/// <summary>
/// 依存句法分析（神经网络句法模型需要-Xms1g -Xmx1g -Xmn512m）
/// 内部采用NeuralNetworkDependencyParser实现，用户可以直接调用NeuralNetworkDependencyParser.compute(sentence)
/// 也可以调用基于MaxEnt的依存句法分析器MaxEntDependencyParser.compute(sentence)
/// 《基于神经网络的高性能依存句法分析器》
///https://www.hankcs.com/nlp/parsing/neural-network-based-dependency-parser.html
///《最大熵依存句法分析器的实现》
///https://www.hankcs.com/nlp/parsing/to-achieve-the-maximum-entropy-of-the-dependency-parser.html
///《基于CRF序列标注的中文依存句法分析器的Java实现》
///https://www.hankcs.com/nlp/parsing/crf-sequence-annotation-chinese-dependency-parser-implementation-based-on-java.html
/// </summary>
public void DependencyParser()
{
CoNLLSentence sentence = HanLP.parseDependency("徐先生还具体帮助他确定了把画雄鹰、松鼠和麻雀作为主攻目标。");
Console.WriteLine(sentence);
// 可以方便地遍历它
foreach (CoNLLWord word in sentence)
{
Console.WriteLine($"{word.LEMMA} --({word.DEPREL})--> {word.HEAD.LEMMA}\n");
}
// 也可以直接拿到数组，任意顺序或逆序遍历
CoNLLWord[] wordArray = sentence.getWordArray();
for (int i = wordArray.Length - 1; i >= 0; i--)
{
CoNLLWord word = wordArray[i];
Console.WriteLine($"{word.LEMMA} --({word.DEPREL})-->{word.HEAD.LEMMA}\n");
}
// 还可以直接遍历子树，从某棵子树的某个节点一路遍历到虚根
CoNLLWord head = wordArray[12];
while ((head = head.HEAD) != null)
{
if (head == CoNLLWord.ROOT)
{
Console.WriteLine(head.LEMMA);
}
else
{
Console.WriteLine($"{head.LEMMA} --({head.DEPREL})--> ");
}
}
}
//static void LemmatizeMate(List<CoNLLSentence> corpus, String exportPath, MateTools mateTools)
//{
// int count = corpus.Count;
// for (int i = 0; i < count; i++)
// {
// CoNLLSentence sentence = corpus[i];
// mateTools.ProcessSentence(sentence);
// Console.WriteLine(i);
// }
// XMLSerializer.Serialize<List<CoNLLSentence>>(corpus, exportPath);
//}
static void LemmatizeTreeTagger(List <CoNLLSentence> corpus, string exportPath, TreeTagger treeTagger)
{
int count = corpus.Count;
for (int j = 0; j < count; j++)
{
CoNLLSentence sentence = corpus[j];
treeTagger.ProcessSentence(sentence);
Console.WriteLine(j);
}
XMLSerializer.Serialize <List <CoNLLSentence> >(corpus, exportPath);
}
public void ProcessSentence(CoNLLSentence sentence)
{
string[] tokenArray = sentence.Tokens.Select(x => x.Form).ToArray();
is2.data.SentenceData09 sentenceMateTools = mateToolsWrapper.TagSentenceLemmatizerAndPOS(tokenArray, true);
for (int i = 0; i < sentence.Tokens.Count; i++)
{
String mateToolsLemma = sentenceMateTools.plemmas[i + 1]; // zero based index is ROOT node
sentence.Tokens[i].PredictedLemmas = new List<string>();
sentence.Tokens[i].PredictedLemmas.Add(mateToolsLemma);
}
}
static void LemmatizeMorphy(List <CoNLLSentence> corpus, string exportPath)
{
Morphy morphy = new Morphy();
morphy.InitMorphy(AppSettingsWrapper.MorphyCSV);
int count = corpus.Count;
for (int i = 0; i < count; i++)
{
CoNLLSentence sentence = corpus[i];
morphy.ProcessSentence(sentence);
//Console.WriteLine(i);
}
XMLSerializer.Serialize <List <CoNLLSentence> >(corpus, exportPath);
}
static void LemmatizeIWNLP(List <CoNLLSentence> corpus, string exportPath)
