public static void RefreshStemDic(string des = null)
{
lock (stemmerLocker)
{
if (stemWordDic != null)
{
if (des == null)
{
des = (string)DefaultParameter.Get(DefaultParameter.Field.stem_map);
//des = (string)GlobalParameter.Get(DefaultParameter.stem_map);
}
if (stemWordDic == null)
{
return;
}
FileWriter writer = new LargeFileWriter(des, FileMode.Create);
foreach (var word in stemWordDic.Keys)
{
writer.WriteLine(word + "\t" + stemWordDic[word]);
}
writer.Close();
stemWordDic = null;
}
}
}
static void Temp()
{
var sourceDir = @"D:\Codes\Project\EntityTyping\Fine-ner\input\tmp\";
var des = @"D:\Codes\Project\EntityTyping\Fine-ner\input\keywords.txt";
var files = Directory.GetFiles(sourceDir);
var reader = new LargeFileReader();
var writer = new LargeFileWriter(des, FileMode.Create);
var line = "";
var keyWords = new HashSet <string>();
foreach (var file in files)
{
reader.Open(file);
int count = 0;
while ((line = reader.ReadLine()) != null)
{
count++;
if (count > 100)
{
break;
}
var array = line.Split('\t');
keyWords.Add(array[0]);
}
}
reader.Close();
foreach (var word in keyWords)
{
writer.WriteLine(word);
}
writer.Close();
}
public static void Temp4()
{
var source = @"D:\Codes\Project\EntityTyping\Fine-ner\input\dictionaries\dbpedia\dbpedia entity type.txt";
var reader = new pml.file.reader.LargeFileReader(source);
var des = @"D:\Codes\Project\EntityTyping\Fine-ner\input\dictionaries\dbpedia\tmp.txt";
var writer = new LargeFileWriter(des, FileMode.Create);
string line;
int count = 0;
var set = new HashSet <string>();
var dic = new Dictionary <string, int>();
var times = 0;
while ((line = reader.ReadLine()) != null)
{
if (++count % 10000 == 0)
{
Console.WriteLine(count);
}
var array = line.Split('\t');
dic.TryGetValue(array[1], out times);
dic[array[1]] = times + 1;
}
reader.Close();
foreach (var type in dic.OrderByDescending(key => key.Value))
{
writer.WriteLine(type.Key + "\t" + type.Value);
}
writer.Close();
}
static void ParseWord2Vec(string word2vecFile, string parsedFile)
{
var writer = new LargeFileWriter(parsedFile, FileMode.Create);
var parser = new ParseBinaryVector(word2vecFile);
int count = 0;
while (!parser.EOF)
{
if (++count % 1000 == 0)
{
Console.WriteLine(count);
}
try
{
var pair = parser.GetNextVector();
var index = pair.first.IndexOf("en/");
writer.Write(pair.first.Substring(index + 3));
foreach (var value in pair.second)
{
writer.Write(string.Format(" {0}", value));
}
writer.Write("\r");
}
catch (Exception)
{
continue;
}
}
writer.Close();
}
public static void Temp5()
{
var sourceDir = @"D:\Codes\Project\EntityTyping\Fine-ner\input\feature\test";
var sourceFiles = Directory.GetFiles(sourceDir).ToList();
var desFile = @"D:\Codes\Project\EntityTyping\Fine-ner\input\feature\test data in dbpedia info.txt";
var writer = new LargeFileWriter(desFile, FileMode.Create);
for (var i = 0; i < sourceFiles.Count; i++)
{
var reader = new EventReaderByLine(sourceFiles[i]);
int count = 0;
int total = 0;
while (reader.HasNext())
{
total++;
var event1 = reader.GetNextEvent();
var rawFeature = event1.Feature.ToList();
if (!rawFeature[(int)Event.Field.dbpediaTypes].Equals("UNKNOW"))
{
count++;
}
}
reader.Close();
writer.WriteLine(Path.GetFileNameWithoutExtension(sourceFiles[i]) + "\t" + count + "\t" + (1.0 * count / total));
}
writer.Close();
}
private void OutputTfIdf()
{
var writer = new LargeFileWriter(vectorPath, FileMode.Create);
int docLabel = 1;
int num = 0;
ReadOneDoc();
while (this.doc != null)
{
var tokenizer = TokenizerPool.GetTokenizer();
var document = tokenizer.Tokenize(doc);
TokenizerPool.ReturnTokenizer(tokenizer);
if (++num % 1000 == 0)
{
Console.WriteLine(num);
}
var vector = GetTfIdf(document);
writer.Write(docLabel);
foreach (var value in vector)
{
writer.Write("\t" + value.first + ":" + value.second);
}
writer.Write("\r");
ReadOneDoc();
docLabel++;
}
writer.Close();
}
/// <summary>
/// Extract features for bayes model
/// </summary>
/// <param name="source">
/// File path storing the data from which this program extract features.
/// </param>
/// <param name="des">
/// File path to store the extracted features.
