从解析器的输出找到依赖关系的流 [英] Find the flow of dependency from the ouput of a parser
本文介绍了从解析器的输出找到依赖关系的流的处理方法，对大家解决问题具有一定的参考价值，需要的朋友们下面随着小编来一起学习吧！
问题描述
我的输入是
String idlist = [26，28]
String argslist = [[26，25]，[26，24]，[26，29]，[26，28] ，[28,27]，[35,26]，[29，30]]
String kindlist = [aux，dobj，pobj，prep，advmod，aux， adj]
我的预期输出是两个列表
Depflowlist是第一个，另一个是kindflowlist。
对于26本身它有不同的流程。
为26预期的depflowlist是[26，25]和其相应的kindflowlist [aux]
然后[26,24]是另一个depflowlist为26和其对应kindflowlist [dobj]
然后[[26，29]，[29，30]]是另一个用于26的depflowlist及其对应的kindflowlist [pobj，adj]。 >
这也应该为28。
我写了代码如下：
`public static void dependencyparse（String a，ArrayList< String> argslist，ArrayList< String> kindlist，Document doc）{
ArrayList< String> selectarg_list = new ArrayList< String>（）; （int j = 0; j< argslist.size（）; j ++）
{
int index = argslist.get（j）.indexOf（a）;
if（index == 1）
{
String next = argslist.get（j）.substring（5，7）;
selectarg_list.add（next）;
}
}
System.out.println（a +---------->+ selectarg_list）;
if（！selectarg_list.isEmpty（））
{
// System.out.println（notempty）;
hm.put（a，selectarg_list）;
}
//System.out.println(\"Hashmap\"+hm）;
if（selectarg_list.size（）> 0）{
dependencyparse（selectarg_list.get（0），argslist，kindlist，doc）;
}
if（selectarg_list.size（）== 0）{
String prevKey = delete（a，selectarg_list，argslist，kindlist，doc）;
System.out.println（prevKey）;
if（！hm.isEmpty（））{
if（hm.keySet（）。contains（prevKey））{
ArrayList< String> list = hm.get（prevKey）;
dependencyparse（list.get（0），argslist，kindlist，doc）;
}
else {
if（！hm.isEmpty（））{
try {
dependencyparse（hm.values（）。iterator（） .get（0），argslist，kindlist，doc）;
} catch（异常e）{}
}
}
迭代器< String> itr3 = hm.keySet（）。iterator（）;
while（itr3.hasNext（））{
String string =（String）itr3.next（）;
System.out.println（string +------------>+ hm.get（string））;
}
}
}
}
私人静态字符串delete（String a，ArrayList< String> selectarg_list，ArrayList< String> argslist，ArrayList< String> kindlist，Document doc）{
AnnotationSet outputAnnotations = doc.getNamedAnnotationSets（）get（Output markups
String retKey =;
集合< ArrayList< String>> values = hm.values（）;
String key =;
//System.out.println(\"Valuess\"+values）;
//System.out.println(\"find\"+a）;
boolean flag = false;
迭代器< ArrayList< String>> Valueitr = values.iterator（）;
while（Valueitr.hasNext（））
{
//System.out.println(\"Valuess ---->+ Valueitr.next（））;
ArrayList< String> contentnext = Valueitr.next（）;
if（contentnext.contains（a））
{
flag = true;
//System.out.println(\"Valuess find ---->+ contentnext）;
for（Entry< String，ArrayList< String>>条目：hm.entrySet（））
{
if（contentnext.equals（entry.getValue（）））
{
key = entry.getKey（）;
// System.out.println（Key find ---->+ key）;
//使用键
进行某些操作String arg =[+ key +，+ a +];
// System.out.println（arg find ---->+ arg）;
depFlowList.addFirst（arg）;
// System.out.println（entry find ---->+ entry.getValue（））;
try {
if（！hm.keySet（）。contains（a））{
entry.getValue（）。remove（a）;
// hm.put（entry.getKey（），value）
// System.out.println（entry.getKey（）+------------- ---+ hm.get（entry.getKey（））+\\\
entry delete ---->+ entry.getValue（））;
}} catch（Exception e）{}
}
}
}
$
if（！flag）{
if（！depFlowList.isEmpty（））{
System.out.println（Dep Flow List ----- >中+ depFlowList）; （int x = 0; x< depFlowList.size（）; x ++）{
String item = depFlowList.get（x）;
int item_index = argslist.indexOf（item）;
String item_kind = kindlist.get（item_index）;