{
Lemmatizer IWNLP = new Lemmatizer();
IWNLP.Load(AppSettingsWrapper.IWNLPPath);
int count = corpus.Count;
for (int i = 0; i < count; i++)
{
CoNLLSentence sentence = corpus[i];
IWNLPSentenceProcessor.ProcessSentence(sentence, IWNLP);
//Console.WriteLine(i);
}
XMLSerializer.Serialize <List <CoNLLSentence> >(corpus, exportPath);
}
public void ProcessSentence(CoNLLSentence sentence)
{
string[] tokenArray = sentence.Tokens.Select(x => x.Form).ToArray();
for (int i = 0; i < sentence.Tokens.Count; i++)
{
CoNLLToken token = sentence.Tokens[i];
if (token.POS == "NN")
{
if (morphyDictionary.ContainsKey(token.Form))
{
token.PredictedLemmas = new List<string>();
token.PredictedLemmas = morphyDictionary[token.Form];
}
//else if (morphyDictionary.ContainsKey(token.Form.ToLower())) // adding a lower case comparison worsens the results
//{
// token.PredictedLemmas = new List<string>();
// token.PredictedLemmas = morphyDictionary[token.Form.ToLower()];
//}
}
else if (token.POS == "ADJA" || token.POS == "ADJD")
{
if (morphyDictionary.ContainsKey(token.Form))
{
token.PredictedLemmas = new List<string>();
token.PredictedLemmas = morphyDictionary[token.Form];
}
else if (morphyDictionary.ContainsKey(token.Form.ToLower()))
{
token.PredictedLemmas = new List<string>();
token.PredictedLemmas = morphyDictionary[token.Form.ToLower()];
}
}
else if (token.POS.StartsWith("V"))
{
if (morphyDictionary.ContainsKey(token.Form))
{
token.PredictedLemmas = new List<string>();
token.PredictedLemmas = morphyDictionary[token.Form];
}
else if (morphyDictionary.ContainsKey(token.Form.ToLower()))
{
token.PredictedLemmas = new List<string>();
token.PredictedLemmas = morphyDictionary[token.Form.ToLower()];
}
}
}
}
public void ProcessSentence(CoNLLSentence sentence)
{
string[] tokenArray = sentence.Tokens.Select(x => x.Form).ToArray();
for (int i = 0; i < sentence.Tokens.Count; i++)
{
CoNLLToken token = sentence.Tokens[i];
if (token.POS == "NN")
{
if (morphyDictionary.ContainsKey(token.Form))
{
token.PredictedLemmas = new List <string>();
token.PredictedLemmas = morphyDictionary[token.Form];
}
//else if (morphyDictionary.ContainsKey(token.Form.ToLower())) // adding a lower case comparison worsens the results
//{
// token.PredictedLemmas = new List<string>();
// token.PredictedLemmas = morphyDictionary[token.Form.ToLower()];
//}
}
else if (token.POS == "ADJA" || token.POS == "ADJD")
{
if (morphyDictionary.ContainsKey(token.Form))
{
token.PredictedLemmas = new List <string>();
token.PredictedLemmas = morphyDictionary[token.Form];
}
else if (morphyDictionary.ContainsKey(token.Form.ToLower()))
{
token.PredictedLemmas = new List <string>();
token.PredictedLemmas = morphyDictionary[token.Form.ToLower()];
}
}
else if (token.POS.StartsWith("V"))
{
if (morphyDictionary.ContainsKey(token.Form))
{
token.PredictedLemmas = new List <string>();
token.PredictedLemmas = morphyDictionary[token.Form];
}
else if (morphyDictionary.ContainsKey(token.Form.ToLower()))
