amyjing_emoji_predict

AmyJing at SemEval 2018 Task 2: Multilingual Emoji Prediction

requirements

python 3

keras >= 2.0.6

tensorflow >= 1.2.1

usage

You need to download the pretrained embeddings and put it under [data/word2vec] dir first.

$ python main.py

model files:

• nn_model.py: super class for each model, common experiment settings are here, you can overwrite param settings within each model's class in init.
• cls_bislstm.py: classification bilstm model
• cls_lstm.py: classification 2-layered lstm model
• vs_bilstm.py: vector similarity bilstm model
• vs_lstm.py: vector similarity 2-layered lstm model

english preptrained embedding:

• https://github.com/fvancesco/acmmm2016

datasets:

• train set: 466,233
• valid set: 50,000
• test set: 50,000