Sense Text

Inspiration

The inspiration for this project stemmed from the shift to online learning during the pandemic, which resulted in many students finding themselves increasingly drawn towards social media platforms. With the unfiltered nature of platforms like Discord, we soon realized that many students could face toxic/harmful comments with no way of of processing. As a result, we created Sense Text.

What It Does

Sense Text is a bot that aims to summarize text sent from users, enhancing the key concepts of paragraphs and eliminating any filler information. It also analyzes the tone of messages based on the context, meaning, and word associations, in order to classify toxic comments into a set number of categories: toxic, severe toxic, obscene, threat, insult, and identity hate. Sense Text uses the NLP BERT model to assess the sentiment of messages and cross-reference it with a given dataset, allowing it to accurately determine the toxicity of messages sent to users.

How we built it

Sense Text was built primarily using a multitude of machine learning models. Pegasus was used for the abstractive text summarization while Node.js was used as a framework to design the bot itself.

Challenges we ran into

We ran into challenges while integrating the models into the discord bot. Applying theoretical NLP principles was also difficult given the unfamiliarity with the topic.

Accomplishments that we're proud of

We are proud of the models and the scope of the project. Sense Text has the capability of actively helping real-world people.

What we learned

We gained experience with NLP, Google Cloud, and other tools that aided in the creation of Sense Text.

What's next for Text Sense

We hope to integrate Sense Text into various educational discord servers. Furthermore, we wish to transition a similar model into video conference software such as Google Meets, Microsoft Teams, and Zoom which also dominate the online learning efforts during the pandemic. By implementing Sense Text on these platforms, it would free up professors and other teaching professionals from monitoring chats so that they can focus on the lesson at hand.