Create deduplication data for hand-annotating training data for a simple classifier

[janehmueller]

• generate every (string) similarity measure score for the names of two subjects
• threshold the mean of every score (to reduce data)
• export duplicate candidates to a file in the following way:
  • for every new subject there should be a list of subjects in the knowledge base that can be linked
  • the grouped subjects should be ranked by the mean of the scores in descending order
  • every subject should contain all the subject data, the similarity measure scores and the mean of the scores
  • export format should be json
• a python script should then use the exported data and enable manually annotating it