diff --git a/index.bat b/index.bat
index 07d4531c..f5bf71b9 100644
--- a/index.bat
+++ b/index.bat
@@ -1 +1 @@
-sir.bat indexwikipedia --directory C:\projects\resin\src\Sir.HttpServer\AppData\database --collection wikipedia --skip 0 --take 1000000 --pageSize 10000 --sampleSize 1000 %*
\ No newline at end of file
+sir.bat indexwikipedia --directory C:\projects\resin\src\Sir.HttpServer\AppData\database --collection wikipedia --skip 0 --take 1000000 --pageSize 1000 --sampleSize 1000 %*
\ No newline at end of file
diff --git a/src/Sir.Cmd/AnalyzeDocumentCommand.cs b/src/Sir.Cmd/AnalyzeDocumentCommand.cs
index 29caa739..d3445c1b 100644
--- a/src/Sir.Cmd/AnalyzeDocumentCommand.cs
+++ b/src/Sir.Cmd/AnalyzeDocumentCommand.cs
@@ -18,7 +18,6 @@ public void Run(IDictionary<string, string> args, ILogger logger)
             var select = new HashSet<string>(args["select"].Split(new char[] { ',', ' ' }, StringSplitOptions.RemoveEmptyEntries));
             var collectionId = collection.ToHash();
             var model = new BagOfCharsModel();
-            var embedding = new SortedList<int, float>();
 
             using (var documentReader = new DocumentRegistryReader(dataDirectory, collectionId))
             {
@@ -26,7 +25,7 @@ public void Run(IDictionary<string, string> args, ILogger logger)
 
                 foreach (var field in doc.Fields)
                 {
-                    var tokens = model.CreateEmbedding(field.Value.ToString(), true, embedding);
+                    var tokens = model.CreateEmbedding(field.Value.ToString(), true);
                     var tree = new VectorNode();
 
                     foreach (var token in tokens)
diff --git a/src/Sir.Cmd/BenchmarkCommand.cs b/src/Sir.Cmd/BenchmarkCommand.cs
index 6578324a..4a0f2201 100644
--- a/src/Sir.Cmd/BenchmarkCommand.cs
+++ b/src/Sir.Cmd/BenchmarkCommand.cs
@@ -30,13 +30,12 @@ public void RunTokenizeBenchmark(IDictionary<string, string> args, ILogger logge
             var model = new BagOfCharsModel();
             var documents = new List<Document>(WikipediaHelper.Read(fileName, skip, take, new HashSet<string> { "text" }));
             var timer = Stopwatch.StartNew();
-            var embedding = new SortedList<int, float>();
 
             for (int i = 0; i < numOfRuns; i++)
             {
                 foreach (var document in documents)
                 {
-                    var embeddings = new List<ISerializableVector>(model.CreateEmbedding((string)document.Fields[0].Value, false, embedding));
+                    var embeddings = new List<ISerializableVector>(model.CreateEmbedding((string)document.Fields[0].Value, false));
                 }
             }
 
diff --git a/src/Sir.Cmd/ValidateCommand.cs b/src/Sir.Cmd/ValidateCommand.cs
index 99898183..870388d5 100644
--- a/src/Sir.Cmd/ValidateCommand.cs
+++ b/src/Sir.Cmd/ValidateCommand.cs
@@ -22,13 +22,12 @@ public void Run(IDictionary<string, string> args, ILogger logger)
             var selectFields = new HashSet<string> { "title" };
             var time = Stopwatch.StartNew();
             var count = 0;
-            var embedding = new SortedList<int, float>();
 