/// </param>
private static void ExtractBayesFeature(string source, string des)
{
FileReader reader = new LargeFileReader(source);
FileWriter writer = new LargeFileWriter(des, FileMode.Create);
var lines = reader.ReadAllLines().ToList();
const int numPerThread = 10000;
var threadNum = (int)Math.Ceiling(1.0 * lines.Count / numPerThread);
var childThreads = new Thread[threadNum];
var tmpFiles = new string[threadNum];
for (var i = 0; i < threadNum; i++)
{
tmpFiles[i] = "./tmp" + i + ".txt";
var threadClass = new BayesFeatureThread(lines.GetRange(numPerThread * i, Math.Min(numPerThread, lines.Count - numPerThread * i)), tmpFiles[i]);
childThreads[i] = new Thread(threadClass.ThreadMain);
childThreads[i].Name = "thread " + i;
childThreads[i].Start();
}
for (var i = 0; i < threadNum; i++)
{
childThreads[i].Join();
}
foreach (var tmpFile in tmpFiles)
{
var text = File.ReadAllText(tmpFile);
writer.Write(text);
File.Delete(tmpFile);
}
}
// save word and cluster id pairs
private void SaveWordClusterId()
{
var writer = new LargeFileWriter(wordClusterIDFile, FileMode.Create);
for (int i = 0; i < words.Count; i++)
{
writer.WriteLine(words[i] + "\t" + labels[i]);
}
writer.Close();
}
private void SaveWordTable()
{
var writer = new LargeFileWriter(wordTablePath, FileMode.Create);
foreach (var word in wordTable.Keys)
{
writer.WriteLine(word + "\t" + wordTable[word]);
}
writer.Close();
}
private void SaveDf()
{
var writer = new LargeFileWriter(dfPath, FileMode.Create);
foreach (var word in df.Keys)
{
writer.WriteLine(word + "\t" + df[word]);
}
writer.Close();
}
public void ThreadMain()
{
FileWriter writer = new LargeFileWriter(this.des, FileMode.Create);
var count = 0;
foreach (var line in this.lines)
{
if ((++count) % 1000 == 0)
{
Console.WriteLine(Thread.CurrentThread.Name + " has processed " + count);
}
try
{
var feature = ExtractBayesFeature(line);
writer.Cache(feature.first);
var dic = feature.second;
foreach (var field in dic.Keys)
{
writer.Cache("\t" + field + ":{");
var values = dic[field];
if (values is IEnumerable && !(values is string))
{
var begin = true;
foreach (var value in (IEnumerable)values)
{
if (begin)
{
writer.Cache(value);
begin = false;
}
else
{
writer.Cache("," + value);
}
}
writer.Cache("}");
}
else
{
writer.Cache(values + "}");
}
}
writer.Cache("\r");
writer.WriteCache();
}
catch (Exception e)
{
Console.WriteLine(line);
Console.WriteLine(e.Message);
writer.ClearCache();
}
}
writer.Close();
}
public static void Analyse()
{
// String basedir = @"E:\Users\v-mipeng\Codes\C#\NLP\Fine-ner\input";
String reportFile = @" E:\Users\v-mipeng\Codes\C#\NLP\Fine-ner\output\report.txt";
FileWriter writer = new LargeFileWriter(reportFile, FileMode.Append);
//// statistic train data number by type and coverage of UIUC by type
//String dicCovReport = Statistic.StatisticDicCoverage(@"E:\Users\v-mipeng\Codes\C#\NLP\Fine-ner\input\dictionary\UIUC.txt", basedir + @"\train\train.txt");
//writer.WriteLine("Coverage of UIUC within train data:\r"+dicCovReport);
//writer.WriteLine("");
//basedir = @"E:\Users\v-mipeng\Codes\C#\NLP\Fine-ner\input\satori";
//// statistic satori develop data number by type and coverage of UIUC by type
//dicCovReport = Statistic.StatisticDicCoverage(@"E:\Users\v-mipeng\Codes\C#\NLP\Fine-ner\input\dictionary\UIUC.txt", basedir + @"\develop.txt");
//writer.WriteLine("Coverage of UIUC within satori develop data:\r" + dicCovReport);
//writer.WriteLine("");
//// statistic satori test data number by type and coverage of UIUC by type
//dicCovReport = Statistic.StatisticDicCoverage(@"E:\Users\v-mipeng\Codes\C#\NLP\Fine-ner\input\dictionary\UIUC.txt", basedir + @"\test.txt");
//writer.WriteLine("Coverage of UIUC within satori test data:\r" + dicCovReport);
//writer.WriteLine("");
//basedir = @"E:\Users\v-mipeng\Codes\C#\NLP\Fine-ner\input\satori_lc";
//// statistic satori_lc develop data number by type and coverage of UIUC by type
//dicCovReport = Statistic.StatisticDicCoverage(@"E:\Users\v-mipeng\Codes\C#\NLP\Fine-ner\input\dictionary\UIUC.txt", basedir + @"\develop.txt");
//writer.WriteLine("Coverage of UIUC within satori_lc develop data:\r" + dicCovReport);
//writer.WriteLine("");
//// statistic satori_lc test data number by type and coverage of UIUC by type
//dicCovReport = Statistic.StatisticDicCoverage(@"E:\Users\v-mipeng\Codes\C#\NLP\Fine-ner\input\dictionary\UIUC.txt", basedir + @"\test.txt");
//writer.WriteLine("Coverage of UIUC within satori_lc test data:\r" + dicCovReport);
//writer.WriteLine("");
string basedir = @"E:\Users\v-mipeng\Codes\C#\NLP\Fine-ner\input";
