Kyle He¹ and Carlos Gonzalez², ¹USA, ²UC Berkeley, USA

SyntaxStar, a desktop program management system, aims to enhance writing quality for users. Ineffective word choices hinder engagement and effective communication, contributing to reading struggles among eighth graders which often leads to a challenging pathway in higher education [7]. SyntaxStar addresses these concerns, emphasizing the importance of writing as a fundamental communication skill, and is driven by the goal of avoiding repetition and improving clarity. Upon creation of this project, things to note are that there exists no current method to update the software automatically for users. There is also no current way of receiving user feedback, which can significantly maintain the longevity and integrity of the software design. Despite this, the pipeline has been solidified and working with an interactive frontend, a comprehensive backend that hosts the application and machine learning model [9]. With regards to performance, the concept of word similarity in the Word2Vec library was applied to not only cross-check words with their paired synonyms but also to assign a score for these synonyms [8]. Cosine similarity scores were applied and distributed onto a plot which showcases how accurately the model classified words with their appropriate synonym. In the future, the model can improve by looking at a more intensive preprocessing section by shuffling words and their parts of speech, prior to recommending a synonym. This is for the sake of adding variability and reducing bias for synonyms that wouldn’t make sense in certain contexts. To propel future growth, more marketing and publications should be incorporated to generate awareness and usage of the app. Developing more components for the software, such as a dedicated website serving as the landing page for a downloadable version, would also be beneficial.

Adaptive, Desktop Program Management System, NLP, Machine Learning