java - 使用斯坦福 Tregex 提取子树

Question

我创建了一个使用 Tregex 提取子树的类。我使用了“TregexPattern.java”中的一些代码片段，因为我不想让程序使用控制台命令。

一般来说，有一个句子的树，我想提取某些子树(没有用户交互)。

到目前为止我所做的如下:

package edu.stanford.nlp.trees.tregex;
import edu.stanford.nlp.ling.StringLabelFactory;
import edu.stanford.nlp.trees.*;
import java.io.*;
import java.util.*;
public abstract class Test {
    abstract TregexMatcher matcher(Tree root, Tree tree, Map<String, Tree> namesToNodes, VariableStrings variableStrings);
    public TregexMatcher matcher(Tree t) {
        return matcher(t, t, new HashMap<String, Tree>(), new VariableStrings());
    }
    public static void main(String[] args) throws ParseException, IOException {
        String encoding = "UTF-8";
        TregexPattern p = TregexPattern.compile("NP < NN & <<DT"); //"/^MWV/" or "NP < (NP=np < NNS)"
        TreeReader r = new PennTreeReader(new StringReader("(VP (VP (VBZ Try) (NP (NP (DT this) (NN wine)) (CC and) (NP (DT these) (NNS snails)))) (PUNCT .))"), new LabeledScoredTreeFactory(new StringLabelFactory()));
        Tree t = r.readTree();
        treebank = new MemoryTreebank();
        treebank.add(t);
        TRegexTreeVisitor vis = new TRegexTreeVisitor(p, encoding);
        **treebank.apply(vis);  //line 26**
        if (TRegexTreeVisitor.printMatches) {
            System.out.println("There were " + vis.numMatches() + " matches in total.");
        }
    }
    private static Treebank treebank; // used by main method, must be accessible
    static class TRegexTreeVisitor implements TreeVisitor {
        private static boolean printNumMatchesToStdOut = false;
        static boolean printNonMatchingTrees = false;
        static boolean printSubtreeCode = false;
        static boolean printTree = false;
        static boolean printWholeTree = false;
        static boolean printMatches = true;
        static boolean printFilename = false;
        static boolean oneMatchPerRootNode = false;
        static boolean reportTreeNumbers = false;
        static TreePrint tp;
        PrintWriter pw;
        int treeNumber = 0;
        TregexPattern p;
        //String[] handles;
        int numMatches;
        TRegexTreeVisitor(TregexPattern p, String encoding) {
            this.p = p;
            //this.handles = handles;
            try {
                pw = new PrintWriter(new OutputStreamWriter(System.out, encoding), true);
            } catch (UnsupportedEncodingException e) {
                System.err.println("Error -- encoding " + encoding + " is unsupported.  Using ASCII print writer instead.");
                pw = new PrintWriter(System.out, true);
            }
            // tp.setPrintWriter(pw);
        }
        public void visitTree(Tree t) {
            treeNumber++;
            if (printTree) {
                pw.print(treeNumber + ":");
                pw.println("Next tree read:");
                tp.printTree(t, pw);
            }
            TregexMatcher match = p.matcher(t);
            if (printNonMatchingTrees) {
                if (match.find()) {
                    numMatches++;
                } else {
                    tp.printTree(t, pw);
                }
                return;
            }
            Tree lastMatchingRootNode = null;
            while (match.find()) {
                if (oneMatchPerRootNode) {
                    if (lastMatchingRootNode == match.getMatch()) {
                        continue;
                    } else {
                        lastMatchingRootNode = match.getMatch();
                    }
                }
                numMatches++;
                if (printFilename && treebank instanceof DiskTreebank) {
                    DiskTreebank dtb = (DiskTreebank) treebank;
                    pw.print("# ");
                    pw.println(dtb.getCurrentFile());
                }
                if (printSubtreeCode) {
                    pw.println(treeNumber + ":" + match.getMatch().nodeNumber(t));
                }
                if (printMatches) {
                    if (reportTreeNumbers) {
                        pw.print(treeNumber + ": ");
                    }
                    if (printTree) {
                        pw.println("Found a full match:");
                    }
                    if (printWholeTree) {
                        tp.printTree(t, pw);
                    } else {
                        **tp.printTree(match.getMatch(), pw);  //line 108**
                    }
                    // pw.println();  // TreePrint already puts a blank line in
                } // end if (printMatches)
            } // end while match.find()
        } // end visitTree
        public int numMatches() {
            return numMatches;
        }
    } // end class TRegexTreeVisitor
}

但它给出以下错误:

Exception in thread "main" java.lang.NullPointerException
        at edu.stanford.nlp.trees.tregex.Test$TRegexTreeVisitor.visitTree(Test.java:108)
        at edu.stanford.nlp.trees.MemoryTreebank.apply(MemoryTreebank.java:376)
        at edu.stanford.nlp.trees.tregex.Test.main(Test.java:26)
Java Result: 1

有什么修改或想法吗？

Answer 1

NullPointerException 通常是软件错误的指示器。

我过去也有同样的任务。使用依存解析器对句子进行解析。我决定将生成的解析树放入 XML(DOM) 中并对其执行 XPath 查询。

为了提高性能，您不需要将 xml 放入 String 中，只需将所有 XML 结构作为 DOM 保留在内存中(例如 http://www.ibm.com/developerworks/xml/library/x-domjava/ )。

使用 XPath 查询树状数据结构给我带来了以下好处:

1. 轻松加载/保存/传输句子解析结果。
2. XPath 的强大语法/功能。
3. 很多人都知道 XPath(每个人都可以自定义您的查询)。
4. XML 和 XPath 是跨平台的。
5. 大量稳定的 XPath 和 XML/DOM 库实现。
6. 能够使用 XSLT。
7. 与现有的基于 XML 的管道集成XSLT+XPath -> XSD -> 执行操作 (例如，用户已在自由空间内的某个位置指定了他们的电子邮件地址和操作)文字投诉)。