             using (var kvReader = new KeyValue.KeyValueReader(dir, collectionId))
             using (var validateSession = new ValidateSession<string>(
                     collectionId,
                     new SearchSession<string>(dir, model, new LogStructuredIndexingStrategy(model), logger),
-                    new QueryParser<string>(kvReader, model, embedding: embedding, logger: logger)))
+                    new QueryParser<string>(kvReader, model, logger: logger)))
             using (var documents = new DocumentStreamSession(dir))
             {
                 foreach (var doc in documents.ReadDocuments<string>(collectionId, selectFields, skip, take))
diff --git a/src/Sir.ImageTests/ImageModelTests.cs b/src/Sir.ImageTests/ImageModelTests.cs
index 6fdb96c3..763bda88 100644
--- a/src/Sir.ImageTests/ImageModelTests.cs
+++ b/src/Sir.ImageTests/ImageModelTests.cs
@@ -103,7 +103,7 @@ public void Can_traverse_streamed()
                                     throw new Exception($"unable to find {word} in tree.");
                                 }
 
-                                if (hit.Score < model.IdenticalAngle)
+                                if (hit.Score.Approximates(model.IdenticalAngle))
                                 {
                                     throw new Exception($"unable to score {word}.");
                                 }
diff --git a/src/Sir.Images/LinearClassifierImageModel.cs b/src/Sir.Images/LinearClassifierImageModel.cs
index c1b60205..a4e024ab 100644
--- a/src/Sir.Images/LinearClassifierImageModel.cs
+++ b/src/Sir.Images/LinearClassifierImageModel.cs
@@ -11,7 +11,7 @@ public class LinearClassifierImageModel : Sir.DistanceCalculator, IModel<IImage>
         public double FoldAngle => 0.75d;
         public override int NumOfDimensions => 784; 
 
-        public IEnumerable<ISerializableVector> CreateEmbedding(IImage data, bool label, SortedList<int, float> embedding = null)
+        public IEnumerable<ISerializableVector> CreateEmbedding(IImage data, bool label)
         {
             var pixels = data.Pixels.Select(x => Convert.ToSingle(x));
 
diff --git a/src/Sir.InformationRetreival/IModel.cs b/src/Sir.InformationRetreival/IModel.cs
index 20a7c4b6..49fa64fb 100644
--- a/src/Sir.InformationRetreival/IModel.cs
+++ b/src/Sir.InformationRetreival/IModel.cs
@@ -9,7 +9,7 @@ namespace Sir
     /// <typeparam name="T">The type of data the model should consist of.</typeparam>
     public interface IModel<T> : IModel
     {
-        IEnumerable<ISerializableVector> CreateEmbedding(T data, bool label, SortedList<int, float> embedding = null);
+        IEnumerable<ISerializableVector> CreateEmbedding(T data, bool label);
     }
 
     /// <summary>
diff --git a/src/Sir.InformationRetreival/IO/GraphBuilder.cs b/src/Sir.InformationRetreival/IO/GraphBuilder.cs
index 90a4b2cd..fd9c7bf4 100644
--- a/src/Sir.InformationRetreival/IO/GraphBuilder.cs
+++ b/src/Sir.InformationRetreival/IO/GraphBuilder.cs
@@ -9,11 +9,10 @@ public static class GraphBuilder
         public static VectorNode CreateTree<T>(this IModel<T> model, IIndexReadWriteStrategy indexingStrategy, params T[] data)
         {
             var root = new VectorNode();
-            var embedding = new SortedList<int, float>();
 
             foreach (var item in data)
             {
-                foreach (var vector in model.CreateEmbedding(item, true, embedding))
+                foreach (var vector in model.CreateEmbedding(item, true))
                 {
                     indexingStrategy.Put<T>(root, new VectorNode(vector));
                 }
diff --git a/src/Sir.InformationRetreival/IndexingStrategies/LogStructuredIndexingStrategy.cs b/src/Sir.InformationRetreival/IndexingStrategies/LogStructuredIndexingStrategy.cs
index 95bc4416..a1094efd 100644
--- a/src/Sir.InformationRetreival/IndexingStrategies/LogStructuredIndexingStrategy.cs
+++ b/src/Sir.InformationRetreival/IndexingStrategies/LogStructuredIndexingStrategy.cs
@@ -29,7 +29,7 @@ public void SerializePage(string directory, ulong collectionId, long keyId, Vect
             var time = Stopwatch.StartNew();
 