// // statistic co-occurrence rate between train and satori develop data
//string corate = Statistic.StatisticCooccurrence(basedir + @"\train\train.txt", basedir + "/satori/develop.txt");
//writer.WriteLine("Co-occurrence rate between train and satori develop data is:\r "+corate);
//writer.WriteLine("");
//// statistic co-occurrence rate between train and satori test data
//corate = Statistic.StatisticCooccurrence(basedir + @"\train\train.txt", basedir + "/satori/test.txt");
//writer.WriteLine("Co-occurrence rate between train and satori test data is:\r " + corate);
//writer.WriteLine("");
//// statistic co-occurrence rate between train and satori_lc develop data
//corate = Statistic.StatisticCooccurrence(basedir + @"\train\train.txt", basedir + "/satori_lc/develop.txt");
//writer.WriteLine("Co-occurrence rate between train and satori_lc develop data is:\r " + corate);
//writer.WriteLine("");
//// statistic co-occurrence rate between train and satori_lc test data
//corate = Statistic.StatisticCooccurrence(basedir + @"\train\train.txt", basedir + "/satori_lc/test.txt");
//writer.WriteLine("co-occurrence rate between train and satori_lc test data is:\r " + corate);
//statistic name list coverage
//String report = Statistic.StatisticNameListCoverageByType(@"E:\Users\v-mipeng\Codes\C#\NLP\Fine-ner\input\dictionary\name-list.txt", @"E:\Users\v-mipeng\Codes\C#\NLP\Fine-ner\input\train\limited train.txt");
//writer.WriteLine("Name list coverage by type is :\r " + report);
// statistic item number by type
//writer.WriteLine("Item number by type:\r" + Statistic.StatisticItemNumberByType(basedir + @"\train\train.txt"));
//writer.WriteLine(Statistic.StatisticRoundTokenInformation(basedir + @"\train\train.txt"));
//writer.WriteLine(Statistic.StatisticWithinTokenInfomation(basedir + @"\train\train.txt"));
writer.Close();
}
/* Train file format:
* Mention Type Context
* Extract word table and word shape table from train data
* Every word is converted to lowercase and stemmed
* /************************************************************************/
public void ExtractWordTable()
{
FileReader reader = new LargeFileReader((string)GlobalParameter.Get(DefaultParameter.Field.train_data_file));
FileWriter writer = new LargeFileWriter((string)GlobalParameter.Get(DefaultParameter.Field.word_table_file), FileMode.Create);
FileWriter wordShapeWriter = new LargeFileWriter((string)GlobalParameter.Get(DefaultParameter.Field.word_shape_table_file), FileMode.Create);
//FileWriter wordShapeWriter = new LargeFileWriter("../../../Fine-ner/input/shape-table-file.txt", FileMode.Create);
string line = null;
var wordTable = new HashSet <string>();
var wordShapeTable = new HashSet <string>();
while ((line = reader.ReadLine()) != null)
{
try
{
var array = line.Split('\t');
var tokenizer = TokenizerPool.GetTokenizer();
var words = tokenizer.Tokenize(array[2]);
TokenizerPool.ReturnTokenizer(tokenizer);
foreach (var w in words)
{
if (!string.IsNullOrEmpty(w)) // w should not be empty
{
var shape = Feature.GetWordShape(w);
if (!wordShapeTable.Contains(shape))
{
wordShapeWriter.WriteLine(shape);
wordShapeTable.Add(shape);
}
var word = Generalizer.Generalize(w);
if (!wordTable.Contains(word))
{
writer.WriteLine(word);
wordTable.Add(word);
}
}
}
}
catch (Exception e)
{
Console.WriteLine("=================error!===============");
Console.WriteLine("\t" + e.Message);
Console.WriteLine("\t" + e.StackTrace);
Console.WriteLine("=================error!===============");
continue;
}
}
reader.Close();
writer.Close();
}
// save the centroid of clusters
private void SaveCentroids()
{
var writer = new LargeFileWriter(centroidInfoFile, FileMode.Create);
//foreach (var centroid in kmeans.Clusters.Centroids)
foreach (var centroid in kmeans.Centroids)
{
foreach (var value in centroid)
{
writer.Write(string.Format("{0}\t", value));
}
writer.Write("\r");
}
writer.Close();
}
public static void ExtractUIUC()
{
string source = @"E:\Users\v-mipeng\Data\Dictionary\name-list.freq.txt";
FileReader reader = new LargeFileReader(source);
string des = @"E:\Users\v-mipeng\Codes\C#\NLP\Fine-ner\data\name-list.txt";
FileWriter writer = new LargeFileWriter(des, FileMode.OpenOrCreate);
String line;
String[] array;
while ((line = reader.ReadLine()) != null)
{
array = line.Split('\t');
writer.WriteLine(array[0]);
}
}
private void SaveKeyWords(Dictionary <string, double> keyWordDic, string des)
{
var writer = new LargeFileWriter(des, FileMode.Create);
int count = 0;
foreach (var item in keyWordDic.OrderByDescending(key => key.Value))
{
count++;
writer.WriteLine(item.Key + "\t" + item.Value);
if (count > 1000)
{
break;
