OPAL Named Entity Disambiguation and indexing component
AGDISTIS (Agnostic Named Entity Disambiguation) aims at delivering a framework for disambiguating a priori annotated named entities. It has been extended in the project LIMBO to integrate the search engine Elasticsearch. In the project OPAL, the geographical database LauNuts was integrated.
Links to other versions:
The underlying indexing system is Elasticsearch 6.6. The installation with Docker is described at the Elasticsearch documentation. The following commands can be used:
docker pull docker.elastic.co/elasticsearch/elasticsearch:6.6.0
docker run -p 9200:9200 -p 9300:9300 -e "discovery.type=single-node" docker.elastic.co/elasticsearch/elasticsearch:6.6.0
To import the required data, perform the following steps:
- Download and extract the LauNuts dataset
- Set the property folderWithTTLFiles in the configuration file agdistis.properties
- Run the org.aksw.agdistis.indexWriter.TripleIndexCreator
- Validate the data status on http://localhost:9200/_cat/indices?v using the Elasticsearch cat indices API
Start the webservices using org.aksw.agdistis.webapp.RunApp.
- Use the AGDISTIS webservice URLs to disambiguate places
- Disambiguation example:
curl --data-urlencode "text='The city of <entity>Paderborn</entity> has over 150,000 inhabitants.'" -d type='agdistis' http://localhost:8080/AGDISTIS
- Candidates example:
curl --data-urlencode "text='The city of <entity>Hamburg</entity> is also a federal state.'" -d type='candidates' http://localhost:8080/AGDISTIS
This work has been supported by the German Federal Ministry of Transport and Digital Infrastructure (BMVI) in the projects Open Data Portal Germany (OPAL) (funding code 19F2028A) and Linked Data Services for Mobility (LIMBO) (funding code 19F2029I).