[PyLucene]Build my own Analyzer using ShingleFilter in tokenStream function

[Wu Ke <wu...@umich.edu>]

Hi All,

I am pretty new to Lucene and Pylucene. This is a problem when I am using pylucene to write a customized analyzer, to tokenize text in to bigrams.

The code for analyzer class is:

class BiGramShingleAnalyzer(PythonAnalyzer):                                                                                                
    def __init__(self, outputUnigrams=False):                                                                                               
        PythonAnalyzer.__init__(self)                                                                                                       
        self.outputUnigrams = outputUnigrams                                                                                                
                                                                                                                                            
    def tokenStream(self, field, reader):                                                                                                   
        result = ShingleFilter(LowerCaseTokenizer(Version.LUCENE_35,reader))                                                                
        result.setOutputUnigrams(self.outputUnigrams)                                                                                       
        #print 'result is', result                                                                                                           
        return result


I used ShingleFilter on the TokenStream produced by LowerCaseTokeinizer. When I call the tokenStream function directly, it works just tine:
str = ‘divide this sentence'
bi = BiGramShingleAnalyzer(False)
sf = bi.tokenStream('f', StringReader(str))
while sf.incrementToken():
    print sf
(divide this,startOffset=0,endOffset=11,positionIncrement=1,type=shingle)
(this sentence,startOffset=7,endOffset=20,positionIncrement=1,type=shingle)

But when I tried to build a query parser using this analyzer, problem occurred:

parser = QueryParser(Version.LUCENE_35, 'f', bi)
query = parser.parse(str)

In query there is nothing.

After I add print statement in the tokenStream function, I found when I call parser.parse(str), the print statement in tokenStream actually get called 3 times (3 words in my str variable). It seems to me the parser pre-processed the str I passed to it, and call the tokenStream function on the result of the pre-processing.

Any thoughts on how should I make the analyzer work, so that when I pass it to query parser, the parser could parse a string into bigrams?

Thanks in advance!

Ke Wu