如何初始化stanfordNLP管道一次并多次使用而无需再次初始化？

Question

我想初次化stanfordNLP pipelince并多次使用它而不再重新初始化，以改善执行时间。

有可能吗？

我有代码：

    public static boolean isHeaderMatched(String string) {

    // creates a StanfordCoreNLP object.
    Properties props = new Properties();
    props.put("annotators", "tokenize, ssplit, pos, lemma, ner");

    RedwoodConfiguration.current().clear().apply();
    StanfordCoreNLP pipeline = new StanfordCoreNLP(props);

    Env env = TokenSequencePattern.getNewEnv();
    env.setDefaultStringMatchFlags(NodePattern.CASE_INSENSITIVE);
    env.setDefaultStringPatternFlags(Pattern.CASE_INSENSITIVE);

    Annotation document = new Annotation(string);

    // use the pipeline to annotate the document we created
    pipeline.annotate(document);
    List<CoreMap> sentences = document.get(SentencesAnnotation.class);

    CoreMapExpressionExtractor extractor = CoreMapExpressionExtractor.createExtractorFromFiles(env, "./app/utils/Summarizer/mapping/career_objective.rule", "./app/utils/Summarizer/mapping/personal_info.rule", "./app/utils/Summarizer/mapping/education.rule", "./app/utils/Summarizer/mapping/work_experience.rule", "./app/utils/Summarizer/mapping/certification.rule", "./app/utils/Summarizer/mapping/publication.rule", "./app/utils/Summarizer/mapping/award_achievement.rule", "./app/utils/Summarizer/mapping/hobbies_interest.rule", "./app/utils/Summarizer/mapping/lang_known.rule", "./app/utils/Summarizer/mapping/project_details.rule", "./app/utils/Summarizer/mapping/skill-set.rule", "./app/utils/Summarizer/mapping/misc_header.rule");

    boolean flag = false;
    for (CoreMap sentence : sentences) {
        List<MatchedExpression> matched = extractor.extractExpressions(sentence);
        //System.out.println("Probable Header is : " + matched);
        Set<String> uniqueMatchedKeyWordSet = DocumentParserUtil.removeDuplicate(matched);
        System.out.println("Matched: " + uniqueMatchedKeyWordSet + " and Size of MatchedSet: " + uniqueMatchedKeyWordSet.size());

        //checked if the more than half the no. of word in header(string) is matched
        if ((matched.size() >= uniqueMatchedKeyWordSet.size()) && !matched.isEmpty() && matched.size() >= Math.floorDiv(string.split("\\s").length, 2)) {
                //System.out.println("This is sure a header!");
            flag = true;
        } else {
            flag = false;
        }
  /*for(MatchedExpression phrase: matched){
    System.out.println("matched header type: " + phrase.getValue().get());
  }*/
    }
    return flag;
}

我想执行这部分代码只在第一次调用上面的方法时执行才能加载模型。

    // creates a StanfordCoreNLP object.
    Properties props = new Properties();
    props.put("annotators", "tokenize, ssplit, pos, lemma, ner");

    RedwoodConfiguration.current().clear().apply();
    StanfordCoreNLP pipeline = new StanfordCoreNLP(props);

    Env env = TokenSequencePattern.getNewEnv();
    env.setDefaultStringMatchFlags(NodePattern.CASE_INSENSITIVE);
    env.setDefaultStringPatternFlags(Pattern.CASE_INSENSITIVE);

提前致谢。

Answer 1

以下是您可以执行的操作的示例：

public class Example {
    private static StanfordCoreNLP pipeline;
    private static Env env;

    static {
        // creates a StanfordCoreNLP object.
        Properties props = new Properties();
        props.put("annotators", "tokenize, ssplit, pos, lemma, ner");

        RedwoodConfiguration.current().clear().apply();
        pipeline = new StanfordCoreNLP(props);

        env = TokenSequencePattern.getNewEnv();
        env.setDefaultStringMatchFlags(NodePattern.CASE_INSENSITIVE);
        env.setDefaultStringPatternFlags(Pattern.CASE_INSENSITIVE);
    }

    public static boolean isHeaderMatched(String string) {
        Annotation document = new Annotation(string);