System.out.println（add kind of -----+ depFlowList.get（x））;
kindFlowList.add（item_kind）;
}
System.out.println（Kind Flow List ----->+ kindFlowList）;
file_write（/ home / cognicor / vagateplugin / testing / dependency.txt，kindFlowList.toString（））;
}
if（！kindFlowList.isEmpty（））{
FeatureMap depfeature = new SimpleFeatureMapImpl（）;
depfeature.put（depflow，kindFlowList）;
try {
outputAnnotations.add（new Long（0），new Long（1），DependencyFlow，depfeature）;
} catch（InvalidOffsetException e）{
// TODO自动生成的catch块
e.printStackTrace（）;
}
}
if（depFlowList.size（）== 0）{
return;
}
retKey = depFlowList.getLast（）。substring（1,3）;
depFlowList.clear（）;
kindFlowList.clear（）;
System.out.println（retKey）;
return retKey;
}
//System.out.println(\"++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ ++++++++++++++++++）;
ArrayList< String> findval = hm.get（key）;
if（findval.isEmpty（））{
hm.remove（key）;
//System.out.println(\" key removed）;
}
//if(!hm.isEmpty（））{
retKey = delete（key，selectarg_list，argslist，kindlist，doc）;
return retKey;
//}
}
public static void main（String [] args）{
// TODO自动生成的方法stub
ArrayList< String> testlist = file_read（/ home / cognicor / vagateplugin / testing / test_set1.txt）;
//System.out.println(testlist）;
ArrayList< String> query_class = new ArrayList< String>（）;
ArrayList< String> property_class = new ArrayList< String>（）;
迭代器< String> testinput = testlist.iterator（）;
字符串条目;
System.out.println（testlist ---->+ testlist）; （int i = 0; i< testlist.size（）; i ++）
{
try {
ArrayList< Integer> idlist = new ArrayList< Integer>（）;
ArrayList< String> argslist = new ArrayList< String>（）;
ArrayList< String> kindlist = new ArrayList< String>（）;
String query =< currDial>+ testlist.get（i）+< / currDial>;
System.out.println（query）;
文档doc = pipline.get_Annoted_Doument（query）;
/ *代码，我得到的输入像idlist，arglist和kindlist * /
AnnotationSet annotset = doc.getNamedAnnotationSets（）。get（Output markups）。get（TokenID）;
迭代器<注释> itr = annotset.iterator（）;
while（itr.hasNext（））
{
注释注释= itr.next（）;
int id =（int）annot.getFeatures（）。get（id）;
idlist.add（id）;
}
AnnotationSet注释集1 = doc.getNamedAnnotationSets（）。get（输出标记）。get（Dependency）;
迭代器<注释> itr1 = annotset1.iterator（）;
while（itr1.hasNext（））
{
注释注释= itr1.next（）;
String arg = annot.getFeatures（）。get（args）。toString（）;
String kind = annot.getFeatures（）。get（kind）。toString（）;
argslist.add（arg）;
kindlist.add（kind）;
}
System.out.println（args+ argslist）; （int j = 0; j< idlist.size（）; j ++）
System.out.println（--------- ----------------------项目： + idlist.get（J）的ToString（）+ ------------ ------------------------------）;
dependencyparse（idlist.get（j）.toString（），argslist，kindlist，doc）;
if（j == idlist.size（））
{
dependencyparse（null，argslist，kindlist，doc）;
}
}
} catch（Exception e）{e.printStackTrace（）; }
}
}
}`
现在我有一个没有结束的循环。我正在使用门的输出。
解决方案
这是我的代码。它在Java 7 + 番石榴。请注意，我还打电话给Stanford Parser，并直接从注释中读取输入（不需要String解析）。
解析的句子来自 nlp.stanford.edu 。
import gate。*;
import gate.creole.SerialAnalyserController;
import java.io.File;
import java.util。*;
import com.google.common.collect.ArrayListMultimap;
public class DepflowFinder {
public static void main（String [] args）throws异常{
Gate.setGateHome（new File（C：\\Program Files\\ \\\GATE_Developer_8.1\" ））;
Gate.init（）;
regiterGatePlugin（ANNIE）; regiterGatePlugin（Stanford_CoreNLP/ *Parser_Stanford* /）;
