Update the functioning of the Full Text Search functionality

[rtroncy]

It seems that currently, the Full Text Search functionality works as an exact string match between the search terms and the textual items, and that an OR operator is being used when multiple terms are entered. This has confused hackathon participants.

Participants wish to rather have a AND by default (when multiple search terms are used) and maybe the possibility to express a OR or a quoted string.

Another option is to rely on SentenceBERT or equivalent. GraphDB does offer embeddings based search to investigate.

[pasqLisena]

Requested features:

• AND (default) and OR
• quoted
• wildcard character *

[ehrhart]

An issue with AND/OR has been fixed by 48ecb69.

Other features already exist:

• AND/OR: Louis AND (Stevenson OR Macie) (AND can be replaced with &&, and OR with ||)
• Quoted: "Robert Louis Stevenson"
• Wildcard: Stevenso*

[rtroncy]

The default still needs to be AND. At the moment:

• terrible smell: https://explorer.odeuropa.eu/browse?type=smells&q=terrible+smell&sort= brings 32415 results
• terrible: https://explorer.odeuropa.eu/browse?type=smells&q=terrible&sort= brings 377 results!
• terribl*: https://explorer.odeuropa.eu/browse?type=smells&q=terribl*&sort= brings 414 results
• terribl: https://explorer.odeuropa.eu/browse?type=smells&q=terribl&sort= brings 0 result

[ehrhart]

This is because GraphDB's Lucene Connector implementation uses the default Lucene operator behavior which is OR.
In order to change the behavior you would have to do it programatically during the instantiation of the QueryParser, by calling setDefaultOperator(QueryParser.Operator.AND)

Unfortunately the plugin is closed-source, but we can always contact OntoText support about it.

[rtroncy]

Thanks for digging this up! I'm all for contacting Ontotext and I have good relay there is needed. @ehrhart How would you like to proceed? Email or open an issue in some of their github?

[dschwabe]

I believe that there should be (possibly a minimal) help text in the tool itself, so that users have a better understanding of at least the default interpretation, until further options are available.