{
token.PredictedLemmas = new List <string>();
token.PredictedLemmas = morphyDictionary[token.Form.ToLower()];
}
}
}
}
public void ProcessSentence(CoNLLSentence sentence)
{
StringBuilder tokenString = new StringBuilder();
foreach (CoNLLToken token in sentence.Tokens)
{
tokenString.AppendLine(String.Format("{0}", token.Form));
}
System.IO.File.WriteAllText(inputPath, tokenString.ToString(), Encoding.UTF8);
//german-utf8.par input.txt output.txt -token -lemma
Process treeTaggerProcess = new Process()
{
StartInfo = new ProcessStartInfo()
{
FileName = AppSettingsWrapper.TreeTagger.TreeTaggerExePath,
Arguments = String.Format("\"{0}\" \"{1}\" \"{2}\" -token -lemma",
AppSettingsWrapper.TreeTagger.TreeTaggerGermanPath,
inputPath,
outputPath),
RedirectStandardOutput = true,
UseShellExecute = false,
CreateNoWindow = true
}
};
treeTaggerProcess.Start();
// the new process runs asynchronous. Wait until it stops to read its output afterwards
treeTaggerProcess.WaitForExit();
String[] allLinesOuputFile = System.IO.File.ReadAllLines(outputPath, Encoding.UTF8);
for (int i = 0; i < sentence.Tokens.Count; i++)
{
String[] line = allLinesOuputFile[i].Split(new String[] { "\t" }, StringSplitOptions.RemoveEmptyEntries);
if (line.Length == 3 && line[2] != "<unknown>")
{
sentence.Tokens[i].PredictedLemmas = line[2].Split(new String[] { "|" }, StringSplitOptions.RemoveEmptyEntries).ToList();
// TreeTagger can return multiple lemmas
// Example: Stiften NN Stift|Stiften
// For instance "Stiften" will return "Stift" and "Stiften".
}
}
}
public void ProcessSentence(CoNLLSentence sentence)
{
StringBuilder tokenString = new StringBuilder();
foreach (CoNLLToken token in sentence.Tokens)
{
tokenString.AppendLine(string.Format("{0}", token.Form));
}
System.IO.File.WriteAllText(inputPath, tokenString.ToString(), Encoding.UTF8);
//german-utf8.par input.txt output.txt -token -lemma
Process treeTaggerProcess = new Process()
{
StartInfo = new ProcessStartInfo()
{
FileName = AppSettingsWrapper.TreeTagger.TreeTaggerExePath,
Arguments = string.Format("\"{0}\" \"{1}\" \"{2}\" -token -lemma",
AppSettingsWrapper.TreeTagger.TreeTaggerGermanPath,
inputPath,
outputPath),
RedirectStandardOutput = true,
UseShellExecute = false,
CreateNoWindow = true
}
};
treeTaggerProcess.Start();
// the new process runs asynchronous. Wait until it stops to read its output afterwards
treeTaggerProcess.WaitForExit();
string[] allLinesOuputFile = System.IO.File.ReadAllLines(outputPath, Encoding.UTF8);
for (int i = 0; i < sentence.Tokens.Count; i++)
{
string[] line = allLinesOuputFile[i].Split(new string[] { "\t" }, StringSplitOptions.RemoveEmptyEntries);
if (line.Length == 3 && line[2] != "<unknown>")
{
sentence.Tokens[i].PredictedLemmas = line[2].Split(new string[] { "|" }, StringSplitOptions.RemoveEmptyEntries).ToList();
// TreeTagger can return multiple lemmas
// Example: Stiften NN Stift|Stiften
// For instance "Stiften" will return "Stift" and "Stiften".
