TidyTuesdayAltText

An R package with the goal of providing insight into the alternative (alt) text accompanying the data visualizations shared on Twitter as part of the TidyTuesday social project

By Silvia Canelón in R Package

May 4, 2021

package related talk tidytuesday databases on notion

About the data Hex logo for the package. White with a thick black border. Inside, the TidyTuesday logo on the top half which are the words TidyTuesday in white against a broad brush stroke of black paint. On the bottom half, the words alt = "text" in black against a white background and within angle brackets to simulate html code.

The original data were collected and made available by Tom Mock ( @thomas_mock) using {rtweet}. These data are available in the TidyTuesday repository.

These tweets were processed and scraped for alternative text by Silvia Canelón ( @spcanelon)

  1. Data were filtered to remove tweets without attached media (e.g. images)
  2. Data were supplemented with reply tweets collected using {rtweet}. This was done to identify whether the original tweet or a reply tweet contained an external link (e.g. data source, repository with source code)
  3. Alternative (alt) text was scraped from tweet images using {RSelenium}. The first image attached to each tweet was considered the primary image and only the primary image from each tweet was scraped for alternative text. The following attributes were used to build the scraper:
  • CSS selector: .css-1dbjc4n.r-1p0dtai.r-1mlwlqe.r-1d2f490.r-11wrixw
  • Element attribute: aria-label
Web inspection tool being used to identify the CSS selector corresponding to the primary image of one of Hao Ye's (@Hao_and_Y) tweets with alt text

Figure 1: Example of web inspection being used to identify the CSS selector utilized for alt-text web scraping

This data package does not include data that could directly identify the tweet author in order to respect any author’s decision to delete a tweet or make their account private after the data was originally collected.1

To obtain the tweet text, author screen name, and many other tweet attributes, you can “rehydrate” the TweetIds (or “status” ids)2) using the {rtweet} package.3

TidyTuesday databases on Notion

I use the data available in the TidyTuesday repository to populate some searchable TidyTuesday databases at tiny.cc/notion-dataviz with data visualizations tagged by the dataset of the week, hashtags, mentions, etc.

The Notion 2021 TidyTuesday database showing a gallery of the most recent data visualizations in the collection, organized in a grid

  1. Developer Policy – Twitter Developers | Twitter Developer ↩︎

  2. Tweet object | Twitter Developer ↩︎

  3. Get tweets data for given statuses (status IDs). — lookup_tweets • rOpenSci: rtweet ↩︎

Posted on:
May 4, 2021
Length:
2 minute read, 318 words
Categories:
R Package
Tags:
TidyTuesday
See Also:
Revealing Room for Improvement in Accessibility within a Social Media Data Visualization Learning Community