Creating accessible search engines for users with dyslexia

Accessibility of systems has become a significant issue for professional who deliver software products and services, and this includes search technologies. Users are often reliant on these technologies in different contexts (think of Google), and therefore accessibility of information retrieval systems is a key issue for professionals. There are many disabilities, but one which stands out is dyslexia – it is estimated the condition is found in around 10% of the population. Some will not know they are dyslexic, and may live substantial parts of their lives without diagnosis – if indeed they are ever diagnosed). Others may not want to reveal this invisible cognitive difficultly either because they fear that it will be career limiting or because they feel stigmatised by society. Either way search professionals need to consider methods and strategies to build accessible systems and services for dyslexic users.

What is dyslexia?

There are many misconceptions of dyslexia. Often when it is discussed, it is treated as a reading and writing problem. However, this is very misleading as the impairment has much wider implications than just reading and writing. Dyslexia is one of many Specific Learning Difficulties (SpLD), in which general intelligence is unaffected, but there are specific problems which cause those that have the condition problems. In the case of dyslexia this is a phonological deficit problem, where words cannot be deciphered or expressed due to the person’s inability to use or manipulate words – dyslexia means ‘problem with words’. Therefore this not only means reading and writing, but speaking and listening also. Dyslexia is not a condition which specifies a single set of characteristics, but a spectrum of conditions in which strengths and weakness in different areas varies. Some users are very good at reading, but have significant problems writing and vice versa. There are a number of key characteristics of dyslexia which include problems with short term or working memory and rapid naming (the ability to retrieve and use words). But way do we need to consider dyslexic users when we are designing search technology software and services.

The problem to address

With legislation in place (The Equalities Act 2010), there is a legal requirement for reasonable adjustments and products and services need to be produced in this light. Of course we want to help our users’ irrespective of any legal requirements given our focus on professional ethics! But how can the condition effect search?

How can it effect search?

The best way to consider the effects are to consider a practical example. Google is a heavily used service worldwide, and web search is therefore a good place to start considering the problems which can occur. Note that other contexts such as enterprise and professional search can be considered, but the basics can easily be outlined using the web context. We can consider web search to have three main phases: 1) issuing a query, 2) inspecting results lists and 3) reading web pages or documents. From this we can identify the problems to address.

The first of these is query formulation – this could either be a misspelt word or an inability to recall or remember a word which of course can inhibit or prevent the retrieval of documents which are relevant to the users information need. Once the query is issued a set of results is presented to the user, at which point problems with reading start to occur. Can the user extract unfamiliar words, which can be useful in query reformulations? Different format of results may cause confusion (e.g. links to images, the knowledge graph seen in universal search schemes). Then there is the issue of browsing – clicking back and forth between results lists and documents may overload the user with information and they might not be able to pick up all they need due to working memory limitations. Once the document or web page is retrieved, the challenge in terms of reading is significantly increased due to the volume of text, which in turn impacts on the ability to absorb information from it due to working memory limitations. All these issues make the search process very time consuming if not exhausting for dyslexic users. What is known about the problems from research done by academics so far?

What is known?

Up until 2010, very little work had been carried out on how dyslexic users search for information. With colleagues I undertook an initial pilot study on the experimental Okapi system  and found a clear difference between how users with and without dyslexia searched for information – dyslexic users did fewer searches but spend more time on them and they did more backtracking in documents (from eye tracking data). That is dyslexic users scanned the document up and down more as they were forgetting what they had read previously. Subsequently, other research has been carried out in all three phases mentioned above.

My colleague Gerd Berget of Oslo Metropolitan University in Norway has investigated the problem with query formulation and reformulation and found that dyslexic users took much longer to search for information when there were no query building aids or spelling support (the kind you get on Google with auto completion of words on the search box). Research done at Microsoft also identified problems querying on compact devices such as smart phones . Clear problems in the initial stage of querying has been identified. Together with other colleagues, I followed up the initial study with a more in depth experiment with more users, and found a clear link between memory and judging the relevance of documents – users with dyslexia tended to judge a significantly less number of relevant documents and this was collated with only one aspect of their cognitive profile, namely poor working memory. This confirmed the problems found in the initial study and research done by Microsoft research has also confirmed that this is a significant issue.

A PhD study carried out by Areej Al-Wabil showed problems navigating through linked structures (e.g. from results lists and clicking through web pages to find information) . She also found that dyslexic users spend more time on tasks, taking longer to complete them due to phonological, working memory and visual stress. This work has implications for examination of both results lists and documents. This is just a flavour of some of the work done and being carried out, but it does give us a good idea of the problems dyslexic users face when searching. We now turn to solutions to the problem.

Practical solutions to problems identified

The solutions to the problems listed above lie mostly in the deployment of assistive technology of various kinds which will vary depending on the phases of search being considered. The initial stage of querying can be supported by spell checking and error tolerance (the kinds of strategies Google regularly uses to help users with entering their search terms). This can make a real difference to the user, but strategies to build such query support can be considered. For example different types or strands of dyslexia imply different kinds of spelling mistakes, and if the search professional has access to the corpus they can analyse the text with these different strands to pick up different forms of words to support when the user is issuing their keyword. This is a good strategy to think about as dyslexic users have their own specific spelling issues, which are different from the general population. In terms of unfamiliar words (which the user has not encountered before), methods to highlight the word, speak it aloud and provide a definition would help the user in adding new words to their query. Different ways of issuing queries can be considered e.g. for some types of dyslexic users voice search may be more effective as they can express their search terms more effectively orally than typing them in.

There are various technologies and methods which can be used to assist dyslexic users examining and interpreting results lists and documents. One key issue to consider is that a substantial minority of users with dyslexia suffer from visual stress, and methods to adapt the presentation of information to the user e.g. choosing colours which help them read the text better. This has been shown to help those types of users with reading. A further method (already mentioned) is to provide read aloud functionality, so that users can highlight blocks of text to be read which helps with comprehension of the material. Systems such as TextHelp can be used as an example, but many applications now provide such facilities e.g. MS Word and Apple products e.g. Siri. To help with working memory, various aids can be considered e.g. text pasted can be saved to a space which can be retrieved by clicking on an icon – the source of that material would also be kept.

In general terms we can consider the use of universal design in search technologies e.g. by resolving the problems for dyslexic users we can also assist the general population. Alternatively we can design accessible technology from first principle to address the problem of this group, and create responsive applications which will adapt to the users’ needs depending on the context e.g. the needs users on desktops will differ significantly from those who search for information on mobile devices.


Professionals working on search technologies need to consider various kinds of users, and we hope that this article highlights the problems dyslexic user face when using search to resolve their information needs, and some solutions which can be considered to resolve the problems. There is still much to learn about dyslexic users’ information searching behaviour as we only have a limited understanding of how dyslexia actually impacts search, but this is a first step. Interested members should watch this space!

Useful resources

Gerd Berget and Andrew MacFarlane. 2019. What is known about the impact of impairments on information seeking and searching? Journal of the Association for Information Science and Technology. Available on:

British Dyslexia Association Web site:

Margaret J. Snowling. 2000. Dyslexia: 2nd Edition. Blackwell Publishing.

About Andy Macfarlane
Andy Macfarlane

Andy is a Reader in Information Retrieval in the Department of Computer Science at City, University of London, and is a member of the Centre for HCI Design. He is the past Chair of the BCS Information Retrieval Specialist Group and is a long standing member of that SG. He is a Fellow of the BCS.

Leave a Reply

You must be logged in to post a comment.