SerialAnalyserController pipeline =（SerialAnalyserController）Factory.createResource（gate.creole.SerialAnalyserController）;
pipeline.add（（ProcessingResource）Factory.createResource（gate.creole.tokeniser.DefaultTokeniser））;
pipeline.add（（ProcessingResource）Factory.createResource（gate.creole.splitter.SentenceSplitter））;
pipeline.add（（ProcessingResource）Factory.createResource（gate.stanford.Parser））;
语料库语料库= Factory.newCorpus（DepflowCorpus）;
文件文件= Factory.newDocument（港口和移民的帐单由堪萨斯共和国参议员布朗克提交。
corpus.add（document）; pipeline.setCorpus（语料库）; pipeline.execute（）;
AnnotationSet defaultAS = document.getAnnotations（）;
DepflowFinder finder = new DepflowFinder（defaultAS.get（Dependency））;
finder.findDepflowsFor（defaultAS.get（Token，Utils.featureMap（string，submitted））。iterator（）。next（）。getId（））; （int i = 0; i< finder.getNumDepflows（）; i ++）{
System.err.println（finder.getDepflow（i））;
System.err.println（finder.getKindflow（i））;
}
}
private ArrayListMultimap< Integer，Dependency> parentDependencies = ArrayListMultimap.create（）;
private List< DependencyLink> depflows = new ArrayList<（）;
public DepflowFinder（AnnotationSet dependencyAS）{
for（Annotation dep：dependencyAS）{
FeatureMap fm = dep.getFeatures（）;
@SuppressWarnings（unchecked）
依赖关系依赖=新的依赖关系（（List< Integer>）fm.get（args），（String）fm.get（kind））;
parentDependencies.put（dependency.getParent（），dependency）;
}
}
public void findDepflowsFor（Integer nodeId）{
depflows = new ArrayList& （Dependency dep：parentDependencies.get（nodeId））
{
findDepflowsInternal（new DependencyLink（dep，null））;
}
}
private void findDepflowsInternal（DependencyLink lastLink）{
整数节点= lastLink.dependency.getChild（）;
列表<依赖关系> children = parentDependencies.get（node）;
if（children.isEmpty（））{
depflows.add（lastLink）;返回;
}
（Dependency child：children）{
findDepflowsInternal（new DependencyLink（child，lastLink））;
}
}
public List< String> getKindflow（int i）{
DependencyLink last = depflows.get（i）;
LinkedList< String> ret = new LinkedList<>（）;
while（last！= null）{
ret.addFirst（last.dependency.kind）;
last = last.parentLink;
}
return ret;
}
public List< List< Integer>> getDepflow（int i）{
DependencyLink last = depflows.get（i）;
LinkedList< List< Integer>> ret = new LinkedList<>（）;
while（last！= null）{
ret.addFirst（last.dependency.args）;
last = last.parentLink;
}
return ret;
}
private static class Dependency {
List< Integer> ARGS;字符串种类
public Dependency（List< Integer> args，String kind）{
this.args = args; this.kind = kind;
}
public Integer getParent（）{return args.get（0）; }
public Integer getChild（）{return args.get（1）; }
}
私有静态类DependencyLink {
依赖关系依赖; DependencyLink parentLink;
public DependencyLink（依赖关系依赖关系，DependencyLink parentLink）{
this.dependency = dependency; this.parentLink = parentLink;
}
}
public int getNumDepflows（）{return depflows.size（）; }
public static void regiterGatePlugin（String name）throws Exception {
Gate.getCreoleRegister（）。registerDirectories（new File（Gate.getPluginsHome（），name）.toURI（）.toURL ））;
}
}
这里是输出：
[[12，10]]
[auxpass]
[[12，14]，[14，18] [18,21]，[21,23]，[23,25]]
[prep，pobj，appos，prep，pobj]
[[12，14]，[14,18] [18，16]]
[prep，pobj，nn]
[[12，0]，[0,2]，[2，4]，[4，8]]
$ s
[[12，0]，[0，2]，[2，4]，[4，6]]
[nsubjpass，prep，pobj，cc ]
I have used the parser in the gate to get the dependency. And from that dependency, I found separate list for the dependency pair and dependency name.
