Is there anything I can do to pass my Unit-Test ?
Or it is impossible ?
Thanks a lot,
Fabrice
Fabrice Robini wrote:
>
> Hi Srikant,
>
> I really thank you for your reply, it's very interesting.
> I have to say I am confused with that now...
> I do not know what I can to for passing this Unit test...
>
> I agree with you, it may be an issue of computing relevance.
>
> Fabrice
>
>
> Srikant Jakilinki-3 wrote:
>>
>> OK, got it to work. Thanks.
>>
>> By a quick scoring comparision, I got the same scores for both hits.
>> Maybe there is a loss of precision somewhere. Or when scores are equal,
>> Lucene is doing something unintended/overlooked and thus putting shorter
>> documents higher as the experiment is a special case where the TF of a
>> queried term (for both suites, the TF of x = 10%) is equal which is very
>> rarely. Or maybe the IDF factor is kicking in in some strange way
>> although it shouldnt. There are a number of varied reasons, but for the
>> naked eye, there isnt much.
>>
>> However, that said, length normalization is not a science but an art and
>> the simple scheme we have here in the FairSimilarity will not perform
>> always as expected in real world scenarios. Maybe I am missing something
>> or have forgot my basics but that is not to say your observation is
>> trivial.
>>
>> Rather, the contrary. Hope there will be more activity on this topic
>> because it is an issue of computing relevance which is the core of
>> search.
>>
>> Cheers,
>> Srikant
>>
>> Fabrice Robini wrote:
>>> Oooops sorry, bad cut/paste...
>>>
>>> Here is the right one :-)
>>>
>>> public void testFairSimilarity() throws CorruptIndexException,
>>> IOException, ParseException
>>> {
>>> Directory theDirectory = new RAMDirectory();
>>> Analyzer theAnalyzer = new StandardAnalyzer();
>>>
>>> IndexWriter theIndexWriter = new IndexWriter(theDirectory,
>>> theAnalyzer);
>>> theIndexWriter.setSimilarity(new FairSimilarity());
>>>
>>> Document doc1 = new Document();
>>> Field name1 = new Field("NAME", "SHORT_SUITE", Field.Store.YES,
>>> Field.Index.UN_TOKENIZED);
>>> Field content1 = new Field("CONTENT", "x 2 3 4 5 6 7 8 9 10",
>>> Field.Store.NO, Field.Index.TOKENIZED);
>>> doc1.add(name1);
>>> doc1.add(content1);
>>> theIndexWriter.addDocument(doc1);
>>>
>>> Document doc2 = new Document();
>>> Field name2 = new Field("NAME", "BIG_SUITE", Field.Store.YES,
>>> Field.Index.UN_TOKENIZED);
>>> Field content2 = new Field("CONTENT", "x x 3 4 5 6 7 8 9 10 11
>>> 12 13
>>> 14 15 16 17 18 19 20", Field.Store.NO, Field.Index.TOKENIZED);
>>> doc2.add(name2);
>>> doc2.add(content2);
>>> theIndexWriter.addDocument(doc2);
>>>
>>> theIndexWriter.close();
>>>
>>> Searcher searcher = new IndexSearcher(theDirectory);
>>> searcher.setSimilarity(new FairSimilarity());
>>>
>>> QueryParser queryParser = new QueryParser("CONTENT",
>>> theAnalyzer);
>>>
>>> Hits hits = searcher.search(queryParser.parse("x"));
>>>
>>> assertEquals(2, hits.length());
>>> assertEquals("BIG_SUITE", hits.doc(0).get("NAME"));
>>> assertEquals("SHORT_SUITE", hits.doc(1).get("NAME"));
>>> }
>>>
>>>
>>>
>>>
>>> Srikant Jakilinki-3 wrote:
>>>
>>>> Well, I cant seem to even get past the assertions of this code.
>>>>
>>>> The first assertion is failing in that I get 0 hits. I am using
>>>> SimpleAnalyzer since I do not have a FrenchAnalyzer.
>>>>
>>>> Any thoughts?
>>>> Srikant
>>>>
>>
>> ----------------------------------------------------------------------
>> Free pop3 email with a spam filter.
>> http://www.bluebottle.com/tag/5
>>
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: [EMAIL PROTECTED]
>> For additional commands, e-mail: [EMAIL PROTECTED]
>>
>>
>>
>
>
--
View this message in context:
http://www.nabble.com/Is-Fair-Similarity-working-with-lucene-2.2---tp15001250p15060757.html
Sent from the Lucene - Java Users mailing list archive at Nabble.com.
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
