-
Screenshot of the ToneTag Config
Inspiration
Many people struggle to read emotion in emails and other forms of media and may rely on tone tags/indicators to keep up, while many others don't understand how to use tone tags. With /ToneTag Suggester, people can begin to learn how to use tone tags in their communications for better inclusivity in the workspace, schools, and more.
What it does
It will suggest 4 basic tone tags for users to insert into their emails and other communications
How I built it
I used Pandas to combine multiple datasets and relabel them to fit the description of 4 basic tone tags, /j,/pos,/neu, and /neg. Once I created the combined dataset I began creating my model and associated each tone tag to a number. The chrome extension was supposed to be able to detect highlighted text and when right clicked would offer a suggestion for a tone tag to use.
Challenges I ran into
I ran into multiple challenges including issues with Chrome Extension Developer Tools which prevented me from finishing my extension on time. I found a lot of things were deprecated (also that my rustiness in JavaScript definitely held me back). I was able to create a pop up window with HTML/CSS, but was unable to figure out the logic for grabbing highlighted text. I also ran into issues with finding appropriate sentiment analysis databases. Most public sentiment analysis databases were crude or were of social media posts and needed to be "cleaned" of hashtags, emojis, @ symbols, and more. This ended with me having to combine multiple datasets to get something I could train my model with. I feel like this ate up most of my time and limited what I could do with the rest of my project. When it came time to try to get my model ready to be turned into a Tensorflowjs, I was unable to install Tensorflowjs due to tech limitations (apparently the dependencies for Tensorflowjs are too great for my measly laptop)
What I learned
I gained a much better understanding of how to use Pandas with csv files and DataFrames. I had a moderate understanding before, but since I have created multiple datasets for this project alone I feel much more familiar with the library. Similarly with Tensorflow I had a baseline understanding of it but now
What's next for Tone Tag Suggester
I would like to get my Tensorflow model in a place where I can serve it via an API. I may transition away from utilizing a chrome extension to making it into a website where one can copy and paste emails and it will suggest a tone tag for each sentence. I would also like to increase the database my project trains on as it is currently very small for what it is and would like to expand to adding more tone indicators such as /srs and /gen.
Log in or sign up for Devpost to join the conversation.