Supplemental Material

Accessible Visualization via Natural Language Descriptions: A Four-Level Model of Semantic Content
Alan Lundgard and Arvind Satyanarayan
IEEE Transactions on Visualization and Computer Graphics (TVCG), Special Issue on the 2021 Visualization Conference (VIS)

corpus_sentences_labeled.json: Our corpus, consisting of all visualization descriptions (582 total) and labeled sentences (2,147 total), organized as follows.
- vis id number: A unique four-digit id number.
  - descriptions: A list of each author id and the description they wrote for a given visualization.
    - sentences: A list of each sentence in the description, labeled according to our four-level model of semantic content.

corpus_summary_and_evaluation.pdf: A typeset document containing the following.
- Corpus Summary
  - Corpus descriptive statistics.
  - Corpus fingerprint visualization.
- Evaluation Design
  - Examples of the rank-choice interfaces.
  - All questions shown in the rank-choice evaluation.
- Evaluation Questionnaire
  - Demographic questions.
  - Visualization questions.

/evaluation: A folder containing data and code from the rank-choice evaluation.
- calculate_stats.py: Code for calculating the statistics reported in the paper.
- blind_rankings_numeric.csv: Rank-choice data from blind readers, in numerical form.
- blind_rankings_text.csv: Rank-choice data from blind readers, textual form.
- sighted_rankings_numeric.csv: Rank-choice data from sighted readers, in numerical form.
- sighted_rankings_text.csv: Rank-choice data from sighted readers, in textual form.

/visualizations: A folder containing the visualizations (50 total) used for gathering the corpus, each file named according to the following.
- vis id number: A unique four-digit number.
- chart type: Bar, line, or scatter.
- difficulty: Easy, medium, or hard.
- topic: Academic, business, or journalism.