Michael Hom

Handwriting Transcription Website

Started Sept. 28, 2021

Abstract or project description

The problem is converting images of handwritten text to computer text. Existing handwriting transcription solutions are expensive and do not offer a method of uploading your own images to be translated. The aspect which makes my solution unique is to make it accessible to the general public and to upload images of handwriting to the website. I am using the library, Tesseract.js, as a foundation of my conversion of the text. Then, I modified the output to measure accuracy and plausibility with the help of an English words frequency and dictionary database. These two separate databases guided the output to be more structured compared to the initial outputs. Plausibility can be measured through finding the number of existing words or finding how many edits it takes to transform the projecting outcome into an existing word. The final deliverable will be a website and a GitHub repository. To create the website, I used the coding languages, HTML, CSS, and JavaScript. The website will allow for the user to upload images of handwritten text to be converted into computerized text and will output the text in consideration with user final adjustments. An example button is provided to see how the program functions. There are drop down menus for words that Tesseract translates which do not exists in the dictionary database for users to pick what should be the correct translation. The GitHub repository holds the code required to create the complexities of the website along with the various databases used in order to interpret the images provided by the user.