Handle negative examples in classifier

Prof Axel asked if it was possible to have a case where Training data consists of both positive and negative examples as it may lead to better learning of the model.

Current chatbot has only positive examples with QA, Sessa and Eliza the three options. Negative feedback removes the training example from the file and re-trains the model

Study whether it is possible to handle negative examples of training data.

Start by tracing how the classifier currently stores positive QA, Sessa, and Eliza training examples and how negative feedback removes an example and triggers retraining. Compare that flow with the requested mixed positive and negative training data, then document whether it is feasible, what behavior should change, and how completion would be verified.

Written by the indexing model from the issue text.