}
}
writer.Close();
}
/// <summary>
/// Combine files given by sourceFiles into one file given by desFile
/// </summary>
/// <param name="sourceFiles">
/// Source file pathes to be combined
/// </param>
/// <param name="desFile">
/// The file path to store the combined file
/// </param>
public static void CombineFiles(IEnumerable <string> sourceFiles, string desFile)
{
var reader = new LargeFileReader();
var writer = new LargeFileWriter(desFile, FileMode.Create);
string line;
foreach (var file in sourceFiles)
{
reader.Open(file);
while ((line = reader.ReadLine()) != null)
{
writer.WriteLine(line);
}
}
reader.Close();
writer.Close();
}
private void OutputDicTypeValue()
{
var dic = DataCenter.GetDicTyeMap();
var writer = new LargeFileWriter((string)GlobalParameter.Get(DefaultParameter.Field.dic_type_value_file), FileMode.OpenOrCreate);
foreach (var key in dic.Keys)
{
if (GlobalParameter.featureNum != 0)
{
writer.WriteLine(key + "\t" + (GlobalParameter.featureNum - DataCenter.GetDicTypeNum() + dic[key]));
}
else
{
writer.WriteLine(key + "\t" + dic[key]);
}
}
writer.Close();
}
/// <summary>
/// Refine disambiguations file download from dbpedia
/// </summary>
/// <param name="sourceFile"></param>
/// <param name="desFile"></param>
public static void RefineAmbiguousItem(string sourceFile, string desFile)
{
var reader = new LargeFileReader(sourceFile);
var writer = new LargeFileWriter(desFile, System.IO.FileMode.Create);
var line = "";
System.Text.RegularExpressions.Regex sourceRegex = new System.Text.RegularExpressions.Regex(@"/([^/>]+)>");
System.Text.RegularExpressions.Regex deleteBraceRegex = new System.Text.RegularExpressions.Regex(@"_?\([^\)]+\)");
System.Text.RegularExpressions.Regex desRegex = new System.Text.RegularExpressions.Regex(@"/([^/>]+)>\s\.$");
var dic = new Dictionary <string, List <string> >(300000);
List <string> list = null;
reader.ReadLine();
while ((line = reader.ReadLine()) != null)
{
var sourceMatch = sourceRegex.Match(line);
var source = sourceMatch.Groups[1].Value;
source = deleteBraceRegex.Replace(source, "");
var desMatch = desRegex.Match(line);
if (dic.TryGetValue(source, out list))
{
list.Add(desMatch.Groups[1].Value);
}
else
{
list = new List <string>();
list.Add(desMatch.Groups[1].Value);
dic[source] = list;
}
}
reader.Close();
foreach (var item in dic)
{
writer.Write(item.Key);
foreach (var des in item.Value)
{
writer.Write("\t" + des);
}
writer.WriteLine("");
}
writer.Close();
}
public void Train()
{
this.statisticModel = BayesTest.LoadModel(@"D:\Codes\C#\EntityTyping\Fine-ner\unit test\output\model.txt");
this.developFeatures = LoadBayesData(@"D:\Codes\C#\EntityTyping\Fine-ner\unit test\output\developFeature.txt");
this.labels = GetLabels().ToList();
this.w = new Dictionary <string, double>(this.labels.Count);
foreach (var label in labels) // Initial weight vector
{
w[label] = 1;
}
var learnSpeed = 0.005;
var lastPositive = -1;
this.positive = -1;
int steps = 0;
var lastWeight = new Dictionary <string, double>(w);
while (lastPositive == -1 || ((this.positive - lastPositive) > 0))
{
steps++;
//learnSpeed /= steps;
lastPositive = this.positive;
var diff = GetDiff();
lastWeight = new Dictionary <string, double>(w);
foreach (var label in labels)
{
w[label] -= diff[label] * learnSpeed;
}
Console.WriteLine("positive: " + positive);
foreach (var label in labels)
{
Console.Write(label + "\t" + w[label] + "\t");
}
Console.WriteLine("");
}
FileWriter writer = new LargeFileWriter(@"D:\Codes\C#\EntityTyping\Fine-ner\unit test\output\weight.txt", FileMode.Create);
foreach (var label in labels)
{
writer.WriteLine(label + "\t" + lastWeight[label]);
}
writer.Close();
}
public static void Temp6()
{
var sourceFile = @"D:\Codes\Project\EntityTyping\Fine-ner\output\conll feature\raw\train.txt";
var desFile = @"D:\Codes\Project\EntityTyping\Fine-ner\output\conll feature\raw\train data in dbpedia info.txt";
var writer = new LargeFileWriter(desFile, FileMode.Create);
var coverNumByType = new Dictionary <string, int>();
var totals = new Dictionary <string, int>();
var reader = new EventReaderByLine(sourceFile);
while (reader.HasNext())
{
var event1 = reader.GetNextEvent();
var rawFeature = event1.Feature.ToList();
try
{
totals[event1.Label.ToString()] += 1;
}
catch (Exception)
{
totals[event1.Label.ToString()] = 1;
}
if (!rawFeature[(int)Event.Field.dbpediaTypes].Equals("UNKNOW"))
{
try
{
coverNumByType[event1.Label.ToString()] += 1;
}
catch (Exception)
{
coverNumByType[event1.Label.ToString()] = 1;
}
}
}
reader.Close();
foreach (var type in totals.Keys)
{
