SemEval Task 3 - SubTask A (Question-Comment Similarity) : Answer Reranking Problem
Given a question and the first ten comments in its question thread, rerank
the 10 comments according to their relevance with respect to the question.
| File | Description |
|---|---|
| train.py | train doc2vec representations |
| test.py | test doc2vec representations |
| myutils.py | utility library |
| pre.py | preprocess library for 'subtask-A' data |
| model.py | model implementations for 'subtask-A' |
| features.py | semantic and metadata features implem. |
| postag.py | helper py script for POS tagging and caching results |
| meta.py | generate & save meta features (Q/C author, Qcategory) |
| lda.py | LDA topic modeling for train-dev-test data |
| cluster.py | KMeans clustering of word vecs from doc2vec.vocab |
| config.json | json config file (system specific parameters) |
| tagger_cache.json | json containing pos tags for train-dev-test data (postag.py) |
| meta_cache.json | json containing meta data about Q & C (meta.py) |
| cluster_cache.json | json containing word to cluster-id mapping (cluster.py) |
| doc2vec.sh | shell script for batch runs of doc2vec model training |
| model.sh | shell script for batch runs of neuralnet model training |
| out/ | results folder (output) |
| papers/ | base research papers and latex documents (thesis, ...) |
| test/ | test data and results |