This data accompanies the paper "Feeding a Gazetteer: Leveraging word embeddings for toponym mining", presented at the 5th ACM SIGSPATIAL International Workshop on Geospatial Humanities, Nov 2 2021. https://ludovicmoncla.github.io/sigspatial-geohumanities-2021/index.html . It includes:
- The Corpus of 5 Hebrew 19th century Newspapers, based on https://github.com/omilab/historical_press.
- A "seed" set of 1000 toponyms that have dated attestation for their use in the relevant period, taken from the Kima Gazetteer https://data.geo-kima.org/
- The notebook "Finding_Locations_in_Historic_Newspapers.ipynb", written by Nitzan Gado, which was used to clean the corpus, train an embedding model, query it for given terms, and then load the seed set and extract the top most similar terms for all seed set terms.
- Sets of our results.
- Set of results of the method gained by extracting the most-similar terms from seed set of non spatial domains.
The corpus may be used to train the model using the notebook. Alternatively, the model we trained is available for immediate use in the notebook through This link . Enjoy!