My input is
String idlist =[26, 28]
String argslist = [[26, 25], [26, 24], [26, 29], [26, 28], [28, 27], [35, 26], [29, 30]]
String kindlist =[ aux, dobj, pobj, prep, advmod, aux, adj]
My expected output is two list
Depflowlist is first one and another one kindflowlist.
for 26 itself it has different flow. for 26 expected depflowlist is [26, 25] and its corresponding kindflowlist [aux]
then [26, 24] is another depflowlist for 26 and its corresponding kindflowlist [dobj]
then [[26, 29] ,[29, 30]] is another depflowlist for 26 and its corresponding kindflowlist [pobj, adj].
This should be done for 28 also.
I have write the code like :
`public static void dependencyparse(String a,ArrayList<String> argslist,ArrayList<String> kindlist,Document doc){
ArrayList<String> selectarg_list = new ArrayList<String>();
for(int j=0;j<argslist.size();j++)
{
int index=argslist.get(j).indexOf(a);
if(index==1)
{
String next=argslist.get(j).substring(5, 7);
selectarg_list.add(next);
}
}
System.out.println(a+"---------->"+selectarg_list);
if(!selectarg_list.isEmpty())
{
// System.out.println("notempty");
hm.put(a, selectarg_list);
}
//System.out.println("Hashmap"+hm);
if(selectarg_list.size()>0){
dependencyparse(selectarg_list.get(0),argslist,kindlist,doc);
}
if(selectarg_list.size()==0){
String prevKey = delete(a,selectarg_list,argslist,kindlist,doc);
System.out.println(prevKey);
if(!hm.isEmpty()){
if(hm.keySet().contains(prevKey)){
ArrayList<String> list = hm.get(prevKey);
dependencyparse(list.get(0),argslist,kindlist,doc);
}
else{
if(!hm.isEmpty()){
try{
dependencyparse(hm.values().iterator().next().get(0), argslist, kindlist,doc);
}catch (Exception e) {}
}
}
Iterator<String> itr3 = hm.keySet().iterator();
while (itr3.hasNext()) {
String string = (String) itr3.next();
System.out.println(string+"------------>"+hm.get(string));
}
}
}
}
private static String delete(String a,ArrayList<String> selectarg_list,ArrayList<String> argslist,ArrayList<String> kindlist,Document doc) {
AnnotationSet outputAnnotations=doc.getNamedAnnotationSets().get("Output markups");
String retKey="";
Collection<ArrayList<String>> values = hm.values();
String key="";
//System.out.println("Valuess"+values);
//System.out.println("find"+a);
boolean flag=false;
Iterator<ArrayList<String>> Valueitr = values.iterator();
while(Valueitr.hasNext())
{
//System.out.println("Valuess---->"+Valueitr.next());
ArrayList<String> contentnext = Valueitr.next();
if(contentnext.contains(a))
{
flag=true;
//System.out.println("Valuess find---->"+contentnext);
for(Entry<String, ArrayList<String>> entry : hm.entrySet())
{
if(contentnext.equals(entry.getValue()))
{
key = entry.getKey();
// System.out.println("Key find---->"+key);
// do something with the key
String arg = "["+key+", "+a+"]";
// System.out.println("arg find---->"+arg);
depFlowList.addFirst(arg);
// System.out.println("entry find---->"+entry.getValue());
try{
if(!hm.keySet().contains(a)){
entry.getValue().remove(a);
// hm.put(entry.getKey(), value)
// System.out.println(entry.getKey()+"----------------"+hm.get(entry.getKey())+"\nentry delete---->"+entry.getValue());
}}catch (Exception e) { }
}
}
}
}
if(!flag){
if(!depFlowList.isEmpty()){
System.out.println("Dep Flow List----->"+depFlowList);
for(int x=0;x<depFlowList.size();x++){
String item=depFlowList.get(x);
int item_index=argslist.indexOf(item);
String item_kind=kindlist.get(item_index);
System.out.println("adding kind of -----"+depFlowList.get(x));
kindFlowList.add(item_kind);
}
System.out.println("Kind Flow List----->"+kindFlowList);
file_write("/home/cognicor/vagateplugin/testing/dependency.txt",kindFlowList.toString());
}
if(!kindFlowList.isEmpty()){
FeatureMap depfeature = new SimpleFeatureMapImpl();
depfeature.put("depflow", kindFlowList);
try {
outputAnnotations.add(new Long(0),new Long(1), "DependencyFlow",depfeature);
} catch (InvalidOffsetException e) {
// TODO Auto-generated catch block
e.printStackTrace();
}
}
if(depFlowList.size()==0){
return "";
}
retKey=depFlowList.getLast().substring(1,3);
depFlowList.clear();
kindFlowList.clear();
System.out.println(retKey);
return retKey;
}
//System.out.println("++++++++++++++++++++++++++++++++++++++++++++++");
ArrayList<String> findval = hm.get(key);
if(findval.isEmpty()){
hm.remove(key);
//System.out.println("key removed");
}
//if(!hm.isEmpty()){
retKey=delete(key, selectarg_list, argslist, kindlist,doc);
return retKey;
//}
}
public static void main(String[] args) {
// TODO Auto-generated method stub
ArrayList<String> testlist = file_read("/home/cognicor/vagateplugin/testing/testing_set1.txt");
//System.out.println(testlist);
ArrayList<String> query_class = new ArrayList<String>();
ArrayList<String> property_class = new ArrayList<String>();
Iterator<String> testinput = testlist.iterator();
String entry;
System.out.println("testlist---->"+testlist);
for(int i=0;i<testlist.size();i++)
{
try {
ArrayList<Integer> idlist = new ArrayList<Integer>();
ArrayList<String> argslist = new ArrayList<String>();
ArrayList<String> kindlist = new ArrayList<String>();
String query = "<currDial>"+testlist.get(i)+"</currDial>";
System.out.println(query);
Document doc = pipline.get_Annoted_Doument(query);
/* code where i get the input like idlist, arglist and kindlist*/
AnnotationSet annotset = doc.getNamedAnnotationSets().get("Output markups").get("TokenID");
Iterator<Annotation> itr = annotset.iterator();
while(itr.hasNext())
{
Annotation annot = itr.next();
int id = (int) annot.getFeatures().get("id");
idlist.add(id);
}
AnnotationSet annotset1 = doc.getNamedAnnotationSets().get("Output markups").get("Dependency");
Iterator<Annotation> itr1 = annotset1.iterator();
while(itr1.hasNext())
{
Annotation annot = itr1.next();
String arg = annot.getFeatures().get("args").toString();
String kind = annot.getFeatures().get("kind").toString();
argslist.add(arg);
kindlist.add(kind);
}
System.out.println("args"+argslist);
for(int j=0;j<idlist.size();j++)
{
System.out.println("-------------------------------Item:"+idlist.get(j).toString()+"------------------------------------------");
dependencyparse(idlist.get(j).toString(),argslist,kindlist,doc);
if(j==idlist.size())
{
dependencyparse("null",argslist,kindlist,doc);
}
}
}catch(Exception e) { e.printStackTrace(); }
}
}
}`
Now I got a loop which is not ending . I am using the output from the gate.
解决方案
Here is my code. Its in Java 7 + Guava. Note that I'm also calling the Stanford Parser and reading the input directly from annotations (no String parsing is needed).
