Skip to content
User Intent Prediction in Information-seeking Conversations
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
features
neural_models
LICENSE
README.md
data_helper.py
fetch_utterances_from_db.ipynb
gen_ground_truth.ipynb

README.md

User Intent Prediction in Information-seeking Conversations

This is the implementation "User Intent Prediction in Information-seeking Conversations". This paper has been accepted to CHIIR 2019. If you find our repo useful in your paper, please cite our work.

Our models predict user intent for utterances in information-seeking conversations. In other words, the models conduct utterance type classification. One utterance can have more than one user intent label. These models are built with Python 2.7 and Keras.

Data

We used the "MSDialog-Intent" data in the MSDialog dataset.

Code

  • ./fetch_utterances_from_db.ipynb and ./gen_ground_truth.ipynb: conduct preprocessing as described in the paper.
  • ./features/: extract features
  • ./neural_models/: different stand-alone neural models

Notes

At the time of developing, we preprocessed the data and extracted features during the process of pulling the data from our database. These codes can be adapted to work with data in json format.

You can’t perform that action at this time.