Daan de Jong Master Thesis Methodology and Statistics for the Biomedical, Behavioral and Social Sciences Negation Scope Detection with a BiLSTM NN

---

The Thesis-Report repository contains two folders: Text and Data and code.

Text

• img folder:
  • Plaatjesmaker.ppt and Plaatjesmaker2.ppt were used to create model.png, model2.png and prediction types.png
  • model.png is outdated
  • model2.png was used in the final Report (by main.tex in the text folder)
  • prediction types.png was used in the final Report (by main.tex in the text folder)
• main.tex contains the text that generates the final report
• refs_report is a bib file that contains the references used by main.tex
• title page.tex is used by main.tex, to split up the first page from the actual content of the report
• Draft 1 Report DDJ.pdf is the main draft which was feedbacked by Ayoub Bagheri and Joost de Jong (feedback not included)
• Research Report Daan de Jong.pdf is the final Report that was handed in
• main.bbl, main.blg, main.log, main.synctex.gz, textut.log, title page.log are byproducts of main.tex and title page.tex

Data and Code

• __pycache__ is a folder that contains byproducts of the .py files
• Original is a folder that contains all data used in this study
  • freely downloadable here
  • the folder bioscope was not used
  • bioscope_abstract.csv and bioscope_full.csv are the data files that were used in this study (by Preprocessing.py)
• All .csv files here are rendered by Preprocessing.py, and are read in Model.py, except for E.csv (the word embeddings, from word2vec.py)
• CustomEmbedding.py defines a keras layer class, to be used in Model.py
• CustomLoss.py defines the loss function used in the study, used by Model.py
• CustomMetric.py defines the metrics used in the study, used by Model.py
• Data description table.py renders the information used in Table 1 of the Report
• Model.py defines, compiles, trains, tests and evaluates the model
• Preprocessing.py bridges the gap between the raw data .csv files and the train and test data suited for the model
• word2vec.py is an implementation of the word2vec algorithm and renders E.csv, the word embeddings