Where and how are Computer Science researchers publishing datasets?

Question

We are publishing articles that use several different datasets acquired from user (clinicians, physicians and radiologists) analysis through interaction to our healthcare systems [1, 2, 3]. We would like to share them so that other researchers can use the same data for post-process analysis and verification.

Our work is a relation between both Human-Computer Interaction (HCI) and Health Informatics (HI) fields by study human behaviour, as a user of an AI-Assisted system for breast cancer diagnosis. From here, we measure several attributes, with several scales (e.g., NASA-TLX, SUS, Time, Number of Clicks, Number of Errors and Qualitative Analysis). It seems to us really interesting to not only publishing around our work but also to provide the information to other researchers.

Our question is as follows:

What strategies are available to share our dataset?

In a near future, we would like to take advantage of Google Dataset Search. However, our idea will be to publish the dataset on Kaggle and then publish an arXiv document to link the dataset. By doing this, we will not only promote the dataset for the scientific community but also have the chance of being cited by.

---

^{[1] Francisco M. Calisto, Alfredo Ferreira, Jacinto C. Nascimento, and Daniel Gonçalves. 2017. Towards Touch-Based Medical Image Diagnosis Annotation. In Proceedings of the 2017 ACM International Conference on Interactive Surfaces and Spaces (ISS '17). ACM, New York, NY, USA, 390-395. DOI: https://doi.org/10.1145/3132272.3134111}

^{[2] Maicas, Gabriel, Gustavo Carneiro, Andrew P. Bradley, Jacinto C. Nascimento, and Ian Reid. "Deep reinforcement learning for active breast lesion detection from dce-mri." In International Conference on Medical Image Computing and Computer-Assisted Intervention, pp. 665-673. Springer, Cham, 2017.}

^{[3] Tiago Dias, Helder Araujo, and Pedro Miraldo. 2016. 3D Reconstruction with Low Resolution, Small Baseline and High Radial Distortion Stereo Images. In Proceedings of the 10th International Conference on Distributed Smart Camera (ICDSC '16). ACM, New York, NY, USA, 98-103. DOI: https://doi.org/10.1145/2967413.2967435}

Answer 1

Zenodo is a EU-funded dataset storage site ran by CERN that seems to be slowly gaining traction. Each dataset gets a DOI immediately when you publish it, so you can refer to the dataset in your manuscript. You can also update your dataset later, which gives a new DOI to the newer version and still keeps the older version there for reference.

A workflow could look something like: (1) Publish dataset in Zenodo, you get DOI for v1. (2) Refer to v1 in your manuscript, submit it. (3) Referee wants to have more stuff in the dataset and you agree. (4) Update your dataset, get a new DOI for v2. Everyone who uses the link to v1 also will be notified that there's a new version available. (5) Submit your edited manuscript, refer to v2 there. (6) Paper gets published, readers see a reference to v2.

Answer 2

What strategies are available to share our dataset?

Various options are available, for instance, you could

• make your dataset publicly available on a website (e.g., your personal website); or
• submit your dataset, along with an accompanying manuscript, to a journal that will make it available.