writer.WriteLine(type + "\t" + coverNumByType[type] + "\t" + totals[type] + "\t" + (1.0 * coverNumByType[type] / totals[type]));
}
writer.Close();
}
public static void SelectInterestWordVector(string interestWordFile, string word2vecFile, string compressedWord2VectorFile)
{
var reader = new LargeFileReader(interestWordFile);
string line;
var set = new HashSet <string>();
while ((line = reader.ReadLine()) != null)
{
set.Add(line.Trim());
}
reader.Close();
var writer = new LargeFileWriter(compressedWord2VectorFile, FileMode.Create);
var parser = new ParseBinaryVector(word2vecFile);
int count = 0;
while (!parser.EOF)
{
if (++count % 1000 == 0)
{
Console.WriteLine(count);
}
try
{
var pair = parser.GetNextVector();
if (set.Contains(pair.first))
{
writer.Write(pair.first);
foreach (var value in pair.second)
{
writer.Write(string.Format(" {0}", value));
}
writer.Write("\r");
}
}
catch (Exception)
{
continue;
}
}
writer.Close();
}
public static void Temp()
{
var source = @"D:\Data\DBpedia\mapping based types";
var desDir = "";
var dic = new Dictionary <string, FileWriter>();
var reader = new pml.file.reader.LargeFileReader(source);
var des = @"D:\Data\DBpedia\entity type pairs.txt";
var writer = new LargeFileWriter(des, FileMode.Create);
string line;
System.Text.RegularExpressions.Regex entityRegex = new System.Text.RegularExpressions.Regex(@"/([^>/]+)>\s<");
System.Text.RegularExpressions.Regex typeRegex = new System.Text.RegularExpressions.Regex(@"ontology/(\w+)>\s\.$");
int count = 0;
while ((line = reader.ReadLine()) != null)
{
string entity = null;
string type = null;
if (entityRegex.IsMatch(line))
{
var match = entityRegex.Match(line);
entity = match.Groups[1].Value;
}
if (typeRegex.IsMatch(line))
{
var match = typeRegex.Match(line);
type = match.Groups[1].Value;
}
if (entity != null && type != null)
{
if (++count % 10000 == 0)
{
Console.WriteLine(count);
}
writer.WriteLine(entity + "\t" + type);
}
}
reader.Close();
writer.Close();
}
/// <summary>
/// Store the trained model into file
/// </summary>
/// <param name="desPath"></param>
/// <param name="array"></param>
/// <format>
/// [Label]
/// [field name]
/// TAB [feature annotation,e.g., last word] TAB [times]
/// TAB [feature annotation,e.g., last word] TAB [times]
/// ###END###
/// </format>
internal static void OutputModel(string desPath, object model)
{
var writer = new LargeFileWriter(desPath, FileMode.Create);
var dics = (Dictionary <string, Dictionary <string, Dictionary <string, int> > >)model;
foreach (var label in dics.Keys) // Check !
{
var dic = dics[label]; // fields-->dic<feature value, times>
foreach (var field in dic.Keys)
{
writer.WriteLine(label);
writer.WriteLine(field); // write field
var featureDic = dic[field];
var keys = Feature.SortKeysByNum(featureDic);
foreach (var featureValue in keys)
{
writer.WriteLine("\t" + featureValue + "\t" + featureDic[featureValue]);
}
}
}
writer.Close();
}
public static void GetItemNumByType()
{
var sourceDir = @"D:\Codes\Project\EntityTyping\Fine-ner\input\feature\test";
var desFile = @"D:\Codes\Project\EntityTyping\Fine-ner\input\feature\data info.txt";
var sourceFiles = Directory.GetFiles(sourceDir).ToList();
var writer = new LargeFileWriter(desFile, FileMode.Append);
writer.WriteLine(sourceDir.Substring(sourceDir.LastIndexOf("\\") + 1) + ":");
for (var i = 0; i < sourceFiles.Count; i++)
{
var reader = new LargeFileReader(sourceFiles[i]);
var line = "";
int count = 0;
while ((line = reader.ReadLine()) != null)
{
count++;
}
reader.Close();
writer.WriteLine(Path.GetFileNameWithoutExtension(sourceFiles[i]) + "\t:\t" + count);
}
writer.Close();
}
public static void Temp3()
{
var source = @"D:\Data\DBpedia\redirects.ttl";
var reader = new pml.file.reader.LargeFileReader(source);
var des = @"D:\Data\DBpedia\redirects.txt";
var writer = new LargeFileWriter(des, FileMode.Create);
string line;
System.Text.RegularExpressions.Regex firstRegex = new System.Text.RegularExpressions.Regex(@"/([^>/]+)>\s<");
System.Text.RegularExpressions.Regex secondRegex = new System.Text.RegularExpressions.Regex(@"/(\w+)>\s\.$");
int count = 0;
while ((line = reader.ReadLine()) != null)
{
string first = null;
string second = null;
if (firstRegex.IsMatch(line))
{
var match = firstRegex.Match(line);
first = match.Groups[1].Value;
}
if (secondRegex.IsMatch(line))
{
var match = secondRegex.Match(line);
second = match.Groups[1].Value;
}
if (first != null && second != null)
{
if (++count % 10000 == 0)
{
Console.WriteLine(count);
}
writer.WriteLine(first + "\t" + second);
}
}
reader.Close();
writer.Close();
}
public static void Temp()
{
if (false)
{
