时间戳

首页 > 解决方案 > 斯坦福 CoreNLP 在不应该的情况下将 2 个单独的实体识别为相同

java - 斯坦福 CoreNLP 在不应该的情况下将 2 个单独的实体识别为相同

问题描述

继承人的代码：

    public static void main(String[] args) {
        String text = "Loryn lives across the street from me. " 
                + "She is 19 years old. " 
                + "Sydney goes to my school. "
                + "She graduated last year. ";
        Properties props = new Properties();
        props.setProperty("annotators", "tokenize,ssplit,pos,lemma,ner,parse,depparse,coref,kbp,quote");
        props.setProperty("coref.algorithm", "neural");
        StanfordCoreNLP pipeline = new StanfordCoreNLP(props);
        CoreDocument document = new CoreDocument(text);
        pipeline.annotate(document);
        for (Entry<Integer, CorefChain> e : document.corefChains().entrySet()) {
            System.out.println(e.getValue() + " - " + e.getKey());
            for (CorefMention s : e.getValue().getMentionsInTextualOrder()) {
                System.out.println(" - " + s);
            }
        }

这是代码输出的内容：

CHAIN7-["me" in sentence 1, "my" in sentence 3] - 7
 - "me" in sentence 1
 - "my" in sentence 3
CHAIN8-["Loryn" in sentence 1, "She" in sentence 2, "She" in sentence 4] - 8
 - "Loryn" in sentence 1
 - "She" in sentence 2
 - "She" in sentence 4

为什么She从第 4 句中提到 Loryn？我怎样才能让它参考悉尼

所需的输出应类似于以下内容：

CHAIN7-["me" in sentence 1, "my" in sentence 3] - 7
 - "me" in sentence 1
 - "my" in sentence 3
CHAIN8-["Loryn" in sentence 1, "She" in sentence 2] - 8
 - "Loryn" in sentence 1
 - "She" in sentence 2
CHAIN9-["Sydney" in sentence 3, "She" in sentence 4] - 8
 - "Sydney" in sentence 3
 - "She" in sentence 4

标签： javastanford-nlp

解决方案

“Sydney”被标记为 CITY，所以这里有一个 NER 错误。

话虽如此，如果您只是将名称更改为“Jane”或类似的名称，它看起来会失败。

不幸的是，共指还没有真正解决，即使是最先进的系统也会犯很多错误。这是一个有趣的问题案例，我们将尝试使用它来为模型添加更多的训练数据！

推荐阅读