             using (var vectorStream = StreamFactory.CreateAppendStream(directory, collectionId, keyId, "vec"))
-            using (var postingsWriter = new PostingsWriter(StreamFactory.CreateSeekableWriteStream(directory, collectionId, keyId, "pos"), indexCache:null))
+            using (var postingsWriter = new PostingsWriter(StreamFactory.CreateSeekableWriteStream(directory, collectionId, keyId, "pos"), indexCache: indexCache))
             using (var columnWriter = new ColumnWriter(StreamFactory.CreateAppendStream(directory, collectionId, keyId, "ix")))
             using (var pageIndexWriter = new PageIndexWriter(StreamFactory.CreateAppendStream(directory, collectionId, keyId, "ixtp")))
             {
diff --git a/src/Sir.InformationRetreival/Parsers/QueryParser.cs b/src/Sir.InformationRetreival/Parsers/QueryParser.cs
index 081f4705..d5765d2c 100644
--- a/src/Sir.InformationRetreival/Parsers/QueryParser.cs
+++ b/src/Sir.InformationRetreival/Parsers/QueryParser.cs
@@ -10,14 +10,12 @@ public class QueryParser<T>
         private readonly KeyValueReader _kvReader;
         private readonly IModel<T> _model;
         private readonly ILogger _logger;
-        private readonly SortedList<int, float> _embedding;
 
-        public QueryParser(KeyValueReader kvReader, IModel<T> model, SortedList<int, float> embedding = null, ILogger logger = null)
+        public QueryParser(KeyValueReader kvReader, IModel<T> model, ILogger logger = null)
         {
             _kvReader = kvReader;
             _model = model;
             _logger = logger;
-            _embedding = embedding ?? new SortedList<int, float>();
         }
 
         public Query Parse(
@@ -243,7 +241,7 @@ private IList<Term> CreateTerms(ulong collectionId, string key, T value, bool an
 
             if (_kvReader.TryGetKeyId(key.ToHash(), out keyId))
             {
-                var tokens = _model.CreateEmbedding(value, label, _embedding);
+                var tokens = _model.CreateEmbedding(value, label);
 
                 foreach (var term in tokens)
                 {
diff --git a/src/Sir.InformationRetreival/SerializableVector.cs b/src/Sir.InformationRetreival/SerializableVector.cs
index 5334bb2a..d1ed6028 100644
--- a/src/Sir.InformationRetreival/SerializableVector.cs
+++ b/src/Sir.InformationRetreival/SerializableVector.cs
@@ -17,6 +17,10 @@ public class SerializableVector : ISerializableVector
         public int[] Indices { get { return ((SparseVectorStorage<float>)Value.Storage).Indices; } }
         public float[] Values { get { return ((SparseVectorStorage<float>)Value.Storage).Values; } }
 
+        public SerializableVector()
+        {
+        }
+
         public SerializableVector(int numOfDimensions, object label = null)
         {
             Value = CreateVector.Sparse<float>(numOfDimensions);
@@ -48,6 +52,11 @@ public SerializableVector(SortedList<int, float> dictionary, int numOfDimensions
             Label = label;
         }
 
+        public bool IsEmptyVector()
+        {
+            return Value == null;
+        }
+
         public SerializableVector(int[] index, float[] values, int numOfDimensions, object label = null)
         {
             var tuples = new Tuple<int, float>[Math.Min(index.Length, numOfDimensions)];
diff --git a/src/Sir.InformationRetreival/Session/DocumentDatabase.cs b/src/Sir.InformationRetreival/Session/DocumentDatabase.cs
index 373cb568..63e354d1 100644
--- a/src/Sir.InformationRetreival/Session/DocumentDatabase.cs
+++ b/src/Sir.InformationRetreival/Session/DocumentDatabase.cs
@@ -39,7 +39,7 @@ public DocumentDatabase(string directory, ulong collectionId, IModel<T> model =
 
