一组过滤器申请答案提取和选择?

时间:2014-03-09 11:16:23

标签: java filter nlp text-mining question-answering

我一直致力于开源Ephyra代码,并将YQL API作为KM的新来源。问题是Ephyra给出了非常简短的答案。即使对于像“化学是什么?”这样的开放式问题,我还会得到一两个简短的答案,即“物质的变化”。或者'有机'。但是这些问题需要有更多解释性和详细的答案。

有什么办法可以修改这个来给我带来更详细的答案吗?也许通过删除一些过滤器并添加其他过滤器。

默认情况下,这是当前的过滤器列表:

// - answer extraction filters
AnswerSelection.addFilter(new AnswerTypeFilter());
AnswerSelection.addFilter(new AnswerPatternFilter());
AnswerSelection.addFilter(new WebDocumentFetcherFilter());
AnswerSelection.addFilter(new PredicateExtractionFilter());
AnswerSelection.addFilter(new FactoidsFromPredicatesFilter());
AnswerSelection.addFilter(new TruncationFilter());
// - answer selection filters
AnswerSelection.addFilter(new StopwordFilter());
AnswerSelection.addFilter(new QuestionKeywordsFilter());
AnswerSelection.addFilter(new ScoreNormalizationFilter(NORMALIZER));
AnswerSelection.addFilter(new ScoreCombinationFilter());
AnswerSelection.addFilter(new FactoidSubsetFilter());
AnswerSelection.addFilter(new DuplicateFilter());
AnswerSelection.addFilter(new ScoreSorterFilter());

我可以使用的附加过滤器是:

AnswerSelection.addFilter(new NumberOfKeywordsFilter());        
AnswerSelection.addFilter(new ResultLengthFilter());
AnswerSelection.addFilter(new AdaptiveNumberOfKeywordsFilter());
AnswerSelection.addFilter(new DirectSpeechFilter());
AnswerSelection.addFilter(new NounPhraseFilter());
AnswerSelection.addFilter(new NuggetEvaluationFilter());
AnswerSelection.addFilter(new PreferNamedEntitiesFilter());
AnswerSelection.addFilter(new ProperNameFilter());
AnswerSelection.addFilter(new ResultLengthFilter());
AnswerSelection.addFilter(new ResultLengthSorterFilter());
AnswerSelection.addFilter(new SentenceSplitterFilter());
AnswerSelection.addFilter(new SubclauseSplitterFilter());
AnswerSelection.addFilter(new TermFilter());
AnswerSelection.addFilter(new TripletFilter());

但我不知道哪些是答案提取过滤器,哪一个是答案选择。我想我们必须遵循一些命令。

你们可以建议一些所有这些过滤器的订单吗?通过它我可以得到我想要的那种答案(解释)。我很困惑

0 个答案:

没有答案