Watch: Releasing a new sentiment classification model.
Our new sentiment model is free on Hugging Face. It scores every heading, list item and sentence from very positive to very negative, gets 84% of labels exactly right and 95% within one point, and is built into DEJAN App and a public demo.
Transcript
A new sentiment analysis model has just been released as a gift to the search engine optimization community. It is small enough to run right on your laptop, and it is designed to classify text chunks into seven different levels of sentiment.
You can use it to analyze everything from customer reviews and brand mentions to competitor pages and search results. The model breaks a web page down into short chunks, scores each one from very positive to very negative, and gives you an overall weighted score. It even generates a visual heatmap of the page, showing you exactly where the tone shifts.
To build it, the creators generated more than eighty thousand text chunks using Google's latest artificial intelligence models. They trained the system to recognize natural shifts in tone across blogs, reviews, social posts, and transcripts. Built on Google's Gemma architecture, the resulting model is incredibly accurate, matching the correct sentiment label eighty-four percent of the time.
The model is currently English-only and is free to download on Hugging Face. You can also try it out right now through a public web demo or inside the Dejan app.