machine-learning - Cleartk:初始化 [class org.cleartk.classifier.jar.DefaultSequenceDataWriterFactory] ​​时出错，需要字段 'dataWriterClassName'

Question

我的主程序看起来像这样:

    public static void main(String args[]) throws UIMAException, IOException{
    //TypeSystemDescription tsd = TypeSystemDescriptionFactory.createTypeSystemDescription(Question.class);

    AggregateBuilder builder = new AggregateBuilder();
    //builder.add(SentenceAnnotator.getDescription());
    builder.add(AnalysisEngineFactory.createPrimitiveDescription(POSAnnotator1.class,
            ExampleComponents.TYPE_SYSTEM_DESCRIPTION,
            GenericJarClassifierFactory.PARAM_CLASSIFIER_JAR_PATH, outputDirectory + File.separator + "model.jar",
            CleartkAnnotator.PARAM_IS_TRAINING, true,
            DefaultDataWriterFactory.PARAM_DATA_WRITER_CLASS_NAME, InstanceDataWriter.class.getName(),
            DirectoryDataWriterFactory.PARAM_OUTPUT_DIRECTORY, new File(outputDirectory)));

    JCas jcas = JCasFactory.createJCas();
    jcas.setDocumentText(testData);

    SimplePipeline.runPipeline(jcas, builder.createAggregateDescription());

}

谁能解释一下这个错误吗？

Caused by: java.lang.IllegalArgumentException: Errors initializing [class org.cleartk.classifier.jar.DefaultSequenceDataWriterFactory] Field 'dataWriterClassName' is required

我尝试用其他数据编写器替换 InstanceDataWriter，但它们不起作用。

Answer 1

POS Annotator 使用序列数据写入器来编写示例。您可以从 DefaultDataWriterFactory.PARAM_DATA_WRITER_CLASS_NAME 设置 DataWriter 类名参数。对于序列数据写入器，它应该是 DefaultSequenceDataWriterFactory.PARAM_DATA_WRITER_CLASS_NAME。序列意味着 cas 中有超过 1 个标签(许多 POS 标签)，而“普通”分类器只为整个文档提供 1 个标签。

http://cleartk.googlecode.com/svn-history/r4142/tags/cleartk-release-1.2.0/apidocs/org/cleartk/classifier/jar/DefaultSequenceDataWriterFactory.html