}
}
}
//protected bool IsExactMatchOrGuess(String goldLemma, String form, List<String> lemmas)
//{
// if (lemmas == null || lemmas.Count == 0)
// {
// return (form == goldLemma);
// }
// if (lemmas.Count > 1)
// {
// return false;
// }
// else
// {
// return goldLemma.ToLower() == lemmas[0].ToLower();
// }
//}
public DetailedLookupResults EvaluateTwoResources(string path, string path2, string comment)
{
Console.WriteLine(comment);
List <CoNLLSentence> sentences = XMLSerializer.Deserialize <List <CoNLLSentence> >(path);
List <CoNLLSentence> sentences2 = XMLSerializer.Deserialize <List <CoNLLSentence> >(path2);
DetailedLookupResults result = new DetailedLookupResults()
{
TotalNounCount = sentences.SelectMany(x => x.Tokens).Count(x => x.POS == "NN" && !x.Lemma.Contains("|") && !x.Lemma.Contains("_") && (x.Lemma != "unknown" && x.Form != "unknown")),
TotalVerbCount = sentences.SelectMany(x => x.Tokens).Count(x => x.POS.StartsWith("V")),
TotalAdjectiveCount = sentences.SelectMany(x => x.Tokens).Count(x => (x.POS == "ADJA" || x.POS == "ADJD") && (x.Lemma != "NULL" && x.Form != "NULL"))// the second condition is for tokens in the HDT corpus that have the lemma "NULL"
};
for (int i = 0; i < sentences.Count; i++)
{
CoNLLSentence sentence = sentences[i];
for (int j = 0; j < sentence.Tokens.Count; j++)
{
CoNLLToken token = sentence.Tokens[j];
if (token.POS == "NN")
{
if (token.Lemma.Contains("|") || token.Lemma.Contains("_") || (token.Lemma == "unknown" && token.Form != "unknown"))
{
continue;
}
if (!(token.PredictedLemmas == null || token.PredictedLemmas.Count == 0))
{
if (token.PredictedLemmas.Count == 1)
{
if (IsLowerCaseExactMatch(token.Lemma, token.PredictedLemmas))
{
result.NounsCorrectlyLemmatizedCount++;
}
}
else // if more than one lemma is found, compare with the second resource
{
if (IsLowerCaseExactMatch(token.Lemma, sentences2[i].Tokens[j].PredictedLemmas))
{
result.NounsCorrectlyLemmatizedCount++;
}
}
}
else // if no lemma is found, compare with the second resource
{
if (IsLowerCaseExactMatch(token.Lemma, sentences2[i].Tokens[j].PredictedLemmas))
{
result.NounsCorrectlyLemmatizedCount++;
}
}
}
else if (token.POS.StartsWith("V"))
{
NormalizeVerbToken(token);
if (!(token.PredictedLemmas == null || token.PredictedLemmas.Count == 0))
{
if (token.PredictedLemmas.Count == 1)
{
if (IsLowerCaseExactMatch(token.Lemma, token.PredictedLemmas))
{
result.VerbsCorrectlyLemmatizedCount++;
}
}
else // if more than one lemma is found, compare with the second resource
{
if (IsLowerCaseExactMatch(token.Lemma, sentences2[i].Tokens[j].PredictedLemmas))
{
result.VerbsCorrectlyLemmatizedCount++;
}
}
}
else // if no lemma is found, compare with the second resource
{
if (IsLowerCaseExactMatch(token.Lemma, sentences2[i].Tokens[j].PredictedLemmas))
{
result.VerbsCorrectlyLemmatizedCount++;
}
}
}
else if (token.POS == "ADJA" || token.POS == "ADJD")
{
if (token.Lemma == "NULL" && token.Form != "NULL") // ~ 2000 adjectives in the HDT corpus have "NULL" as lemma. Ignore them for the evaluation
{
continue;
}
if (!(token.PredictedLemmas == null || token.PredictedLemmas.Count == 0))
{
if (token.PredictedLemmas.Count == 1)
{
if (IsLowerCaseExactMatch(token.Lemma, token.PredictedLemmas))
{
result.AdjectivesCorrectlyLemmatizedCount++;