var reader = new LargeFileReader(@"E:\Users\v-mipeng\Codes\C#\NLP\Fine-ner\input\train\train.txt");
var writer = new LargeFileWriter(@"E:\Users\v-mipeng\Codes\C#\NLP\Fine-ner\input\train\limited train.txt", FileMode.OpenOrCreate);
Dictionary <string, int> numByType = new Dictionary <string, int>(16);
String line;
String[] array;
int count = 0;
int num = 0;
while ((line = reader.ReadLine()) != null)
{
count++;
if (count % 1000 == 0)
{
Console.Error.WriteLine(count + " items processed!");
}
array = line.Split('\t');
try
{
num = numByType[array[1]];
}
catch (Exception)
{
num = 0;
}
if (num > 100000) // do not limit train data number by type
{
continue;
}
writer.WriteLine(line);
numByType[array[1]] = ++num;
}
reader.Close();
writer.Close();
}
if (false)
{
string result = @"E:\Users\v-mipeng\Codes\C#\NLP\Fine-ner\output\result\satori_lc\-1.inst.txt";
string source = @"E:\Users\v-mipeng\Codes\C#\NLP\Fine-ner\input\satori_lc\develop.txt";
String tmpFile = "./tmp.txt";
FileReader reader1 = new LargeFileReader(result);
FileReader reader2 = new LargeFileReader(source);
FileWriter writer = new LargeFileWriter(tmpFile, FileMode.OpenOrCreate);
String line;
String line2;
writer.WriteLine(reader1.ReadLine());
while ((line = reader1.ReadLine()) != null)
{
line2 = reader2.ReadLine();
writer.WriteLine(line2.Split('\t')[0] + "\t" + line.Split(new char[] { '\t' }, 2)[1]);
}
reader1.Close();
reader2.Close();
writer.Close();
File.Copy(tmpFile, @"E:\Users\v-mipeng\Codes\C#\NLP\Fine-ner\output\result\satori_lc\.inst.txt");
File.Delete(tmpFile);
}
if (false)
{
string wordTableFile = @"E:\Users\v-mipeng\Codes\C#\NLP\Fine-ner\input\word table\wordTable.txt";
FileReader reader = new LargeFileReader(wordTableFile);
FileWriter writer = new LargeFileWriter();
HashSet <string> wordSet = new HashSet <string>();
string line;
while ((line = reader.ReadLine()) != null)
{
//var stemmer = StemmerPool.GetStemmer();
//wordSet.Add(stemmer.Stem(line.Split('\t')[0])[0]);
//StemmerPool.ReturnStemmer(stemmer);
//stemmer = null;
}
reader.Close();
writer.Open(wordTableFile);
int i = 0;
foreach (String word in wordSet)
{
writer.WriteLine(word + '\t' + (i++));
}
writer.Close();
}
if (false)
{
String dir = @"E:\Users\v-mipeng\Codes\C#\NLP\Fine-ner\input\dictionary\names";
string des = @"E:\Users\v-mipeng\Codes\C#\NLP\Fine-ner\input\dictionary\name-all.txt";
FileReader reader = new LargeFileReader();
FileWriter writer = new LargeFileWriter(des, FileMode.Create);
string[] files = Directory.GetFiles(dir, "*.txt");
string line;
foreach (String file in files)
{
reader.Open(file);
while ((line = reader.ReadLine()) != null)
{
writer.WriteLine(line.Split(',')[0]);
}
}
reader.Close();
writer.Close();
}
if (false)
{
string path1 = @"E:\Users\v-mipeng\Codes\C#\NLP\Fine-ner\input\dictionary\name-all.txt";
string path2 = @"E:\Users\v-mipeng\Codes\C#\NLP\Fine-ner\input\train\limited train.txt";
string des = @"E:\Users\v-mipeng\Codes\C#\NLP\Fine-ner\input\dictionary\tmp.txt";
FileReader reader = new LargeFileReader(path1);
FileWriter writer = new LargeFileWriter(des);
String line;
HashSet <String> set = new HashSet <string>();
String[] array;
while ((line = reader.ReadLine()) != null)
{
set.Add(line);
array = line.Split(' ');
}
reader.Close();
reader.Open(path2);
while ((line = reader.ReadLine()) != null)
{
array = line.Split('\t');
if (set.Contains(array[0].ToLower()))
{
if (!array[1].Equals("people.person"))
{
set.Remove(array[0].ToLower());
}
}
}
reader.Close();
foreach (String name in set)
{
writer.WriteLine(name);
}
writer.Close();
}
if (false)
{
FileReader reader = new LargeFileReader(@"E:\Users\v-mipeng\Codes\C#\NLP\Fine-ner\input\satori_lc\backup\version 1-2\develop.txt");
FileWriter writer = new LargeFileWriter(@"E:\Users\v-mipeng\Codes\C#\NLP\Fine-ner\input\satori_lc\develop.txt", FileMode.OpenOrCreate);
String line;
string[] array;
HashSet <string> interestTypes = new HashSet <string>();
interestTypes.Add("people.person");
interestTypes.Add("location.location");
interestTypes.Add("organization.organization");
while ((line = reader.ReadLine()) != null)
{
array = line.Split('\t');
if (interestTypes.Contains(array[1]))
{
writer.WriteLine(line);
}
}
reader.Close();
writer.Close();
}
if (false)
{
FileReader reader = new LargeFileReader(@"E:\Users\v-mipeng\Codes\C#\NLP\Fine-ner\output\tmp.txt");
FileWriter writer = new LargeFileWriter(@"E:\Users\v-mipeng\Codes\C#\NLP\Fine-ner\output\tmp2.txt", FileMode.Create);
String line;
string[] array;
string[] pairString;
List <Pair <string, int> > list = new List <Pair <string, int> >();
Pair <string, int> pair = new Pair <string, int>();
