lambretta

Source code and labeled dataset for Lambretta

#Steps of Running the code and generating intermediate files.

• Run candidate_query_generator.py (Running claimextractor.py would be needed if we started from raw tweets directly, without the claims extracted). But that's been already done, and training_claims.csv is a result of that. This should output candidate_queries.txt and dict_claim_query.json.
• Run fetch_results.py . Make sure to update the awk_source_path (path of the datasource) so that the awk_query function works properly. In case you are using ElasticSearch or other interfaces, you can update the function accordingly. Additionally, also specify awk_output_export_path which will be written as an intermediate output file and needed on further step.
• Run generate_semantic_features.py. This needs file written in awk_output_export_path in fetch_results.py as input. The output of running this will be results_scoring.json.
• Run export_all_features.py. This needs the file results_scoring.json from previous step. This will write export_ltr_train.txt and export_ltr_test.txt , the files that shall go into the Learning To Rank Java programs.

Dataset

The file lambretta_dataset.json contains the dictionary of claims and list of tweets discussing the claim, alongside their moderation status (0 or 1).