}
}
else // if more than one lemma is found, compare with the second resource
{
if (IsLowerCaseExactMatch(token.Lemma, sentences2[i].Tokens[j].PredictedLemmas))
{
result.AdjectivesCorrectlyLemmatizedCount++;
}
}
}
else // if no lemma is found, compare with the second resource
{
if (IsLowerCaseExactMatch(token.Lemma, sentences2[i].Tokens[j].PredictedLemmas))
{
result.AdjectivesCorrectlyLemmatizedCount++;
}
}
}
}
}
Console.WriteLine(result.ToString());
return(result);
}
public DetailedLookupResults Evaluate(string path, string comment)
{
Console.WriteLine(comment);
List <CoNLLSentence> sentences = XMLSerializer.Deserialize <List <CoNLLSentence> >(path);
DetailedLookupResults result = new DetailedLookupResults()
{
TotalNounCount = sentences.SelectMany(x => x.Tokens).Count(x => x.POS == "NN" && !x.Lemma.Contains("|") && !x.Lemma.Contains("_") && (x.Lemma != "unknown" && x.Form != "unknown")),
TotalVerbCount = sentences.SelectMany(x => x.Tokens).Count(x => x.POS.StartsWith("V")),
TotalAdjectiveCount = sentences.SelectMany(x => x.Tokens).Count(x => (x.POS == "ADJA" || x.POS == "ADJD") && (x.Lemma != "NULL" && x.Form != "NULL"))// the second condition is for tokens in the HDT corpus that have the lemma "NULL"
};
for (int i = 0; i < sentences.Count; i++)
{
CoNLLSentence sentence = sentences[i];
for (int j = 0; j < sentence.Tokens.Count; j++)
{
CoNLLToken token = sentence.Tokens[j];
if (token.POS == "NN")
{
if (token.Lemma.Contains("|") || token.Lemma.Contains("_") || (token.Lemma == "unknown" && token.Form != "unknown"))
{
continue;
}
if (IsLowerCaseExactMatch(token.Lemma, token.PredictedLemmas))
{
result.NounsCorrectlyLemmatizedCount++;
}
else
{
result.AddLookup(PartOfSpeech.Noun, token);
}
}
else if (token.POS.StartsWith("V"))
{
NormalizeVerbToken(token);
if (IsLowerCaseExactMatch(token.Lemma, token.PredictedLemmas))
{
result.VerbsCorrectlyLemmatizedCount++;
}
else
{
result.AddLookup(PartOfSpeech.Verb, token);
}
}
else if (token.POS == "ADJA" || token.POS == "ADJD")
{
if (token.Lemma == "NULL" && token.Form != "NULL") // ~ 2000 adjectives in the HDT corpus have "NULL" as lemma. Ignore them for the evaluation
{
continue;
}
if (IsLowerCaseExactMatch(token.Lemma, token.PredictedLemmas))
{
result.AdjectivesCorrectlyLemmatizedCount++;
}
else
{
result.AddLookup(PartOfSpeech.Adjective, token);
}
//else
//{
// if (token.PredictedLemmas != null && token.PredictedLemmas.Count == 1)
// {
// String key = String.Format("{0}->{1} != {2}", token.Form, token.PredictedLemmas[0], token.Lemma);
// if (!wrongMappings.ContainsKey(key))
// {
// wrongMappings.Add(key, 0);
// }
// wrongMappings[key] = wrongMappings[key] + 1;
// }
//}
}
}
}
//var wrongMappingSorted = wrongMappings.OrderByDescending(x => x.Value);
//foreach (var entry in wrongMappingSorted.Take(100))
//{
// Console.WriteLine(entry.Key + ": " + entry.Value);
//}
Console.WriteLine(result.ToString());
return(result);
}
public static void ProcessSentence(CoNLLSentence sentence, Lemmatizer iwnlp)
{
string[] tokenArray = sentence.Tokens.Select(x => x.Form).ToArray();
//is2.data.SentenceData09 sentenceMateTools = mateToolsWrapper.TagSentenceLemmatizerAndPOS(tokenArray, true);
for (int i = 0; i < sentence.Tokens.Count; i++)
{
