uima-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Michael Tanenblatt <sloth...@park-slope.net>
Subject Re: ConceptMApper
Date Wed, 20 Mar 2013 15:00:42 GMT
Any tokenizer should work--I only suggested using the whitespace tokenizer as a way to see
if the tokenizer you are using may be the problem in this case. As to the DictionaryAnnotator,
I really can't say much, since I haven't used it. 


On Mar 20, 2013, at 10:56 AM, Andreas Niekler <aniekler@informatik.uni-leipzig.de> wrote:

> I will investigate if this is the case. I will try to only use a
> whitespace Tokenizer.
> 
> Is there any information if the DictionaryAnnotator would help me more
> then? And if so is it as fast as the conceptmapper?
> 
> Thanks for clarification
> 
> Andreas
> 
> Am 20.03.2013 14:16, schrieb Michael Tanenblatt:
>> One thing that looks odd to me is that each entry is the "1. " prefix. Perhaps that
is causing a problem, if the tokenizer is putting a sentence break at that point? Just a guess.

>> 
>> 
>> On Mar 20, 2013, at 8:38 AM, Andreas Niekler <aniekler@informatik.uni-leipzig.de>
wrote:
>> 
>>> This is how my dict looks like:
>>> 
>>> <?xml version="1.0" encoding="UTF-8" ?>
>>> <synonym>
>>> <token canonical="mwu" SemClass="mwu">
>>> <variant base="1. FC Straubing"/>
>>> <variant base="1. FC Styrum"/>
>>> <variant base="1. FC Tatran Presov"/>
>>> <variant base="1. FC Tatran Prešov"/>
>>> <variant base="1. FC Trogen"/>
>>> <variant base="1. FC Union"/>
>>> <variant base="1. FC Union Berlin"/>
>>> <variant base="1. FC Union Solingen"/>
>>> <variant base="1. FC Viersen"/>
>>> <variant base="1. FC Viersen 05"/>
>>> <variant base="1. FC Vöcklabruck"/>
>>> <variant base="1. FC Weißenfels"/>
>>> <variant base="1. FC Wernigerode"/>
>>> <variant base="1. FC Wilmersdorf"/>
>>> <variant base="1. FC Windeck"/>
>>> <variant base="1. FC Wolfsburg"/>
>>> <variant base="1. FC Wunstorf"/>
>>> <variant base="1. FC Zeitz"/>
>>> <variant base="1. FFC"/>
>>> <variant base="1. FFC 08 Niederkirchen"/>
>>> <variant base="1. FFC Fortuna Dresden-Rähnitz"/>
>>> <variant base="1. FFC Frankfurt"/>
>>> <variant base="1. FFC Montabaur"/>
>>> </token>
>>> </synonym>
>>> 
>>> Am 20.03.2013 12:26, schrieb Michael Tanenblatt:
>>>> I have never seen this issue--under no circumstances should anything less
than the full dictionary entry be matched. The only things I can think of are either errors
in the dictionary, though that's unlikely, or issues with the tokenizer. Or a bug… My guess
is that the dictionary entry, "FC Barcelona" is being tokenized such that only "FC" is annotated,
therefore that is the only part that needs to match. You can test if it is a tokenization
issue by using the sample whitespace tokenizer that comes with ConceptMapper just to test
and see what results you get.
>>>> 
>>>> 
>>>> On Mar 20, 2013, at 7:09 AM, Andreas Niekler <aniekler@informatik.uni-leipzig.de>
wrote:
>>>> 
>>>>> Hello,
>>>>> 
>>>>> i try to use the ConceptMapper to annotate Multi Word Units in german.
I
>>>>> face the problem that all the tokens within the dictionary are matched
>>>>> somehow like.
>>>>> 
>>>>> In the dict -> FC Barcelona
>>>>> 
>>>>> Annotated in a Text "The FC scored today" FC is annotated as DictEntry
>>>>> 
>>>>> Why does conceptMapper annotate this. Here are my Parameters
>>>>> 
>>>>> AnalysisEngineDescription mapper =
>>>>> AnalysisEngineFactory.createPrimitiveDescription(
>>>>> 				ConceptMapper.class,
>>>>> 				ts,
>>>>> 				ConceptMapper.PARAM_ANNOTATION_NAME,
>>>>> "org.apache.uima.conceptMapper.DictTerm",
>>>>> 	    		ConceptMapper.PARAM_ENCLOSINGSPAN, "enclosingSpan",
>>>>> 	    		ConceptMapper.PARAM_TOKENANNOTATION, "opennlp.uima.Token",
>>>>> 	    		ConceptMapper.PARAM_ATTRIBUTE_LIST, new String[] {"canonical"},
>>>>> 	    		ConceptMapper.PARAM_FEATURE_LIST, new String[] {"DictCanon"},
    		
>>>>> 	    		ConceptMapper.PARAM_MATCHEDFEATURE, "matchedText",
>>>>> 	    		ConceptMapper.PARAM_TOKENIZERDESCRIPTOR, "TokenizerDE.xml",
>>>>> 	    		//ConceptMapper.PARAM_DATA_BLOCK_FS, "uima.tcas.DocumentAnnotation",
>>>>> 	    		ConceptMapper.PARAM_DATA_BLOCK_FS, "opennlp.uima.Sentence",
>>>>> 	    		ConceptMapper.PARAM_SEARCHSTRATEGY, "ContiguousMatch",
>>>>> 	    		ConceptMapper.PARAM_MATCHEDTOKENSFEATURENAME, "matchedTokens",
>>>>> 	    		TokenNormalizer.PARAM_CASE_MATCH, "ignoreall");
>>>>> 
>>>>> Thank you
>>>>> 
>>>>> Andreas
>>>> 
>>>> 
>>> 
>>> -- 
>>> Andreas Niekler, Dipl. Ing. (FH)
>>> NLP Group | Department of Computer Science
>>> University of Leipzig
>>> Johannisgasse 26 | 04103 Leipzig
>>> 
>>> mail: aniekler@informatik.uni-leipzig.deg.de
>> 
>> 
> 
> -- 
> Andreas Niekler, Dipl. Ing. (FH)
> NLP Group | Department of Computer Science
> University of Leipzig
> Johannisgasse 26 | 04103 Leipzig
> 
> mail: aniekler@informatik.uni-leipzig.deg.de


Mime
View raw message