GeLIC - German Library Indexing Collection

Gelic is Proto-Germanic for "like, alike, similar" (see Wiktionary).

This repository contains the following items:

• components contains the documents, relevance assessments and topics
  • assessments.txt contains relevance assessments
  • corpus.zip contains the title data of the documents
  • topics.xml contains topics used for evaluation
• src contains a bunch of conversion and analysis scripts
• scripts contains scripts to automizie different tasks that emerged in the course of the project
  • evalsummary.py is a script to search a local solr instance with the titles of the topics.xml as queries. fieldnames.json specifies in which fields solr searches. Variations could be: 1. only in "subject_auto_txt_de" 2. only in "subject_gnd_txt_de" 3. "subject_auto_txt_de" and "subject_gnd_txt_de". The results of each fieldnames-variation of queries are then evaluated by trec_eval and summarized. Thus the script is most suitable if you want a quick comparision between variations of fields that are searched in. These summary-evaluations are then saved as a csv-file.
  • evalpertopic.py automizes similar processes as evalsummary.py and only the endresult differs. Instead of writing a summary in a single csv-file, the script creates as many csv-files as there are variations in fieldnames.json. In these csv-files are the trec_eval result for each query of the variation e.g. a csv-file for the variation "subject_auto_txt_de" in which the recall, precision and f measure of topics like "Kritische Theorie" can be found.
  • erschliessung.py results in a list of all topics with counts regarding the occurences of relevant documents divided in machine content indexing, conventional content indexing etc.
  • synonyms.py extracts the gnd-synonyms and their preferred term from the jsonld-file authorities-sachbegriff_lds_20190613.jsonld and transforms the terms in a solr-readable synonyms.txt-file. Warning! Before using, follow the instructions of src/README.md Plus: if you want a .csv-list, there is also a commented out option in the script.

Authors

This repository is joined work of the following people:

• Philipp Schaer (phschaer, project lead)
• Klaus Lepsky (klepsky, project lead)
• Ina Böckmann (iboeckma)
• Sebastian Pommerencke (SebastianPommerencke)
• Sven Gaida (SvenGaida)
• Felix van Tellingen (fvantellingen)

Alumni

• Johanna Munkelt (FH Dortmund)