Welcome to Lite Lingo
A plain language text generator.
We created a plain language translator, accessible through a google chrome extension, to help individuals with learning disabilities. The tool we created is important for helping those with reading comprehension disabilities and learning disabilities navigate complex text. We found sources like WCAG guidelines stating that plain language helps people with disabilities that “make it harder to read, including a large range of cognitive and learning disabilities, mental health conditions, aging-related conditions, print disabilities, and cognitive impacts of some chronic or other health conditions.” To achieve our goals, we wanted to make the process of generating plain language and viewing plain language on the web as simple as possible through using generative language-models like OpenAI’s GPT-3.5 and crowd-sourced annotations. We wanted to streamline the process so that it includes highlighting existing text, right clicking, and selecting the option “generate plain language text”.
In terms of the technical details of the project, we developed a back-end server using Node.js and Amazon Relational Database Service (RDS). The plain language translation is generated by OpenAI's ChatGPT-3.5 API using a personalized prompt that we engineered and validated. The initial prototypes, designs, and branding of Lite Lingo were created in Figma. The tool itself is hosted as a Google Chrome Extension that is accessible in the context menu whenever a user highlights text and right clicks to open the context menu.
When looking for first person accounts, we had trouble finding people with reading comprehension learning disabilities such as: dyslexia, autism, ADHD, and hyperlexia who use a plain language tool for improving their comprehension. However, we did find multiple accounts of how trouble with reading comprehension was a significant obstacle in their lives and were able to manage with their disabilities. A first hand account we found was an individual named Julie who has hyperlexia, “the ability to decode words but inability to comprehend written and oral material” in her own words. She mentions that she struggled through school, particularly in achieving passing reading scores. plain language text is essential for individuals with hyperlexia because it leverages their advanced decoding skills, facilitating improved comprehension and reducing cognitive challenges associated with complex language.
Another first person account we had was an individual named Kathryn who has dyslexia and Specific Reading Comprehension Deficit (SRC D). Dyslexia is a neurodevelopmental disorder characterized by difficulties in accurate and/or fluent word recognition and by poor spelling and decoding abilities, often despite having average or above-average intelligence. Specific Reading Comprehension Deficit (SRC D) refers to a condition where individuals experience challenges in understanding and making meaning from written text, even when their decoding and word recognition skills may be intact. According to Kathryn, “you read something and you don't understand it and normally you don't have problems reading”. However, Kathryn mentions that with SRC D, the way she reads is in a mechanical technical level, meaning she finds the subject then finds a noun then find a verb. This process is similar to how to write plain language. In summary, providing plain language text is essential for individuals with dyslexia and SRC D as it simplifies the reading process, reduces cognitive load, improves reading fluency and comprehension, promotes independent reading, supports learning, reduces anxiety, facilitates communication, and allows for customizable accessibility.
Our project addresses disability justice as we aim to champion the diverse experiences of people with disabilities and actively incorporate the encounters people with learning disabilities have had in their education into our design and implementation processes. We recognize that there is a large community of people that face academic disparities due to their disabilities, and we hope to create a tool that challenges it. By prioritizing the needs and perspectives of users with learning disabilities, we hope to contribute to developing a tool that empowers users in educational environments and enable them to foster a sense of agency and advocacy. The two disability justice principles we focused on were intersectionality and Leadership. Intersectionality through recognizing the diverse needs and experiences within the community of individuals with reading comprehension learning disabilities. Also, ensuring that the tool is user-friendly and accessible to a wide range of users. Leadership through providing individuals with reading comprehension learning disabilities with the ability to create and access plain language text content easily.
After implementing our tool there were many things to reflect on. One thing were the limitations to our tool. One limitation was needing to cap the word limit per translation so that the original text and translation could fit adequately on the screen. Along with this, there are still some issues with AI generated plain language as from our validation it is not 100% perfect at following all of the plain language guidelines. Also, Chat GPT-3.5 does not take in images as input meaning that image description generation is not feasible using just Chat GPT-3.5.
In terms of future plans for Lite Lingo we plan to incorporate user feedback into our extension. We plan to have a feature that allows the user to select “Yes'' or “No” if the translation is helpful. The user feedback will be taken into account to tweak the prompt to ensure that the AI-generated text is helpful to users and to rank the translations for users to find easier. Another future plan we have is to have multi turn processing for the prompt. This basically means to iterate upon the AI-generated translation, meaning running it through Chat GPT multiple times to get an even more accurate plain language translation. The third feature we plan to implement is generation of image descriptions through the use of Chat GPT-4.0 or another generative AI that takes in images as input.
To make our app accessible, we first decided to include the extension in the context menu. This allows for seamless access as a user just has to highlight a block of text and right click to open the context menu and click on the Lite Lingo extension to access the tool. In terms of the user interface of the app, we made sure to have distinct color contrast throughout all the pages of the app. We realized color contrast is very important from an accessibility standpoint to ensure that text and visual elements like buttons have distinct and easily perceivable differences in color. Another accessibility feature we made to include were instructions on how to use the tool. As soon as the user opens the tool they are met with a popup window that gives them the full instructions on how to use Lite Lingo. Additionally, we made sure to include two panels for the original text and the plain language text translation as a side by side comparison. This was important to include as users with reading comprehension disabilities can benefit from having a visual reference to compare the original text with the plain language text translation. The side-by-side format aids in understanding and processing the information. We also made sure to follow W3C guidelines in our implementation. Lastly, we ensured that the extension was usable with a screen-reader by using the ARIA web tools for finding highlighted annotation and navigating the interface.
A video demonstrating how to use Lite Lingo on Google Chrome.
