User Intent Prediction in Information-seeking Conversations

This is the reproduction of "User Intent Prediction in Information-seeking Conversations". The source code can be found in the github. We also implement the classification with BERT. The BERT is used to encode the utterances, and BiLSTM is trained for the classification.

Data

• We used the "MSDialog-Intent" data in the MSDialog dataset.
• The data is removed in this repository.

Code

• ./BERT : the preprocessing, the generatation of encoded vector from BERT and the training of BiLSTM
• baselines_classification.ipynb: the training of the baseline ML methods
• ./neural_models: Neural models from the source code
• ./features: Feature generation for ML methods