Socialized Word Embeddings

Preparation

You need to download the dataset and some tools:

• Download Yelp dataset

• Convert following datasets from json format to csv format by using json_to_csv_converter.py:

  yelp_academic_dataset_review.json

  yelp_academic_dataset_user.json

• Download LIBLINEAR

  After downloading liblinear, you can refer to Installation to install it.

  It is suggested that you put liblinear under the directory SocializedWordEmbeddings.

• Download Stanford CoreNLP

  Only stanford-corenlp.jar is required. SocializedWordEmbeddings/preprocess/Split_NN.jar and SocializedWordEmbeddings/preprocess/Split_PPL.jar need to reference stanford-corenlp.jar.

  It is suggested that after getting stanford-corenlp.jar, you put it under the directory SocializedWordEmbeddings/resources, otherwise, you should modify the default Class-Path in Split_NN.jar and Split_PPL.jar.

Preprocessing

cd SocializedWordEmbeddings/preprocess

Modify ./run.py by specifying --input (Path to yelp dataset).

python run.py

Training

cd SocializedWordEmbeddings/train

You may modify the following arguments in ./run.py:

• --para_lambda The trade off parameter between log-likelihood and regularization term
• --para_r The constraint of L2-norm of the user vector
• --yelp_round The round number of yelp data, e.g. {8,9}

python run.py

Sentiment Classification

cd SocializedWordEmbeddings/sentiment

You may modify the following arguments in ./run.py:

• --para_lambda The trade off parameter between log-likelihood and regularization term
• --para_r The constraint of L2-norm of the user vector
• --yelp_round The round number of yelp data, e.g. {8,9}

python run.py

Perplexity

cd SocializedWordEmbeddings/perplexity

You may modify the following arguments in ./run.py:

• --para_lambda The trade off parameter between log-likelihood and regularization term
• --para_r The constraint of L2-norm of the user vector
• --yelp_round The round number of yelp data, e.g. {8,9}

python run.py

User Vectors for Attention

We thank Tao Lei as our code is developed based on his code.

You can simply re-implement our results of different settings (Table 5 in the paper) by modifying the SocializedWordEmbeddings/attention/run.sh:

[1] add user and word embeddings by specifying --user_embs and --embedding.

[2] add train/dev/test files by specifying --train, --dev, and --test respectively.

[3] three settings for our experiments could be achieved by specifying --user_atten and --user_atten_base:

setting '--user_atten 0' for 'Without attention'.

setting '--user_atten 1 --user_atten_base 1' for 'Trained attention'

setting '--user_atten 1 --user_atten_base 0' for 'Fixed user vector as attention'.

Dependencies

• Python 2.7
• Theano >= 0.7
• Numpy
• Gensim
• PrettyTable