Currently we store the predictions as json. We want to be able to extract triples from this and store in a database (a SQL database to start with) to run semantic queries.
One way to get relations from the predictions is get sentence level and paragraph level co-occurences and use that as the relations data.
Currently we store the predictions as json. We want to be able to extract triples from this and store in a database (a SQL database to start with) to run semantic queries.
One way to get relations from the predictions is get sentence level and paragraph level co-occurences and use that as the relations data.