AUTOR is based on scientific research involving adults with a diagnosis of autism. The research studies were conducted at the University of Wolverhampton, UK.
You can scroll down for a list of peer-reviewed publications but before that we explain our main findings underpinning the development of AUTOR. You can also watch a TEDx video about AUTOR here: https://www.youtube.com/watch?v=5jNwceqD06g
The main technology we used to study how people with autism read and use the web is called eye tracking. It follows the eyes of a person on the screen and this way we are able to analyse the places which they look at and for how long. Sometimes these “places” could be single words or entire phrases, which gives us valuable information about which parts of the text are difficult to understand. Below you can see an image of how eye tracking works. A person is reading a text about chemical elements. When they read a complex word such as “sulphur” or “chlorine” their eyes spend longer time processing the difficult word, which is why the dots on the complex words are bigger than the dots on the easy words.
Some of the main findings from our eye tracking studies are listed below.
- Linguistic factors which significantly affect the reading comprehension of people with autism include (but are not limited to): the number of words per sentence, the number of metaphors per text, the average number of words occurring before the main verb in a sentence, and the similarity of the syntactic structure in adjacent sentences. If a text contains less of these, then readers with autism will be able to comprehend it better.
- Participants with autism spent significantly longer time looking at images inserted in the text, compared to readers without autism. This means that attention works differently in autism and may have implications for how people with autism read. This is no surprise given all previous studies on this subject but mind that inclusion of logos, advertisements or any other visual information, which is not directly relevant to the meaning of the text will distract readers with autism more than the rest of the readers.
- Images inserted into text have statistically significant effects on the subjective perception of readers with autism on how well they comprehend and memorise the text but not on their actual comprehension and memorisation. This means that they strongly prefer to read texts with images, even though this may not actually help them comprehend and memorise the texts better.
- Both photographs and symbols are suitable for adults with autism. Bear in mind that this may not be the case for children.
- Participants with autism take significantly longer to read a text. This means that in the case of videos, you may have to allow longer times for the users to read the text or captions and to process the visual information.
- Participants with autism found it significantly more difficult to find information on web pages compared to people without autism. We still do not know why this is the case but we have found that the two groups search for information in a different way: participants with autism tend to “check” more elements of the web page before they arrive at the one they are looking for. Below you can see the scanpath of a person without autism (in green) and a person with autism (purple).
We are currently conducting more experiments to find what helps readers with autism read better. Keep an eye on this page if you want to find out about our future results.
The research studies about the development of AUTOR are published in a number of peer-reviewed publications. You can see the list of already publications below. Some of the publications on AUTOR are still under peer review and this list will be regularly updated to feature the new content.
- Yaneva, V., Temnikova, I. and Mitkov, R. 2016. A Corpus of Text Data and Gaze Fixations from Autistic and Non-autistic Adults. Proceedings of the 10th edition of the Language Resources and Evaluation Conference (LREC), Portoroz, Slovenia, 25 – 28 May
- Yaneva, V., Temnikova, I. and Mitkov, R. 2016. Evaluating the Readability of Text Simpliﬁcation Output for Readers with Cognitive Disabilities. Proceedings of the 10th edition of the Language Resources and Evaluation Conference (LREC), Portoroz, Slovenia, 25 – 28 May
- Yaneva, V., Evans, R. and Temnikova, I. 2016. Predicting Reading Difﬁculty for Readers with Autism Spectrum Disorder. Proceedings of Workshop on Improving Social Inclusion using NLP: Tools and Resources (ISI-NLP) held in conjunction with LREC 2016, Portoroz, Slovenia, 23 May
- Yaneva, V., Temnikova, I. and Mitkov, R. 2015. Accessible Texts for Autism: An Eye-Tracking Study. ASSETS 2015. The 17th International ACM SIGACCESS Conference of Computers and Accessibility, Lisbon, Portugal, 26-28 October. pp. 49-57
- Yaneva, V. and Evans, R. 2015. Six Good Predictors of Autistic Text Comprehension. In: Proceedings of the International Conference Recent Advances in Natural Language Processing (RANLP 2015), Hissar, Bulgaria, 5-11 September 2015. pp. 697 – 706
- Yaneva, V. 2015. Easy-read Documents as a Gold Standard for Evaluation of Text Simplification Output. In Proceedings of the Student Research Workshop at the International Conference on Recent Advances in Natural Language Processing (RANLP 2015) , Hissar, Bulgaria, 5-11 September 2015. pp. 30-36
- Niculae, V. and Yaneva, V. 2013. Computational considerations of comparisons and similes. In: Proceedings of ACL 2013 Student Research Workshop, pp. 89-95. Sofia, Bulgaria, August 2013.