Hi,

I am using Hibernate Search for performing search. I have a JPA/Hibernate entity say 'Inventory' which gets indexed whenever the entity changes (I use hibernate annotations for indexing). My requirement is that, for some of the attributes of Inventory say 'vehicle color', I want to implement Synonym as well as Metaphone(Soundex). I have corresponding analyzers which works perfectly fine if I use one analyzer at a time (I understand that only one analyzeris allowed for an entity attribute). My question is how can I enable both sysnonym and metaphone search for this field.

Any help in this regard would be highly appreciated.

Thanks,
Deepa[/quote]

Hi,

really there are several options here.

1. You just index your attribute twice using the @Fields annotations containing two @Field annotations each with its own analyzer (@Field has a 'analyzer' attribute).

2. You write your own custom analyzer. If you look at the implementation of a Analyzer you will see that all it does is to define an initial tokenizer (often the StandardTokenizer), followed by a chain of TokenFilters doing additional operations like adding synonyms to the token stream and/or adding soundex tokens. All your custom analyzer would have to do is create the right chain of filters. Probably you want to first insert synonyms and then even create soundex tokens for these synonyms.

3. You can use the @AnalyzerDef framework of HibernateSearch and define your tokenizers and filters declarative. Have a look at the online documentation of Hibernate Search. @AnalyzerDef depends on the Solr analyzer framework which uses a factory pattern to instantiate the right analyzer. In fact there is SynonymFilter and a PhoneticFilter available in the Solr framework. Maybe you can use them? The Hibernate Search codebase contains some tests using these two filters in case you are looking for some examples.

If I understand your usecase correctly I would recommend option 2 or 3.
I hope this helped.

--Hardy

Thanks Hardy. I tried writing a custom analyzer with a chain of desired filters. This works well. Thanks for the pointer.

Regards,
Deepa