Comparer <Pair <string, int> > comparer = pair.GetBySecondReverseComparer();
while ((line = reader.ReadLine()) != null)
{
array = line.Split('\t');
for (int i = 1; i < array.Length; i++)
{
pairString = new string[] { array[i].Substring(0, array[i].LastIndexOf(":")), array[i].Substring(array[i].LastIndexOf(":") + 1) };
pair = new Pair <string, int>();
pair.first = pairString[0];
pair.second = int.Parse(pairString[1]);
list.Add(pair);
}
list.Sort(comparer);
foreach (Pair <string, int> item in list)
{
writer.Write("\t" + item.first + ":" + item.second);
}
writer.Write("\r");
list.Clear();
}
reader.Close();
writer.Close();
}
if (true)
{
FileReader reader = new LargeFileReader(@"E:\Users\v-mipeng\Codes\C#\NLP\Fine-ner\output\tmp.txt");
FileWriter writer = new LargeFileWriter(@"E:\Users\v-mipeng\Codes\C#\NLP\Fine-ner\output\tmp2.txt", FileMode.Create);
String line;
string[] lines = new string[3];
string[] array;
string[] pairString;
Dictionary <string, int>[] dics = new Dictionary <string, int> [3];
List <Pair <string, int> > list = new List <Pair <string, int> >();
Pair <string, int> pair = new Pair <string, int>();
Comparer <Pair <string, int> > comparer = pair.GetBySecondReverseComparer();
for (int i = 0; i < 3; i++)
{
line = reader.ReadLine();
array = line.Split('\t');
dics[i] = new Dictionary <string, int>();
if (i == 0)
{
for (int j = 1; j < array.Length; j++)
{
pairString = new string[] { array[j].Substring(0, array[j].LastIndexOf(":")), array[j].Substring(array[j].LastIndexOf(":") + 1) };
pair = new Pair <string, int>();
pair.first = pairString[0];
pair.second = int.Parse(pairString[1]);
dics[i][pair.first] = pair.second;
list.Add(pair);
}
}
else
{
for (int j = 1; j < array.Length; j++)
{
pairString = new string[] { array[j].Substring(0, array[j].LastIndexOf(":")), array[j].Substring(array[j].LastIndexOf(":") + 1) };
dics[i][pairString[0]] = int.Parse(pairString[1]);
}
}
}
list.Sort(comparer);
int count = 10;
int locNum;
int orgNum;
foreach (Pair <string, int> item in list)
{
count++;
try
{
locNum = dics[1][item.first];
}
catch (Exception)
{
locNum = 0;
}
try
{
orgNum = dics[2][item.first];
}catch (Exception)
{
orgNum = 0;
}
writer.Write("\t" + item.first + ":(" + item.second + "|" + locNum + "|" + orgNum + ")");
if (count % 5 == 0)
{
writer.Write("\r");
}
}
reader.Close();
writer.Close();
}
}
/// <summary>
/// map types
/// </summary>
public static void Temp2()
{
var dbpediaToSatoriDic = new Dictionary <string, Dictionary <string, int> >();
var satoriMentionDic = new Dictionary <string, string>();
var satoriEntityDic = new Dictionary <string, string>();
var dbpedia = @"D:\Data\DBpedia\entity type pairs.txt";
var satori = @"D:\Codes\C#\EntityTyping\Fine-ner\input\feature\train.txt";
var des = @"D:\Codes\C#\EntityTyping\Fine-ner\input\db2satori.txt";
var dbpediaReader = new LargeFileReader(dbpedia);
var satoriReader = new LargeFileReader(satori);
var writer = new LargeFileWriter(des, FileMode.Create);
var line = "";
Dictionary <string, int> dic = null;
while ((line = satoriReader.ReadLine()) != null)
{
var array = line.Split('\t');
satoriMentionDic[array[0]] = array[2];
satoriEntityDic[array[1]] = array[2];
}
satoriReader.Close();
System.Text.RegularExpressions.Regex regex = new System.Text.RegularExpressions.Regex(@"\s+");
System.Text.RegularExpressions.Regex deleteBrace = new System.Text.RegularExpressions.Regex(@"\(\w+\)");
var count = 0;
while ((line = dbpediaReader.ReadLine()) != null)
{
if (++count % 10000 == 0)
{
Console.WriteLine(count);
}
var array = line.Split('\t');
var entity = deleteBrace.Replace(array[0], "");
entity = regex.Replace(entity, " ").Trim();
if (satoriEntityDic.ContainsKey(entity))
{
try
{
dic = dbpediaToSatoriDic[array[1]];
}
catch (Exception)
{
dic = new Dictionary <string, int>();
dbpediaToSatoriDic[array[1]] = dic;
}
try
{
dic[satoriEntityDic[entity]] += 1;
}
catch (Exception)
{
dic[satoriEntityDic[entity]] = 1;
}
}
else if (satoriMentionDic.ContainsKey(entity))
{
try
{
dic = dbpediaToSatoriDic[array[1]];
}
catch (Exception)
{
dic = new Dictionary <string, int>();
dbpediaToSatoriDic[array[1]] = dic;
}
try
{
dic[satoriMentionDic[entity]] += 1;
}
catch (Exception)
{
dic[satoriMentionDic[entity]] = 1;
}
}
}
dbpediaReader.Close();
foreach (var item in dbpediaToSatoriDic)
{
foreach (var d in item.Value)
{
writer.WriteLine(item.Key + "\t" + d.Key + "\t" + d.Value);
}
}
writer.Close();
}
public void SplitData()
{
var sourceDic = LoadTotalNumByType();
var mentionNumDic = new Dictionary <string, int>();
var uniqueMentionNumDic = new Dictionary <string, HashSet <string> >();
// create reader by file
