asappresearch / dialog-intent-induction

Code and data for paper "Dialog Intent Induction with Deep Multi-View Clustering", Hugh Perkins and Yi Yang, 2019, EMNLP 2019

Code and data for paper "Dialog Intent Induction with Deep Multi-View Clustering", Hugh Perkins and Yi Yang, 2019, to appear in EMNLP 2019.

Data is available in the sub-directory data, with a specific LICENSE file.

Dialog Intent Induction

Dialog intent induction aims at automatically discovering dialog intents from human-human conversations. The problem is largely overlooked in the prior academic works, which created a hugh gap between academics and industry.

In particular, academic dialog datasets such as ATIS and MultiWoZ assume dialog intents are given; they also focus on simple dialog intents like BookRestaurant or BookHotel. However, many complex dialog intents emerge in industrial settings that are hard to predefine; the dialog intents are also undergoing dynamic changes.

Deep multi-view clustering

In this work, we propose to tackle this problem using multi-view clustering. Consider the following example dialogs:

The user query utterances (query view) are lexically and syntactically dissimilar. However, the solution trajectories (content view) are similar.

Alternating-view k-means (AV-KMeans)

We propose a novel method for joint representation learning and multi-view cluster: alternating-view k-means (AV-KMeans).

We perform clustering on view 1 and project the assignment to view 2 for classification. The encoders are fixed for clustering and updated for classification.

Experiments

We construct a new dataset to evaluate this new intent induction task: Twitter Airlines Customer Support (TwACS).

We compare three competitive clustering methods: k-means, Multi-View Spectral Clustering (MVSC), and AV-Kmeans (ours). We experiment with three approaches to parameter initialization: PCA for k-means and MVSC; autoencoders; and quick thoughts.

The F1 scores are presented below:

Algo	PCA/None	autoencoders	quick thoughts
k-means	28.2	29.5	42.1
MVSC	27.8	31.3	40
AV-Kmeans (ours)	35.4	38.9	46.2

Usage

Pre-requisites

• decompress the .bz2 files in datafolder
• download http://nlp.stanford.edu/data/glove.840B.300d.zip, and unzip glove.840B.300d.txt into data folder

To run AV-Kmeans

• run one of:

# no pre-training
cmdlines/airlines_mvc.sh

# ae pre-training
cmdlines/airlines_ae.sh
cmdlines/airlines_ae_mvc.sh

# qt pre-training
cmdlines/airlines_qt.sh
cmdlines/airlines_qt_mvc.sh

• to train on askubuntu, replace airlines with ubuntu in the above command-lines

To run k-means baseline

• for qt pretraining run:

PYTHONPATH=. python train_qt.py --data-path  data/airlines_processed.csv --pre-epoch 10 --view1-col first_utterance --view2-col context --scenarios view1

• to train on askubuntu, replace airlines with askubuntu in the above command-line, and remove the --*-col command-line options