Multi-word synonyms about solrtexttagger HOT 2 CLOSED

The problem ist that the tokens are emitted in the order you write the mappings. So if you use the mapping

Domain Name Service, DNS

you get the tokens

1. domain, posInc: 1, posLen: 1, offset: [0,3]
2. dns, posInc: 0, posLen: 1, offset: [0,3]
3. name, posInc: 1, posLen: 1, offset: [0,3]
4. service, posInc: 1, posLen: 1, offset: [0,3]

So using your suggestion one would build the phrases

1. domain
2. dns name service

The SynonymFilter does simple not provide enough information to correctly build phrases. But I am not really concerned about cases like that as configuring analyzers like that would also result in unexpected query results as nicely described by Lucene's TokenStreams are actually graphs! in the last few paragraphs (also using the DNS -> Domain Name Service example).

from solrtexttagger.

Thanks for the explanation. I committed clarifications to this just now.

Also, for real-world usage outside the tests, I think using SynonymFilter with expand=false at index & query time is ideal as it won't grow the FST like expand=true will.

from solrtexttagger.