GitHub - statham-stone/DACI: "This is the code and data release of the 2018 IJCAI paper 《Domain adapatation via Tree Kernel Based Maximus Mean Discrepancy for User Consumption Intention Identification》"

Dependency：

pytorch : Version 0.2.0 tqdm : Version 4.19.4 pyltp : Version: 0.1.9.1

Preparation:

Download the ltp model version 3.3.1: https://pan.baidu.com/share/link?shareid=1988562907&uk=2738088569#list/path=%2F 'pos.model' 'cws.model' 'parser.model' is needed and Please put the model in path '/data/ltp_data/'
Download the embedding data : https://pan.baidu.com/s/1GwX8lSjthxoveRfxB6k5Cw Please put it in the path "./data/embedding_weibo.data"

The code is runned in three steps:

Use the script file in fold "generate_tree" to pre-precess the origin data. This step can get the depenency tree presentation of the origin data which is saved in the fold "train_tree","phone_tree" etc. command: sh gen_tree.sh We've already provided the output of the script in the zip file
Run the train_source_model.sh to get the ".pkl" model file while can be used to to predict the origin domain data.
command : sh ./script/train_source_model.sh
Run the trans_train.sh to get the domain transfer answer. command : sh ./script/trans_train.sh

Name		Name	Last commit message	Last commit date
Latest commit History 4 Commits
baby_tree/prop_relation_split		baby_tree/prop_relation_split
data		data
flight_tree/prop_relation_split		flight_tree/prop_relation_split
generate_tree		generate_tree
laptop_tree/prop_relation_split		laptop_tree/prop_relation_split
movie_tree/prop_relation_split		movie_tree/prop_relation_split
script		script
train_tree/prop_relation_split		train_tree/prop_relation_split
README.md		README.md
config.py		config.py
create_tree.py		create_tree.py
create_wordvec.py		create_wordvec.py
main.py		main.py
test_pytorch_lstm.py		test_pytorch_lstm.py
title_to_tree.py		title_to_tree.py
train_model.py		train_model.py
trans_main.py		trans_main.py
treelstm_with_pytorch.py		treelstm_with_pytorch.py

statham-stone/DACI