        // use the pipeline to annotate the document we created
        pipeline.annotate(document);
        List<CoreMap> sentences = document.get(SentencesAnnotation.class);

        CoreMapExpressionExtractor extractor = CoreMapExpressionExtractor.createExtractorFromFiles(env, "./app/utils/Summarizer/mapping/career_objective.rule", "./app/utils/Summarizer/mapping/personal_info.rule", "./app/utils/Summarizer/mapping/education.rule", "./app/utils/Summarizer/mapping/work_experience.rule", "./app/utils/Summarizer/mapping/certification.rule", "./app/utils/Summarizer/mapping/publication.rule", "./app/utils/Summarizer/mapping/award_achievement.rule", "./app/utils/Summarizer/mapping/hobbies_interest.rule", "./app/utils/Summarizer/mapping/lang_known.rule", "./app/utils/Summarizer/mapping/project_details.rule", "./app/utils/Summarizer/mapping/skill-set.rule", "./app/utils/Summarizer/mapping/misc_header.rule");

        boolean flag = false;
        for (CoreMap sentence : sentences) {
            List<MatchedExpression> matched = extractor.extractExpressions(sentence);
            //System.out.println("Probable Header is : " + matched);
            Set<String> uniqueMatchedKeyWordSet = DocumentParserUtil.removeDuplicate(matched);
            System.out.println("Matched: " + uniqueMatchedKeyWordSet + " and Size of MatchedSet: " + uniqueMatchedKeyWordSet.size());

            // checked if the more than half the no. of word in header(string) is matched
            if ((matched.size() >= uniqueMatchedKeyWordSet.size()) && !matched.isEmpty() && matched.size() >= Math.floorDiv(string.split("\\s").length, 2)) {
                flag = true;
            } else {
                flag = false;
            }

        }

        return flag;
    }

}

在上面的代码中，加载类时将执行static块。如果您不希望出现此行为，则允许访问init方法，如下所示：

public class Example {
    private static StanfordCoreNLP pipeline;
    private static Env env;

    public static init() {
        // creates a StanfordCoreNLP object.
        Properties props = new Properties();
        props.put("annotators", "tokenize, ssplit, pos, lemma, ner");

        RedwoodConfiguration.current().clear().apply();
        pipeline = new StanfordCoreNLP(props);

        env = TokenSequencePattern.getNewEnv();
        env.setDefaultStringMatchFlags(NodePattern.CASE_INSENSITIVE);
        env.setDefaultStringPatternFlags(Pattern.CASE_INSENSITIVE);
    }

    public static boolean isHeaderMatched(String string) {
        // code left out for brevity
    }

}

可以使用以下方法从另一个类调用：

Example.init();
Example.isHeaderMatched("foobar");

在写这个答案的时候，我发现你的逻辑可能存在缺陷。以下代码可能无法产生您想要的行为。

boolean flag = false;
for (CoreMap sentence : sentences) {
    List<MatchedExpression> matched = extractor.extractExpressions(sentence);
    //System.out.println("Probable Header is : " + matched);
    Set<String> uniqueMatchedKeyWordSet = DocumentParserUtil.removeDuplicate(matched);
    System.out.println("Matched: " + uniqueMatchedKeyWordSet + " and Size of MatchedSet: " + uniqueMatchedKeyWordSet.size());

    // checked if the more than half the no. of word in header(string) is matched
    if ((matched.size() >= uniqueMatchedKeyWordSet.size()) && !matched.isEmpty() && matched.size() >= Math.floorDiv(string.split("\\s").length, 2)) {
        flag = true;
    } else {
        flag = false;
    }

}

您正在迭代CoreMap集合List<CoreMap>中的每个sentences。每次迭代都会将flag设置为条件的结果，这就是问题所在。布尔值flag仅反映最后一次sentence运行条件的结果。如果您需要知道每个sentence的结果，那么您应该有一个布尔列表来跟踪结果，否则删除循环并只检查最后一句（因为那是你的循环是什么无论如何）。