Demonstration of a Basic Sentiment Analysis Natural Language Processing Model

Do you agree with the prediction?


Each time you click 'Generated Text' a chat is returned from the server and a sentiment prediction of the chat is automatically displayed on the bar chart. The submit button is intended for your own custom text. The submit button is the one you should click after you have entered your own text. The model making the predictions was trained on 40,000 IMDb movie reviews and was validated on 5,000 IMDb movie reviews and was subsequently evaluated on another 5,000 IMDb movie reviews.These movie reviews were released as a dataset curated by the Artificial Intelligence Lab at Stanford Univeristy.

The generated text, however, are part of a human-to-human chat dataset released by Amazon from their Mechanical Turkers. The underlying knowledge spans eight broad topics and conversation partners don’t have explicitly defined roles.

Hence, you might notice that the model struggles with the predictions from the generated text. When you enter a text that sounds like a movie review, the model might perform better.

> The training of this model and this demo wouldn't be possible without the resources obtained from the links below: