Expert finding for Community Question Answering
Code based on theano tutorial
Usage
Run data_engine.py to generate data, then fed generated pickle file to model.py and train model.
Input
- sample_i2q: two columns, (question_id, normalized question text)
- sample_i2t: two columns, (user_id, user_twitter_id)
- sample_twitter_links: each row represents a twitter user and its following users, sperated by a colon, every id in the row is a twitter id
- sample_q2u2s: three columns, (question_id, user_id, rating)
Output
*_graph.*: deepwalk edgelist input, refer to deepwalk code for description*_paths.*: deepwalk path samples output, for debug purpose*_expf.pkl: pickle file oftrain,train_for_test, valid,testdatasets.trainis list of triplets, each element of the triplet can be represented as (question, user_pos, user_neg).train_for_test,valid,testdatasets are for testing purpose, each dataset is organized as (questions list, users list, expert users list). Each element with same index of these three lists corresponds to a test sample (question, users answered the questions, expert users).questionis represented as list of word indices.
Read the Code for detailed representations.