         public QueryParser<T> CreateQueryParser()
         {
-            return new QueryParser<T>(SearchSession.GetKeyValueReader(_collectionId), _model, IndexSession.EmptyEmbedding, _logger);
+            return new QueryParser<T>(SearchSession.GetKeyValueReader(_collectionId), _model, _logger);
         }
 
         public IEnumerable<Document> StreamDocuments(HashSet<string> fieldsOfInterest, int skip, int take)
diff --git a/src/Sir.InformationRetreival/Session/DocumentStreamSession.cs b/src/Sir.InformationRetreival/Session/DocumentStreamSession.cs
index 40466a99..448e62a8 100644
--- a/src/Sir.InformationRetreival/Session/DocumentStreamSession.cs
+++ b/src/Sir.InformationRetreival/Session/DocumentStreamSession.cs
@@ -201,12 +201,8 @@ public IEnumerable<VectorNode> ReadDocumentValuesAsVectors<T>(
             HashSet<string> select,
             DocumentRegistryReader documentReader,
             IModel<T> model,
-            bool label,
-            SortedList<int, float> embedding = null)
+            bool label)
         {
-            if (embedding == null)
-                embedding = new SortedList<int, float>();
-
             var docInfo = documentReader.GetDocumentAddress(doc.docId);
             var docMap = documentReader.GetDocumentMap(docInfo.offset, docInfo.length);
 
@@ -222,7 +218,7 @@ public IEnumerable<VectorNode> ReadDocumentValuesAsVectors<T>(
                 {
                     var vInfo = documentReader.GetAddressOfValue(kvp.valId);
 
-                    foreach (var vector in documentReader.GetValueConvertedToVectors<T>(vInfo.offset, vInfo.len, vInfo.dataType, value => model.CreateEmbedding(value, label, embedding)))
+                    foreach (var vector in documentReader.GetValueConvertedToVectors<T>(vInfo.offset, vInfo.len, vInfo.dataType, value => model.CreateEmbedding(value, label)))
                     {
                         tree.AddIfUnique(new VectorNode(vector, docId:doc.docId, keyId:kvp.keyId), model);
                     }
diff --git a/src/Sir.InformationRetreival/Session/IIndexSession.cs b/src/Sir.InformationRetreival/Session/IIndexSession.cs
deleted file mode 100644
index 7d2f385b..00000000
--- a/src/Sir.InformationRetreival/Session/IIndexSession.cs
+++ /dev/null
@@ -1,19 +0,0 @@
-using System.Collections.Generic;
-
-namespace Sir
-{
-    public interface IIndexSession<T> : IIndexSession
-    {
-        void Put(long docId, long keyId, T value, bool label);
-    }
-
-    public interface IIndexSession
-    {
-        IndexInfo GetIndexInfo();
-        void Put(long docId, long keyId, IEnumerable<ISerializableVector> tokens);
-        void Put(VectorNode documentTree);
-        void Commit();
-        void Commit(long keyId);
-        IDictionary<long, VectorNode> GetInMemoryIndices();
-    }
-}
\ No newline at end of file
diff --git a/src/Sir.InformationRetreival/Session/IndexCache.cs b/src/Sir.InformationRetreival/Session/IndexCache.cs
index 2adb4a7d..c27672bd 100644
--- a/src/Sir.InformationRetreival/Session/IndexCache.cs
+++ b/src/Sir.InformationRetreival/Session/IndexCache.cs
@@ -35,7 +35,7 @@ public void Put(VectorNode node)
             {
                 var hit = PathFinder.ClosestMatch(tree, vector, _model);
 
-                if (hit.Score >= _model.IdenticalAngle)
+                if (hit.Score.Approximates(_model.IdenticalAngle))
                 {
                     return hit.Node.PostingsOffset == -1 ? null : hit.Node.PostingsOffset;
                 }
@@ -49,7 +49,7 @@ public void UpdatePostingsOffset(long keyId, ISerializableVector vector, long po
             {
                 var hit = PathFinder.ClosestMatch(tree, vector, _model);
 
-                if (hit.Score >= _model.IdenticalAngle)
+                if (hit.Score.Approximates(_model.IdenticalAngle))
                 {
                     hit.Node.PostingsOffset = postingsOffset;
                 }
diff --git a/src/Sir.InformationRetreival/Session/IndexDebugger.cs b/src/Sir.InformationRetreival/Session/IndexDebugger.cs
index 50bafab8..f1e01bbb 100644
--- a/src/Sir.InformationRetreival/Session/IndexDebugger.cs
+++ b/src/Sir.InformationRetreival/Session/IndexDebugger.cs
@@ -24,7 +24,7 @@ public IndexDebugger(ILogger logger, int sampleSize = 1000)
             _logger = logger;
         }
 
-        public void Step(IIndexSession indexSession, string message = null)
+        public void Step<T>(IndexSession<T> indexSession, string message = null)
         {
             _steps++;
 
diff --git a/src/Sir.InformationRetreival/Session/IndexSession.cs b/src/Sir.InformationRetreival/Session/IndexSession.cs
index 77eb1834..a53b519d 100644
--- a/src/Sir.InformationRetreival/Session/IndexSession.cs
+++ b/src/Sir.InformationRetreival/Session/IndexSession.cs
@@ -1,5 +1,4 @@
 using Microsoft.Extensions.Logging;
-using Sir.IO;
 using System;
 using System.Collections.Generic;
 using System.Diagnostics;
@@ -10,7 +9,7 @@ namespace Sir
     /// Write a paged index.
     /// </summary>
     /// <typeparam name="T"></typeparam>
-    public class IndexSession<T> : IIndexSession<T>, IDisposable
+    public class IndexSession<T> :  IDisposable
     {
         private readonly IModel<T> _model;
         private readonly IIndexReadWriteStrategy _indexingStrategy;
@@ -20,8 +19,6 @@ public class IndexSession<T> : IIndexSession<T>, IDisposable
         private readonly ILogger _logger;
         private readonly IndexCache _indexCache;
 
-        public SortedList<int, float> EmptyEmbedding = new SortedList<int, float>();
-
         public IndexSession(
             string directory,
             ulong collectionId,
@@ -41,12 +38,12 @@ public IndexSession(
 
         public void Put(long docId, long keyId, T value, bool label)
         {
-            var tokens = _model.CreateEmbedding(value, label, EmptyEmbedding);
+            var tokens = _model.CreateEmbedding(value, label);
 
             Put(docId, keyId, tokens);
         }
 
-        public void Put(long docId, long keyId, IEnumerable<ISerializableVector> tokens)
+        private void Put(long docId, long keyId, IEnumerable<ISerializableVector> tokens)
         {
             VectorNode column;
 
@@ -58,27 +55,10 @@ public void Put(long docId, long keyId, IEnumerable<ISerializableVector> tokens)
 
             foreach (var token in tokens)
             {
-                _indexingStrategy.Put<T>(
-                                    column,
-                                    new VectorNode(vector:token, docId:docId, keyId:keyId));
-            }
-        }
-
-        public void Put(VectorNode token)
-        {
-            VectorNode column;
-
-            if (!_index.TryGetValue(token.KeyId.Value, out column))
-            {
-                column = new VectorNode();
-                _index.Add(token.KeyId.Value, column);
-            }
-
-            foreach (var node in PathFinder.All(token))
-            {
-                _indexingStrategy.Put<T>(
-                    column,
-                    new VectorNode(node.Vector, docIds: node.DocIds));
+                if (!token.IsEmptyVector())
+                    _indexingStrategy.Put<T>(
+                                        column,
+                                        new VectorNode(vector:token, docId:docId, keyId:keyId));
             }
         }
 
diff --git a/src/Sir.InformationRetreival/Session/SearchSession.cs b/src/Sir.InformationRetreival/Session/SearchSession.cs
index caa32729..318a06c6 100644
--- a/src/Sir.InformationRetreival/Session/SearchSession.cs
+++ b/src/Sir.InformationRetreival/Session/SearchSession.cs
@@ -156,7 +156,7 @@ private void Scan(Query query, bool identicalMatchesOnly)
 
                     if (hit != null)
                     {
-                        if (!identicalMatchesOnly || (hit.Score >= _model.IdenticalAngle))
+                        if (!identicalMatchesOnly || hit.Score.Approximates(_model.IdenticalAngle))
                         {
                             term.Score = hit.Score;
                             term.PostingsOffsets = hit.PostingsOffsets;
diff --git a/src/Sir.KeyValue/ISerializableVector.cs b/src/Sir.KeyValue/ISerializableVector.cs
index bf4f55fe..cf33e3c3 100644
--- a/src/Sir.KeyValue/ISerializableVector.cs
+++ b/src/Sir.KeyValue/ISerializableVector.cs
@@ -20,5 +20,6 @@ public interface ISerializableVector
         void AverageInPlace(ISerializableVector vector);
         ISerializableVector Append(ISerializableVector vector);
         ISerializableVector Shift(int numOfPositionsToShift, int numOfDimensions);
+        bool IsEmptyVector();
     }
 }
\ No newline at end of file
diff --git a/src/Sir.Strings/BagOfCharsModel.cs b/src/Sir.Strings/BagOfCharsModel.cs
index 0f212874..cc641f96 100644
--- a/src/Sir.Strings/BagOfCharsModel.cs
+++ b/src/Sir.Strings/BagOfCharsModel.cs
@@ -7,17 +7,15 @@ public class BagOfCharsModel : DistanceCalculator, IModel<string>
         public double IdenticalAngle => 0.998d;
         public double FoldAngle => 0.55d;
         public override int NumOfDimensions => System.Text.Unicode.UnicodeRanges.All.Length;
+        private readonly SortedList<int, float> _embedding = new SortedList<int, float>();
 
-        public IEnumerable<ISerializableVector> CreateEmbedding(string data, bool label, SortedList<int, float> embedding = null)
+        public IEnumerable<ISerializableVector> CreateEmbedding(string data, bool label)
         {
             var source = data.ToCharArray();
 
             if (source.Length > 0)
             {
-                if (embedding == null)
-                    embedding = new SortedList<int, float>();
-                else
-                    embedding.Clear();
+                _embedding.Clear();
 
                 var offset = 0;
                 int index = 0;
@@ -28,20 +26,20 @@ public IEnumerable<ISerializableVector> CreateEmbedding(string data, bool label,
 
                     if (char.IsLetterOrDigit(c) || char.GetUnicodeCategory(c) == System.Globalization.UnicodeCategory.MathSymbol)
                     {
-                        embedding.AddOrAppendToComponent(c, 1);
+                        _embedding.AddOrAppendToComponent(c, 1);
                     }
                     else
                     {
-                        if (embedding.Count > 0)
+                        if (_embedding.Count > 0)
                         {
                             var len = index - offset;
 
                             var vector = new SerializableVector(
-                                embedding,
+                                _embedding,
                                 NumOfDimensions,
                                 label ? new string(source, offset, len) : null);
 
-                            embedding.Clear();
+                            _embedding.Clear();
                             yield return vector;
                         }
 
@@ -49,12 +47,12 @@ public IEnumerable<ISerializableVector> CreateEmbedding(string data, bool label,
                     }
                 }
 
-                if (embedding.Count > 0)
+                if (_embedding.Count > 0)
                 {
                     var len = index - offset;
 
                     var vector = new SerializableVector(
-                                embedding,
+                                _embedding,
                                 NumOfDimensions,
                                 label ? new string(source, offset, len) : null);
 
diff --git a/src/Sir.Strings/BagOfWordsModel.cs b/src/Sir.Strings/BagOfWordsModel.cs
new file mode 100644
index 00000000..de4ae1f2
--- /dev/null
+++ b/src/Sir.Strings/BagOfWordsModel.cs
@@ -0,0 +1,68 @@
+using System.Collections.Generic;
+
+namespace Sir.Strings
+{
+    public class BagOfWordsModel : DistanceCalculator, IModel<string>
+    {
+        public double IdenticalAngle => 0.998d;
+        public double FoldAngle => 0.55d;
+        public override int NumOfDimensions => System.Text.Unicode.UnicodeRanges.All.Length;
+        private readonly SortedList<int, float> _embedding = new SortedList<int, float>();
+
+        public IEnumerable<ISerializableVector> CreateEmbedding(string data, bool label)
+        {
+            var source = data.ToCharArray();
+
+            if (source.Length > 0)
+            {
+                _embedding.Clear();
+
+                var offset = 0;
+                int index = 0;
+
+                for (; index < source.Length; index++)
+                {
+                    char c = char.ToLower(source[index]);
+
+                    if (char.IsPunctuation(c))
+                    {
+                        yield return new SerializableVector();
+                    }
+                    else if (char.IsLetterOrDigit(c) || char.GetUnicodeCategory(c) == System.Globalization.UnicodeCategory.MathSymbol)
+                    {
+                        _embedding.AddOrAppendToComponent(c, 1);
+                    }
+                    else
+                    {
+                        if (_embedding.Count > 0)
+                        {
+                            var len = index - offset;
+
+                            var vector = new SerializableVector(
+                                _embedding,
+                                NumOfDimensions,
+                                label ? new string(source, offset, len) : null);
+
+                            _embedding.Clear();
+                            yield return vector;
+                        }
+
+                        offset = index + 1;
+                    }
+                }
+
+                if (_embedding.Count > 0)
+                {
+                    var len = index - offset;
+
+                    var vector = new SerializableVector(
+                                _embedding,
+                                NumOfDimensions,
+                                label ? new string(source, offset, len) : null);
+
+                    yield return vector;
+                }
+            }
+        }
+    }
+}
\ No newline at end of file
diff --git a/src/Sir.Strings/NGramModel.cs b/src/Sir.Strings/NGramModel.cs
index 699fa9f2..34e467fa 100644
--- a/src/Sir.Strings/NGramModel.cs
+++ b/src/Sir.Strings/NGramModel.cs
@@ -17,18 +17,13 @@ public NGramModel(BagOfCharsModel wordTokenizer)
             NumOfDimensions = wordTokenizer.NumOfDimensions * 2;
         }
 
-        public IEnumerable<ISerializableVector> CreateEmbedding(string data, bool label, SortedList<int, float> embedding = null)
+        public IEnumerable<ISerializableVector> CreateEmbedding(string data, bool label)
         {
-            if (embedding == null)
-                embedding = new SortedList<int, float>();
-            else
-                embedding.Clear();
-
             ISerializableVector vec0 = null;
 
             var i = 0;
 
-            foreach (var token in _wordTokenizer.CreateEmbedding(data, label, embedding))
+            foreach (var token in _wordTokenizer.CreateEmbedding(data, label))
             {
                 if (vec0 == null)
                 {
diff --git a/write.bat b/write.bat
index 4568d69f..5db8472a 100644
--- a/write.bat
+++ b/write.bat
@@ -1 +1 @@
-sir.bat writewikipedia --directory C:\projects\resin\src\Sir.HttpServer\AppData\database --file d:\enwiki-20211122-cirrussearch-content.json.gz --collection wikipedia --skip 0 --take 1000000 --sampleSize 1000 %*
\ No newline at end of file
+sir.bat writewikipedia --directory C:\projects\resin\src\Sir.HttpServer\AppData\database --file d:\enwiki-20211122-cirrussearch-content.json.gz --collection wikipedia --skip 0 --take 10000 --sampleSize 1000 %*
\ No newline at end of file