ToneIt!

Inspiration

Millions of people are switching to remote work since the pandemic hit and most are remaining in remote or semi-remote positions. As students we noticed ourselves getting exhausted and confused trying to discern the meaning of different articles, emails, and websites. We further found our neurodivergent friends and community being disproportionally affected by the lack of in-person interactions (limited if not non-existent non-verbal communication). It quickly became apparent with the increasing amount of remote work that there needs to be a system to support people in assessing tones and meaning of passages across all website!

What it does

ToneIt is a browser extension (currently designed on Chrome but we plan to expand across different browsers) that leverages natural language processing (sparkNLP in python) to identify the tones of paragraphs across any website. It includes features like:

• Marking the tone of the entire paragraph
• Highlighting the sentences matching the paragraph tone
• Customization in two forms:
  • Adding custom tones based on user-inputted paragraphs matching the custom tone
  • Changing tone colors to facilitate high-contrast colors as well as user preference

How we built it

1. Defined the problem space and area of need
   • Researched what support systems are already out there
   • Identified feasible plans by brainstorming via ideation multiple ideas
   • Selected plan via the time it would take and how closely it matched the need
2. Designed the product
   • Wireframed multiple configurations
   • Voted and discussed ideas and went back to the drawing board
   • Settled on final layout of the pages and extension
3. Identified the technological aspect of this product
   • Searched through and tested multiple NLP frameworks
   • Tested different programming languages to see which meshed best
   • Examined how to accommodate for differential programming language experience
4. Created the example design and product
   • Coded a proof of concept example (basically backend related code)
   • Constructed a Figma example for high-fidelity prototype

Challenges we ran into

1. Going back and forth between different languages
2. Considering the different forms websites may come under
3. Finding a valid NLP model that could train quickly and be tuned
4. Identifying how a user would navigate and what features they might need

Accomplishments that we're proud of

1. Built runnable E2E scripts that analyzes texts and outputs tones by leveraging machine learning
2. Designed a complicated yet organized software architecture
3. Created a high-fidelity prototype collaboratively through Figma
4. Learned a completely new NLP framework on the go
5. Divided up work quickly and worked together to solve problems
6. Communicated effectively and weren't afraid to reach out with questions

What we learned

1. Figma - Designed prototype with components, interactions, and possible cases
2. Machine Learning NLP - Developed engine trained with datasets to predict the tone of paragraphs
3. Building runnable software - Developed data pipeline from raw HTML to analyzed information
4. Collaboration & teamwork - Work collaboratively with peers to brainstorm, ideate, and build a product

What's next for ToneIt!

• Finish Chrome extension development
• Develop a website for account management and better user experience
  
• Lighter ML model
• Collect and process our own data to train the engine
• Improve on defining sentences and paragraphs
• Increase confidence level
• Implement sharing tone function
• Develop a solid business model
• Test on actual users and gather feedback

Built With

• figma
• html
• java
• machine-learning
• natural-language-processing
• python