A collaboration between the University of Central Florida (UCF) Virtual Readability Lab and Adobe has developed an AI learning model designed to provide personalized font recommendations to enhance personal reading experiences and help people Read more efficiently . Their research shows that machine learning models can improve reading speed by matching reader characteristics with recommended fonts.
The team consists of Adobe machine learning engineers and researchers working with vision scientists, typographers, data scientists, and UCF readability researchers on Adobe’s machine learning model, FontMART.
Adobe is a member of The Readability Consortium, which leads UCF’s digital readability research, using personalized typography to enhance the reader’s reading experience.
Ben D. Sawyer, Director of the Reading Consortium and UCF Virtual Reading Lab, said: “Imagine a future where a device can identify human reading habits and customize reading formats to help people read at their best. We look forward to the day when people can read Pick up the device and read and receive information in your own unique way.”
Web page screenshot | References [1]
Sawyer and Adobe scientist Zoya Bylinskii contributed to the ideation of the study and provided guidance throughout the research process. Tianyuan Cai, a machine learning engineer at Acrobat.com, led the FontMART-related research. The study used the Font Preference Test on the UCF Virtual Reading Lab website (link at the end of the article) as a baseline value to evaluate the recommendations provided by FontMART.
The results of the study show that the recommended fonts can improve reading speed after the FontMART model matches reader characteristics with specific font characteristics.
Recommended Fonts Improve Reading Speed | Pixabay
The FontMART model, trained on data from 252 workers, learned to associate fonts with specific reader characteristics. Based on interviews with typographers, the research ultimately settled on eight fonts, including serif (i.e. Georgia, Merriweather, Times, and Source Serif Pro) and sans-serif (i.e. Arial, Open Sans, Poppins, and Roboto).
The researchers found that the effect of fonts varies from person to person . FontMART learns to build relationships between font characteristics and reader characteristics to predict which fonts are suitable for readers with different font familiarity, self-reported reading speed, and age. Among them, the age factor has the greatest influence when recommending fonts.
For example, thicker font strokes are easier to read for people with low vision, and such fonts can benefit the reading experience of older adults.
More research is needed in the later stages of the model to expand the age range of participants, to evaluate the model’s effectiveness for other reading environments like different lengths, and to expand language and associated font features to better accommodate reader diversity. Subsequent collaborations and research will help expand the features explored by the model to improve the FontMART model and enhance the personal reading experience.
references
[1] Cai, T., Wallace, S., Rezvanian, T., Dobres, J., Kerr, B., Berlow, S., … & Bylinskii, Z. (2022, June). Personalized Font Recommendations: Combining ML and Typographic Guidelines to Optimize Readability. Designing Interactive Systems Conference (pp. 1-25). 10.1145/3532106.3533457
[2]
https://ift.tt/SELXb8Y
#Related Links:
[1] Link to Virtual Readability Lab: Virtual Readability Lab
[2] Readability Consortium Forms at UCF to Push Reading Research Boundaries | University of Central Florida News
[3] Font preference test website: https://ift.tt/1NaluwB
Compilation: Cod
Editor: Jin Xiaoming
Typesetting: Yin Ningliu
Title image source: Pixabay
research team
The first/corresponding author Tianyuan Cai/ Ben D. Sawyer
Author unit
-
Adobe, United States
-
University of Central Florida, United States
Paper information
Posted in Designing Interactive Systems Conference
Posted on June 16, 2022
Paper titlePersonalized Font Recommendations: Combining ML and Typographic Guidelines to Optimize Readability
(DOI:
https://ift.tt/nY0IwPb)
Article Field Artificial Intelligence
The Future Light Cone Accelerator is an early-stage technology entrepreneurship accelerator initiated by Nutshell Technology. It provides scientists with solutions at different stages ranging from company registration, intellectual property rights, to financing needs, and team formation. Accelerate the transformation of scientific and technological achievements from the laboratory to the market, and accelerate the iteration of some scientists to become CEOs.
The Nutshell team has 12 years of experience in serving scientists. We always make suggestions from the perspective of scientists and be good friends of science and technology creators. If you are planning to start a technology business, whether you are looking for money, people, resources, or orders, you are welcome to chat with the Future Light Cone team. You can send bp or other project information to [email protected] , and leave your contact information, or add the Wechat of Guoke Hard Technology Enterprise to communicate by private message.
✦
✦
Click to read the original text to view the original paper
This article is reproduced from: http://www.guokr.com/article/462409/
This site is for inclusion only, and the copyright belongs to the original author.