Researchers in the Faculty of Data and Decision Sciences have developed computational models that combine eye tracking and natural language processing, and are able to identify the purpose of reading with an accuracy of about 90% within two seconds.
Researchers in the Faculty of Data and Decision Sciences are presenting at a prestigious international conference this week, study An innovative work that deals with deciphering specific aspects of reader-text relationships. The work was led by doctoral student Omer Shobi together with Kfir Hadar, a master's student, under the supervision of Dr. Yevgeny Barzak, and is presented atACL, one of the world's most prestigious conferences in the field of Natural Language Processing, taking place this week in Vienna.
Different readers, the researchers explain, have different purposes when reading a given text. Whether it is a novel, a cooking recipe, a newspaper article, or a scientific paper – any text can be read with several different purposes. Two of the purposes are הבנה (normal reading) andInformation search. The research team developed computational models that combine eye-tracking and text processing. These models are able to identify the reading target with an accuracy of about 90%. The recognition speed is also impressive – about two seconds from the start of reading.
According to Dr. Barzak, "The work is part of a broader research program in which we are developing artificial intelligence models that will allow us to infer, in real time and based solely on eye movements, the reader's linguistic knowledge, their interaction with the text, the difference between a first reading of a text and a second reading, the readability of a given text, and even the The specific information "This research paves the way for new ways to assess linguistic knowledge, personalize texts according to the reader's reading level, make textual information widely accessible to different populations, and more."
Eye tracking systems are becoming more accessible, affordable, and accurate, and some of the technologies allow this monitoring to be performed using common devices such as iPads and phones. The researchers hope that these technologies will accelerate the use of the models they have developed, to the benefit of users and content providers in worlds such as educational institutions, government agencies, media outlets, and more.
Dr. Yevgeny Barzak, faculty member in the Faculty of Data and Decision Sciences and head of The Language, Computation, and Cognition Laboratory, joined the faculty in 2021 after a doctorate and postdoctoral fellowship at MIT.
More of the topic in Hayadan:
