fix keyword extraction

TFIDFbee/TFIDFbee/Program.cs

@@ -2,31 +2,81 @@
 using System;
 using System.Collections.Concurrent;
 using System.Collections.Generic;
-using System.IO;
 using System.Linq;
+using System.Threading;
 using TFIDFbee.Reader;
-using ZeroLevel;
 using ZeroLevel.Services.Semantic;
 using ZeroLevel.Services.Semantic.Helpers;
-using ZeroLevel.Services.Serialization;
 
 namespace TFIDFbee
 {
+    public class IDF
+    {
+        private ConcurrentDictionary<string, int> _terms =
+            new ConcurrentDictionary<string, int>();
+        private long _documents_count = 0;
+
+        public void Learn(BagOfTerms bag)
+        {
+            _documents_count++;
+            foreach (var term in bag.ToUniqueTokens())
+            {
+                _terms.AddOrUpdate(term, 1, (w, o) => o + 1);
+            }
+        }
+
+        public double Idf(string term)
+        {
+            if (_terms.ContainsKey(term))
+            {
+                double count_documents_with_term = (double)_terms[term];
+                double total_documents = (double)_documents_count;
+                return Math.Log(1.0d + (total_documents / count_documents_with_term));
+            }
+            return 0.0d;
+        }
+    }
+
+    public static class TFIDF
+    {
+        public static IDictionary<string, double> TfIdf(BagOfTerms document, IDF idf)
+        {
+            var freg = document.Freguency();
+            return document
+                .ToUniqueTokensWithoutStopWords()
+                .ToDictionary(t => t, t => idf.Idf(t) * (double)freg[t] / (double)document.Words.Length);
+        }
+    }
+
     class Program
     {
-        private const string source = @"E:\Desktop\lenta-ru-data-set_19990901_20171204\lenta-ru-data-set_19990901_20171204_limit_1000.json";
+        private const string source = @"D:\Desktop\lenta-ru-data-set_19990901_20171204_limit_1000.json";
         private readonly static ILexProvider _lexer = new LexProvider(new LemmaLexer());
         private readonly static ConcurrentDictionary<string, double> _scoring = new ConcurrentDictionary<string, double>();
 
         static void Main(string[] args)
         {
-            var terms = new BagOfTerms("На практике эти расширения используют нечасто, особенно те расширения, которые для расчёта", _lexer);
+            IDF idf = new IDF();
-
+            IDocumentReader reader = new JsonByLineReader(source, _lexer);
-            Console.WriteLine(string.Join('-', terms.ToTokens()));
+            foreach (var batch in reader.ReadBatches(1000))
-            Console.WriteLine(string.Join('-', terms.ToUniqueTokens()));
+            {
-            Console.WriteLine(string.Join('-', terms.ToUniqueTokensWithoutStopWords()));
+                foreach (var doc in batch)
-            Console.WriteLine(string.Join('\n', terms.Freguency().Select(pair => $"{pair.Key}: {pair.Value}")));
+                {
-
+                    idf.Learn(doc);
+                }
+            }
+            foreach (var batch in reader.ReadBatches(1000))
+            {
+                foreach (var doc in batch)
+                {
+                    var tfidf = TFIDF.TfIdf(doc, idf);
+                    Console.WriteLine(String.Join(" ", tfidf.OrderByDescending(p => p.Value).Take(10).Select(p => p.Key)));
+                    Console.WriteLine();
+                    Console.WriteLine("                                     ***");
+                    Console.WriteLine();
+                    Thread.Sleep(1000);
+                }
+            }
 
 
             /*

TFIDFbee/TFIDFbee/Reader/IDocumentReader.cs

@@ -1,11 +1,10 @@
-using System;
+using System.Collections.Generic;
-using System.Collections.Generic;
+using ZeroLevel.Services.Semantic.Helpers;
 
 namespace TFIDFbee.Reader
 {
     public interface IDocumentReader
     {
-        IEnumerable<string[][]> ReadBatches(int size);
+        IEnumerable<IEnumerable<BagOfTerms>> ReadBatches(int size);
-        public IEnumerable<IEnumerable<Tuple<string, string>>> ReadRawDocumentBatches(int size);
     }
 }

TFIDFbee/TFIDFbee/Reader/JsonByLineReader.cs

@@ -2,6 +2,8 @@
 using System.Collections.Generic;
 using System.IO;
 using System.Linq;
+using ZeroLevel.Services.Semantic;
+using ZeroLevel.Services.Semantic.Helpers;
 
 namespace TFIDFbee.Reader
 {
@@ -9,78 +11,18 @@ namespace TFIDFbee.Reader
         : IDocumentReader
     {
         private readonly string _file;
-        private readonly Func<string, IEnumerable<string>> _lexer;
+        private readonly ILexProvider _lexer;
 
-        public JsonByLineReader(string file, Func<string, IEnumerable<string>> lexer)
+        public JsonByLineReader(string file, ILexProvider lexer)
         {
             _file = file;
             _lexer = lexer;
         }
 
-        public IEnumerable<string[][]> ReadBatches(int size)
+        public IEnumerable<IEnumerable<BagOfTerms>> ReadBatches(int size)
-        {
-            var list = new List<string[]>();
-            foreach (var batch in ReadDocumentBatches(size))
-            {
-                yield return batch.ToArray();
-                list.Clear();
-            }
-        }
-
-        private IEnumerable<IEnumerable<string[]>> ReadDocumentBatches(int size)
-        {
-            string line;
-            var batch = new List<string[]>();
-            string title = null;
-            string text = null;
-            using (StreamReader reader = new StreamReader(_file))
-            {
-                while ((line = reader.ReadLine()) != null)
-                {
-                    var titleIndex = line.IndexOf("\"metaTitle\":");
-                    if (titleIndex >= 0)
-                    {
-                        var start = line.IndexOf("\"", titleIndex + 12);
-                        var end = line.LastIndexOf("\"");
-                        if (start < end && start != -1 && end != -1)
-                        {
-                            title = line.Substring(start + 1, end - start - 1);
-                        }
-                    }
-                    else
-                    {
-                        var textIndex = line.IndexOf("\"plaintext\":");
-                        if (textIndex >= 0 && title != null)
-                        {
-                            var start = line.IndexOf("\"", textIndex + 12);
-                            var end = line.LastIndexOf("\"");
-                            if (start < end && start != -1 && end != -1)
-                            {
-                                text = line.Substring(start + 1, end - start - 1);
-                                batch.Add(_lexer(title).Concat(_lexer(text)).ToArray());
-                                if (batch.Count >= size)
-                                {
-                                    yield return batch;
-                                    batch.Clear();
-                                    GC.Collect(2);
-                                }
-                                title = null;
-                                text = null;
-                            }
-                        }
-                    }
-                }
-            }
-            if (batch.Count > 0)
-            {
-                yield return batch;
-            }
-        }
-
-        public IEnumerable<IEnumerable<Tuple<string, string>>> ReadRawDocumentBatches(int size)
         {
             string line;
-            var batch = new List<Tuple<string, string>>();
+            var batch = new List<BagOfTerms>();
             string title = null;
             string text = null;
             using (StreamReader reader = new StreamReader(_file))
@@ -107,7 +49,7 @@ namespace TFIDFbee.Reader
                             if (start < end && start != -1 && end != -1)
                             {
                                 text = line.Substring(start + 1, end - start - 1);
-                                batch.Add(Tuple.Create(title, text));
+                                batch.Add(new BagOfTerms(title + " " + text, _lexer));
                                 if (batch.Count >= size)
                                 {
                                     yield return batch;

TFIDFbee/TFIDFbee/Reader/StateMachineReader.cs

@@ -1,7 +1,7 @@
-using System;
+using System.Collections.Generic;
-using System.Collections.Generic;
 using System.IO;
-using System.Linq;
+using ZeroLevel.Services.Semantic;
+using ZeroLevel.Services.Semantic.Helpers;
 
 namespace TFIDFbee.Reader
 {
@@ -9,9 +9,9 @@ namespace TFIDFbee.Reader
         : IDocumentReader
     {
         private readonly string _file;
-        private readonly Func<string, IEnumerable<string>> _lexer;
+        private readonly ILexProvider _lexer;
 
-        public StateMachineReader(string file, Func<string, IEnumerable<string>> lexer)
+        public StateMachineReader(string file, ILexProvider lexer)
         {
             _file = file;
             _lexer = lexer;
@@ -46,30 +46,12 @@ namespace TFIDFbee.Reader
             }
         }
 
-        public IEnumerable<string[][]> ReadBatches(int size)
+        public IEnumerable<IEnumerable<BagOfTerms>> ReadBatches(int size)
         {
-            var list = new List<string[]>();
+            var list = new List<BagOfTerms>();
             foreach (var record in Parse())
             {
-                list.Add((_lexer(record[0]).Concat(_lexer(record[1])).ToArray()));
+                list.Add(new BagOfTerms(record[0] + " " + record[1], _lexer));
-                if (list.Count > size)
-                {
-                    yield return list.ToArray();
-                    list.Clear();
-                }
-            }
-            if (list.Count > 0)
-            {
-                yield return list.ToArray();
-            }
-        }
-
-        public IEnumerable<IEnumerable<Tuple<string, string>>> ReadRawDocumentBatches(int size)
-        {
-            var list = new List<Tuple<string, string>>();
-            foreach (var record in Parse())
-            {
-                list.Add(Tuple.Create(record[0], record[1]));
                 if (list.Count > size)
                 {
                     yield return list.ToArray();