CoNLLToken token = sentence.Tokens[i];
if (token.POS == "NN")
{
List<POS> pos = new List<POS>() { POS.Noun, POS.X };
if (iwnlp.ContainsEntry(token.Form, POS.Noun))
{
token.PredictedLemmas = iwnlp.GetLemmas(token.Form, POS.Noun);
}
else if (iwnlp.ContainsEntry(token.Form, POS.X))
{
token.PredictedLemmas = iwnlp.GetLemmas(token.Form, POS.X);
}
else if (iwnlp.ContainsEntry(token.Form, POS.AdjectivalDeclension))
{
token.PredictedLemmas = iwnlp.GetLemmas(token.Form, POS.AdjectivalDeclension);
}
else if(iwnlp.ContainsEntry(token.Form, pos, true))
{
token.PredictedLemmas = iwnlp.GetLemmas(token.Form, pos, true);
}
}
else
{
if (token.POS == "ADJA" || token.POS == "ADJD")
{
if (iwnlp.ContainsEntry(token.Form, POS.Adjective))
{
token.PredictedLemmas = iwnlp.GetLemmas(token.Form, POS.Adjective);
}
else if (iwnlp.ContainsEntry(token.Form, POS.Adjective, true))
{
token.PredictedLemmas = iwnlp.GetLemmas(token.Form, POS.Adjective, true);
}
else if (iwnlp.ContainsEntry(token.Form, POS.Noun, true))
{
token.PredictedLemmas = iwnlp.GetLemmas(token.Form, POS.Noun, true);
}
else if (iwnlp.ContainsEntry(token.Form, POS.X, true))
{
token.PredictedLemmas = iwnlp.GetLemmas(token.Form, POS.X, true);
}
else if (iwnlp.ContainsEntry(token.Form, POS.Verb, true))
{
token.PredictedLemmas = iwnlp.GetLemmas(token.Form, POS.Verb, true);
}
//else if (iwnlp.ContainsEntry(token.Form,true))
//{
// token.PredictedLemmas = iwnlp.GetLemmas(token.Form, true);
//}
}
else if (token.POS.StartsWith("V"))
{
if (iwnlp.ContainsEntry(token.Form, POS.Verb, true))
{
token.PredictedLemmas = iwnlp.GetLemmas(token.Form, POS.Verb, true);
}
// test
//else if (iwnlp.ContainsEntry(token.Form, true))
//{
// token.PredictedLemmas = iwnlp.GetLemmas(token.Form, true);
//}
}
}
}
}
public List <CoNLLSentence> ReadFile(string path, Corpus corpus)
{
System.IO.StreamReader file = new System.IO.StreamReader(path);
List <CoNLLSentence> sentences = new List <CoNLLSentence>();
int sentenceNumber = 1;
CoNLLSentence nextSentence = new CoNLLSentence();
nextSentence.ID = sentenceNumber++;
nextSentence.Tokens = new List <CoNLLToken>();
string line;
while ((line = file.ReadLine()) != null)
{
if (line.StartsWith("#"))
{
continue;
}
if (string.IsNullOrEmpty(line))
{
sentences.Add(nextSentence);
nextSentence = new CoNLLSentence();
nextSentence.ID = sentenceNumber++;
nextSentence.Tokens = new List <CoNLLToken>();
}
else
{
string[] values = line.Split(new string[] { "\t" }, StringSplitOptions.RemoveEmptyEntries);
if (corpus == Corpus.Tiger || corpus == Corpus.HDT)
{
nextSentence.Tokens.Add(new CoNLLToken()
{
ID = values[0],
Form = values[1],
Lemma = values[2],
POS = values[4],
});
}
else
{
string lemma = values[6];
if (lemma == "#refl")
{
lemma = string.Empty;
}
if (lemma.Contains("#"))
{
lemma = lemma.Replace("#", string.Empty);
}
nextSentence.Tokens.Add(new CoNLLToken()
{
ID = values[2],
Form = values[3],
Lemma = lemma,
POS = values[4],
});
}
}
}
return(sentences);
}
public List<CoNLLSentence> ReadFile(String path, Corpus corpus)
{
System.IO.StreamReader file = new System.IO.StreamReader(path);
List<CoNLLSentence> sentences = new List<CoNLLSentence>();
int sentenceNumber = 1;
CoNLLSentence nextSentence = new CoNLLSentence();
nextSentence.ID = sentenceNumber++;