var files = Directory.GetFiles(this.sourceDir);
var reader = new LargeFileReader();
// create file path to store train, develop and test data
var trainFiles = new List <string>();
var devFiles = new List <string>();
var testFiles = new List <string>();
foreach (var file in files)
{
trainFiles.Add(Path.Combine(trainDir, Path.GetFileName(file)));
devFiles.Add(Path.Combine(developDir, Path.GetFileName(file)));
testFiles.Add(Path.Combine(testDir, Path.GetFileName(file)));
}
var writers = new List <FileWriter>();
// random value generator to seperate develop and test data
var random = new Random();
string line;
string[] array;
int num = 0;
int trainNumLimit = 500000;
HashSet <string> set = null;
int count = 0;
int limitMentionNumPerEntity = 10;
int numByEntity = 0;
int devNumLimit = 4000;
int i = 0;
files = new string[] { @"E:\Users\v-mipeng\Codes\Projects\EntityTyping\Fine-ner\input\satori\refined-satori\time_event.txt" };
trainFiles.Clear();
trainFiles.Add(@"E:\Users\v-mipeng\Codes\Projects\EntityTyping\Fine-ner\input\satori\train\time_event.txt");
devFiles.Clear();
devFiles.Add(@"E:\Users\v-mipeng\Codes\Projects\EntityTyping\Fine-ner\input\satori\develop\time_event.txt");
testFiles.Clear();
testFiles.Add(@"E:\Users\v-mipeng\Codes\Projects\EntityTyping\Fine-ner\input\satori\test\time_event.txt");
foreach (var file in files)
{
reader.Open(file);
string lastEntity = "";
writers.Clear();
writers.Add(new LargeFileWriter(devFiles[i], FileMode.Create));
writers.Add(new LargeFileWriter(testFiles[i], FileMode.Create));
var trainWriter = new LargeFileWriter(trainFiles[i], FileMode.Create);
i++;
int devOrTestNum = 0;
while ((line = reader.ReadLine()) != null)
{
if (++count % 10000 == 0)
{
Console.WriteLine(count);
}
array = line.Split('\t');
if (array[1].Equals(lastEntity))
{
numByEntity++;
}
else
{
numByEntity = 1;
lastEntity = array[1];
}
if (numByEntity > limitMentionNumPerEntity)
{
continue;
}
mentionNumDic.TryGetValue(array[2], out num);
if (num < trainNumLimit && num < 0.8 * sourceDic[array[2]] / limitMentionNumPerEntity)
{
SaveForTrain(trainWriter, line);
mentionNumDic[array[2]] = num + 1;
uniqueMentionNumDic.TryGetValue(array[2], out set);
if (set == null)
{
set = uniqueMentionNumDic[array[2]] = new HashSet <string>();
}
if (!set.Contains(array[0]))
{
set.Add(array[0]);
}
}
else if (devOrTestNum < devNumLimit * 2)
{
devOrTestNum++;
SaveForDevOrTest(writers[random.Next(0, 2)], line);
}
}
reader.Close();
trainWriter.Close();
writers[0].Close();
writers[1].Close();
}
var writer = new LargeFileWriter(statisticInfoFile, FileMode.Create);
foreach (var key in mentionNumDic.Keys)
{
writer.WriteLine(key + "\t" + mentionNumDic[key]);
writer.WriteLine(key + "\t" + uniqueMentionNumDic[key].Count);
}
writer.Close();
foreach (var file in trainFiles)
{
File.SetAttributes(file, FileAttributes.ReadOnly);
}
foreach (var file in devFiles)
{
File.SetAttributes(file, FileAttributes.ReadOnly);
}
foreach (var file in testFiles)
{
File.SetAttributes(file, FileAttributes.ReadOnly);
}
}
public void EvaluateResult(string resultFile, string evaluationFile)
{
var reader = new LargeFileReader(resultFile);
var line = "";
var result = new Dictionary <string, Dictionary <string, int> >(); // class-->(predicted class --> number)
int times = 0;
var trueLabelIndex = 1;
var predictLabelIndex = 2;
var writer = new LargeFileWriter(evaluationFile, FileMode.Create);
Dictionary <string, int> dic = null;
line = reader.ReadLine();
while ((line = reader.ReadLine()) != null)
{
var array = line.Split('\t');
try
{
dic = result[array[trueLabelIndex]];
try
{
times = dic[array[predictLabelIndex]];
dic[array[predictLabelIndex]] = times + 1;
}
catch (Exception)
{
dic[array[predictLabelIndex]] = 1;
}
}
catch (Exception)
{
dic = new Dictionary <string, int>();
dic[array[2]] = 1;
result[array[trueLabelIndex]] = dic;
}
}
reader.Close();
writer.Write("True|Predict");
var keys = result.Keys;
foreach (var key in keys)
{
writer.Write("\t" + key);
}
writer.WriteLine("");
foreach (var key in keys)
{
writer.Write(key);
var info = result[key];
foreach (var k in keys)
{
if (info.TryGetValue(k, out times))
{
writer.Write("\t" + times);
}
else
{
writer.Write("\t" + 0);
}
}
writer.WriteLine("");
}
var macroPre = Util.GetMacroPrecision(result);
var macroRec = Util.GetMacroRecall(result);
var macroF1 = Util.GetF1(macroPre, macroRec);
writer.WriteLine("macro-precision: " + macroPre);
writer.WriteLine("macro-recall : " + macroRec);
writer.WriteLine("macro-F1 : " + macroF1);
var microPre = Util.GetMicroPrecision(result);
writer.WriteLine("micro-precision: " + microPre);
writer.Close();
}
