Web Table Formatting Affects Reading Speed on Mobile Devices

This repo contains the code and data collected for our WWW '23 paper, Web Table Formatting Affects Reading Speed on Mobile Devices. This includes tools for designing table formats and performing our data collection and analysis.

Code Documentation

This code is composed of two parts: (1) a node JS web server and (2) Python3 scripts for offline data analysis.

Web Server Documentation

To run the web server (default port is 3002) on localhost, run

node server.js

You might need to install some node modules, which you can do with npm install fs.promises express. The served files live in the public directory.

Designing Tables

http://localhost:3002/style_table.html?table_id=1 is the endpoint for our table design tool.
The table_id query param can be set to reference any HTML table file in the public/tables directory, which is currently populated with the tables 1-28 used in our study. The tool itself has instructions for how to use the tool, but when the table design is saved, it writes a JSON file to style_out directory.

Viewing Table Designs

There are 2 basic endpoints for viewing the table designs saved in the style_out directory:

• http://localhost:3002/view_result?worker_id=designer1&table_id=1
• http://localhost:3002/view_result2?worker_id=designer1&table_id=1 The only difference is whether the resulting table is viewed within an iframe sized to mimic a mobile device size (view_result2) or not (view_result). The query params are used to look up a saved result. The required worker_id query param refers to the User Id field used when saving a design on the style_table.html tool. table_id is required as well. If the date query param is provided, then a result saved on that particular datetime (represented as integer epoch offset, e.g., 1636392296340) is retrieved. If date is omitted, then the most recent result matching worker_id and table_id is used.

Under the hood, both of these endpoints redirect to http://localhost:3002/view_table.html and http://localhost:3002/view_table2.html with the loaded style parameters encoded as query params.

The repo contains the designs produced by 5 of our Expert Designers (anonymized with worker_id = designer[1-5]) for tables with id 1-14. It also includes style files used in the quantitative readability study for tables ids 1-28. The (faux) worker_ids of these files include spacing[2,5,10,15], plain, colborders, rowborders, striped, bad, colfreeze, rowfreeze, bothfreeze.

Readability Study

http://localhost:3002/readability.html?version=v4 is the endpoint for users to participate in the readability study. If you visit this page on a non-mobile device, it will not let you proceed and will offer you a QR code to open the study on your phone. Study responses are put into the readability_results directory as several individual JSON files so that if the participant does not complete the study, we have all responses they did complete.

Viewing Readability Tasks

http://localhost:3002/view_task.html?task=task_v2 looks for a task_config file in the data_for_viewing directory with a filename equal to the task query parameter (with a .json extension). The task_config files normally live in the task_versions/v[1-4]/config.json files, but they have been copied to the data_for_viewing directory so that view_task.html can access them. This page displays all tasks, which are composed of question, context, answer choices, answer, and table format variations.

Viewing Readability Study Results

After running the python analysis code on the readability study responses, the code produces a pair of data objects in the data_for_viewing directory. One data object has the unnormalized task timings and the other has the normalized time z-scores. This data can be visualized at http://localhost:3002/view_readability_results.html?results=spacing_zscore_data. The results selects which data object file is viewed. Optionally, pair can be passed as a query parameter to view pairwise comparisons of the table formats. E.g., http://localhost:3002/view_readability_results.html?results=spacing_zscore_data&pair=true

Offline Code Documentation

In the readability/code directory, there are a number of Python3 and shell scripts.
The ./do_all.sh script is the entry point.
First, open the script and set these script parameters appropriately:

name=spacing
task_file=../../task_versions/v2/config.json
input_dir="../study_responses/spacing"

The input_dir contains the raw responses of the readability study, i.e. input_dir should have a subdirectory for every participant who took the study named with the datetime (i.e. session id) when they started the study. In each of those subdirectories, there should be a few JSON files for each task. task_file is the task_config file used for the readability study when the responses were corrected. name you get to choose since it's just used to name filenames for the output files.

First, do_all.sh will perform the iterative filtering to remove readability study responses/participants who are outliers. Then it will aggregate the data into data objects (both normalized and unnormalized task timings) and put them in the code/out and data_for_viewing directories. A CSV file, where each row is the data for a participant (including their survey responses) is also put in the code/out directory. It creates a file with the mean/std used per table for the z-score transformation in code/data_files. It also puts a file with the computed outlier ranges in that same directory. Logging and Error files are recorded in the data_files directory so you can see what is going on with the script.

Citation

If you find this code or data helpful in your work, please consider citing our paper:

@inproceedings{10.1145/3543507.3583506,
author = {Tensmeyer, Christopher and Bylinski, Zoya and Cai, Tianyuan and Miller, Dave and Nenkova, Ani and Niklaus, Aleena and Wallace, Shaun},
title = {Web Table Formatting Affects Readability on Mobile Devices},
year = {2023},
isbn = {9781450394161},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
url = {https://doi.org/10.1145/3543507.3583506},
doi = {10.1145/3543507.3583506},
booktitle = {Proceedings of the ACM Web Conference 2023},
pages = {1334–1344},
numpages = {11},
keywords = {mobile device, quantitative study, reading, web tables},
location = {Austin, TX, USA},
series = {WWW '23}
}