All supporting documents and scripts for the Full Ignorance Corpus focusing on prenatal nutrition. The full corpus can be found here: https://github.com/UCDenver-ccp/Ignorance-Question-Corpus
-
Corpus_Construction: The python and bash scripts used to create the corpus including
a. automatic_ontology_insertion: scripts for updating the ontology (taxonomy of ignorance) and the annotations after each round of annotation incorporating new lexical cues into both.
b. IAA_calculations: scripts for calculating the inter-annotator agreement (IAA) while also creating .xml files of the combined annotations. IAA is calculated as F1 score.
-
Preprocess_Corpus: Python scripts and output folders for use in classification algorithms - PMCID sentence files and Tokenzied file
a. run_preprcess_docs.sh: tokenizes all annotation files into sentences and BIO(-) tags by ignorance taxonomy category (preprocess_docs.py)
b. run_combine_all_tokenized_files_by_pmcid.sh: combines all tokenized files into a binary ignorance or not and all combined (combine_all_tokenized_files_by_pmcid.py)
c. run_split_articles_into_train_eval.sh: split all files annotation files into a train set of size 65 and a held-out evaluation set of 26 (split_all_articles_into_train_eval.py)
d. run_make_eval_and_train_lists.sh: scripts to create PMCID lists for the eval and train sets (make_eval_and_train_lists.py)
d. Output_Folders: all folders for the outputs of the above two scripts
i. PMCID_files_sentences: all sentence files per pmcid ii. Tokenized_Files: all tokenized files by ignorance taxonomy category -
Sentence_Analysis:
-
Word_Analysis: all files to create classification algorithms on the word-level identifying the lexical cues. For more details see the README.md in the folder.