The parsed sentence comes from nlp.stanford.edu.
import gate.*;
import gate.creole.SerialAnalyserController;
import java.io.File;
import java.util.*;
import com.google.common.collect.ArrayListMultimap;
public class DepflowFinder {
public static void main(String[] args) throws Exception {
Gate.setGateHome(new File("C:\\Program Files\\GATE_Developer_8.1"));
Gate.init();
regiterGatePlugin("ANNIE"); regiterGatePlugin("Stanford_CoreNLP" /*"Parser_Stanford"*/);
SerialAnalyserController pipeline = (SerialAnalyserController) Factory.createResource("gate.creole.SerialAnalyserController");
pipeline.add((ProcessingResource) Factory.createResource("gate.creole.tokeniser.DefaultTokeniser"));
pipeline.add((ProcessingResource) Factory.createResource("gate.creole.splitter.SentenceSplitter"));
pipeline.add((ProcessingResource) Factory.createResource("gate.stanford.Parser"));
Corpus corpus = Factory.newCorpus("DepflowCorpus");
Document document = Factory.newDocument("Bills on ports and immigration were submitted by Senator Brownback, Republican of Kansas.");
corpus.add(document); pipeline.setCorpus(corpus); pipeline.execute();
AnnotationSet defaultAS = document.getAnnotations();
DepflowFinder finder = new DepflowFinder(defaultAS.get("Dependency"));
finder.findDepflowsFor(defaultAS.get("Token", Utils.featureMap("string", "submitted")).iterator().next().getId());
for (int i=0; i<finder.getNumDepflows(); i++) {
System.err.println(finder.getDepflow(i));
System.err.println(finder.getKindflow(i));
}
}
private ArrayListMultimap<Integer, Dependency> parentDependencies = ArrayListMultimap.create();
private List<DependencyLink> depflows = new ArrayList<>();
public DepflowFinder(AnnotationSet dependencyAS) {
for (Annotation dep : dependencyAS) {
FeatureMap fm = dep.getFeatures();
@SuppressWarnings("unchecked")
Dependency dependency = new Dependency((List<Integer>) fm.get("args"), (String) fm.get("kind"));
parentDependencies.put(dependency.getParent(), dependency);
}
}
public void findDepflowsFor(Integer nodeId) {
depflows = new ArrayList<>();
for (Dependency dep : parentDependencies.get(nodeId)) {
findDepflowsInternal(new DependencyLink(dep, null));
}
}
private void findDepflowsInternal(DependencyLink lastLink) {
Integer node = lastLink.dependency.getChild();
List<Dependency> children = parentDependencies.get(node);
if (children.isEmpty()) {
depflows.add(lastLink); return;
}
for (Dependency child : children) {
findDepflowsInternal(new DependencyLink(child, lastLink));
}
}
public List<String> getKindflow(int i) {
DependencyLink last = depflows.get(i);
LinkedList<String> ret = new LinkedList<>();
while (last != null) {
ret.addFirst(last.dependency.kind);
last = last.parentLink;
}
return ret;
}
public List<List<Integer>> getDepflow(int i) {
DependencyLink last = depflows.get(i);
LinkedList<List<Integer>> ret = new LinkedList<>();
while (last != null) {
ret.addFirst(last.dependency.args);
last = last.parentLink;
}
return ret;
}
private static class Dependency {
List<Integer> args; String kind;
public Dependency(List<Integer> args, String kind) {
this.args = args; this.kind = kind;
}
public Integer getParent() {return args.get(0); }
public Integer getChild() {return args.get(1); }
}
private static class DependencyLink {
Dependency dependency; DependencyLink parentLink;
public DependencyLink(Dependency dependency, DependencyLink parentLink) {
this.dependency = dependency; this.parentLink = parentLink;
}
}
public int getNumDepflows() {return depflows.size(); }
public static void regiterGatePlugin(String name) throws Exception {
Gate.getCreoleRegister().registerDirectories(new File(Gate.getPluginsHome(), name).toURI().toURL());
}
}
And here is the output:
[[12, 10]]
[auxpass]
[[12, 14], [14, 18], [18, 21], [21, 23], [23, 25]]
[prep, pobj, appos, prep, pobj]
[[12, 14], [14, 18], [18, 16]]
[prep, pobj, nn]
[[12, 0], [0, 2], [2, 4], [4, 8]]
[nsubjpass, prep, pobj, conj]
[[12, 0], [0, 2], [2, 4], [4, 6]]
[nsubjpass, prep, pobj, cc]
这篇关于从解析器的输出找到依赖关系的流的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！
查看全文