nextSentence.Tokens = new List<CoNLLToken>();
String line;
while ((line = file.ReadLine()) != null)
{
if (line.StartsWith("#"))
{
continue;
}
if (String.IsNullOrEmpty(line))
{
sentences.Add(nextSentence);
nextSentence = new CoNLLSentence();
nextSentence.ID = sentenceNumber++;
nextSentence.Tokens = new List<CoNLLToken>();
}
else
{
String[] values = line.Split(new String[] { "\t" }, StringSplitOptions.RemoveEmptyEntries);
if (corpus == Corpus.Tiger || corpus == Corpus.HDT)
{
nextSentence.Tokens.Add(new CoNLLToken()
{
ID = values[0],
Form = values[1],
Lemma = values[2],
POS = values[4],
});
}
else
{
String lemma = values[6];
if (lemma == "#refl")
{
lemma = String.Empty;
}
if (lemma.Contains("#"))
{
lemma = lemma.Replace("#", String.Empty);
}
nextSentence.Tokens.Add(new CoNLLToken()
{
ID = values[2],
Form = values[3],
Lemma = lemma,
POS = values[4],
});
}
}
}
return sentences;
}
public static void ProcessSentence(CoNLLSentence sentence, Lemmatizer iwnlp)
{
string[] tokenArray = sentence.Tokens.Select(x => x.Form).ToArray();
//is2.data.SentenceData09 sentenceMateTools = mateToolsWrapper.TagSentenceLemmatizerAndPOS(tokenArray, true);
for (int i = 0; i < sentence.Tokens.Count; i++)
{
CoNLLToken token = sentence.Tokens[i];
if (token.POS == "NN")
{
List <POS> pos = new List <POS>()
{
POS.Noun, POS.X
};
if (iwnlp.ContainsEntry(token.Form, POS.Noun))
{
token.PredictedLemmas = iwnlp.GetLemmas(token.Form, POS.Noun);
}
else if (iwnlp.ContainsEntry(token.Form, POS.X))
{
token.PredictedLemmas = iwnlp.GetLemmas(token.Form, POS.X);
}
else if (iwnlp.ContainsEntry(token.Form, POS.AdjectivalDeclension))
{
token.PredictedLemmas = iwnlp.GetLemmas(token.Form, POS.AdjectivalDeclension);
}
else if (iwnlp.ContainsEntry(token.Form, pos, true))
{
token.PredictedLemmas = iwnlp.GetLemmas(token.Form, pos, true);
}
}
else
{
if (token.POS == "ADJA" || token.POS == "ADJD")
{
if (iwnlp.ContainsEntry(token.Form, POS.Adjective))
{
token.PredictedLemmas = iwnlp.GetLemmas(token.Form, POS.Adjective);
}
else if (iwnlp.ContainsEntry(token.Form, POS.Adjective, true))
{
token.PredictedLemmas = iwnlp.GetLemmas(token.Form, POS.Adjective, true);
}
else if (iwnlp.ContainsEntry(token.Form, POS.AdjectivalDeclension))
{
token.PredictedLemmas = iwnlp.GetLemmas(token.Form, POS.AdjectivalDeclension);
}
else if (iwnlp.ContainsEntry(token.Form, POS.AdjectivalDeclension, true))
{
token.PredictedLemmas = iwnlp.GetLemmas(token.Form, POS.AdjectivalDeclension, true);
}
else if (iwnlp.ContainsEntry(token.Form, POS.Noun, true))
{
token.PredictedLemmas = iwnlp.GetLemmas(token.Form, POS.Noun, true);
}
else if (iwnlp.ContainsEntry(token.Form, POS.X, true))
{
token.PredictedLemmas = iwnlp.GetLemmas(token.Form, POS.X, true);
}
else if (iwnlp.ContainsEntry(token.Form, POS.Verb, true))
{
token.PredictedLemmas = iwnlp.GetLemmas(token.Form, POS.Verb, true);
}
//else if (iwnlp.ContainsEntry(token.Form,true))
//{
// token.PredictedLemmas = iwnlp.GetLemmas(token.Form, true);
//}
}
else if (token.POS.StartsWith("V"))
{
if (iwnlp.ContainsEntry(token.Form, POS.Verb, true))
{
token.PredictedLemmas = iwnlp.GetLemmas(token.Form, POS.Verb, true);
}
// test
//else if (iwnlp.ContainsEntry(token.Form, true))
//{
// token.PredictedLemmas = iwnlp.GetLemmas(token.Form, true);
//}
}
}
}
}
