or imply that HCRC, the University of Edinburgh, the University of Glasgow, the corpus authors, or any participant endorses this project or its model-generated interpretations. The site does not host or use MapTask video. The HCRC corpus website explains that its subject videos were never made freely available for download in order to protect participant privacy. This project also does not publish corpus audio. ### GMMT and SINS The site reformats the full released annotation layer, attaches human-readable labels, and joins GMMT interpretation records to SINS model judgments by reference ID. Interactive filtering, hotspot coordinates, color coding, and model/source badges are project-page additions. The underlying annotations and judgments are not presented as newly authored website code. Model-generated interpretations are labelled with their model. Human-verified records are labelled separately. A SINS Yes/No prediction is an interpretation-matching judgment and must not be described as the model's own prediction of the giver and follower landmark IDs. ## Citation request Publications, screenshots, or derivatives based on this visualization should cite: 1. Anderson, Anne H., et al. (1991). “The HCRC Map Task Corpus.” *Language and Speech*, 34(4), 351–366. 2. Li, Nan; Gatt, Albert; and Poesio, Massimo (2026). “Grounded Misunderstandings in Asymmetric Dialogue: A Perspectivist Annotation Scheme for MapTask.” *Proceedings of LREC 2026*, 4988–5001. 3. Li, Nan; Gatt, Albert; and Poesio, Massimo (2026). “Seeing Is Not Sharing: Some Vision-Language Models Overestimate Common Ground in Asymmetric Dialogue.” *Proceedings of SIGDIAL 2026*, 694–710. ## Downstream use Anyone who copies, republishes, or adapts the corpus-derived files must preserve the relevant attribution, license link, and modification notice. Do not remove source metadata, add technological restrictions that prevent recipients from exercising CC BY 4.0 rights, or use attribution in a way that suggests endorsement.