GP-TSM Demo

This is a demo of GP-TSM (Grammar-Preserving Text Saliency Modulation), a text rendering technique aimed at enhancing reading and skimming experiences. This technique is detailed in our paper presented at CHI 2024:

An AI-Resilient Text Rendering Technique for Reading and Skimming Documents.
Ziwei Gu, Ian Arawjo, Kenneth Li, Jonathan K. Kummerfeld, Elena L. Glassman.
In the 2024 ACM CHI conference on Human Factors in Computing Systems
CHI 2024, May 11-16, 2024, Honolulu, HI, USA

Read the full paper here.

The purpose of this web demo is for you to explore and assess whether GP-TSM may be helpful to you. Please input any text you are comfortable sharing below. Do not input any private or sensitive information. For quality assessment and tool improvement purposes, your text inputs and GP-TSM's outputs will be logged. Note that while GP-TSM has been tested on GRE reading passages, its effectiveness on other types of text may vary. GP-TSM works on text of any length but the demo is intentionally limited to 1000 characters. Feel free to use the code yourself to apply it to longer text.

You will be asked to enter your OpenAI API key. Please note that we do not track or store your API key in any way—it is only required for making calls to OpenAI's GPT-4 large language model, which GP-TSM depends on. Typically, the system makes about 50-100 API calls for each paragraph you input. If you have concerns about this usage, or if you wish to try our tool but do not have an OpenAI API key, please contact us at ziweigu@g.harvard.edu.





paragraph # paragraphs
User activity logs downloaded successfully!