private static void TestForAttachments(string text, Dictionary <string, int> words)
{
if (Regex.IsMatch(text, "Content-Disposition:.*attachment", RegexOptions.Compiled))
{
ExtraFeatures.AddOrIncrementWords(words, "!Attachment", 1);
}
}
private static bool TestForIsEdu(string line, Dictionary <string, int> words)
{
// From a .edu email address
if (Regex.IsMatch(line, "^From:.*@.*edu>", RegexOptions.Compiled))
{
ExtraFeatures.AddOrIncrementWords(words, "!FromEdu", 1);
return(true);
}
return(false);
}
private static bool TestForIsReply(string line, Dictionary <string, int> words)
{
// -----Original Message-----
if (Regex.IsMatch(line, "-----Original Message-----", RegexOptions.Compiled))
{
ExtraFeatures.AddOrIncrementWords(words, "!IsReply", 1);
return(true);
}
return(false);
}
private static bool TestForKnownSender(string line, Dictionary <string, int> words)
{
// From: "Some Friend"
if (Regex.IsMatch(line, "^From:", RegexOptions.Compiled) && !Regex.IsMatch(line, "@", RegexOptions.Compiled))
{
ExtraFeatures.AddOrIncrementWords(words, "!KnownSender", 1);
return(true);
}
return(false);
}
private static void TestForCatchPhrases(string text, Dictionary <string, int> words)
{
if (Regex.IsMatch(text, "free money", RegexOptions.IgnoreCase & RegexOptions.Compiled))
{
ExtraFeatures.AddOrIncrementWords(words, "!CatchPhrase", 1);
}
if (Regex.IsMatch(text, "only $", RegexOptions.IgnoreCase & RegexOptions.Compiled))
{
ExtraFeatures.AddOrIncrementWords(words, "!CatchPhrase", 1);
}
if (Regex.IsMatch(text, "over 21", RegexOptions.IgnoreCase & RegexOptions.Compiled))
{
ExtraFeatures.AddOrIncrementWords(words, "!CatchPhrase", 1);
}
}
private void LoadTestingData(string testingSetPath)
{
Trace.TraceInformation("Loading testing data from {0}", testingSetPath);
this.Documents = new Dictionary <string, Document>();
int progress = 0;
using (StreamReader sr = File.OpenText(testingSetPath))
{
string s = String.Empty;
while ((s = sr.ReadLine()) != null)
{
string[] parts = s.Split(NB.Delimiter);
string id = parts[0];
string value = parts[1];
if (!this.Documents.ContainsKey(id))
{
this.Documents.Add(id, new Document(id, value));
}
if (this.UseExtraFeatures)
{
Trace.TraceInformation("Finding extra features on file {0}", progress++.ToString());
ExtraFeatures.AddExtraFeatures(id, this.Documents[id].Words);
}
// Add the words to the Document
for (int i = 2; i < parts.Length; i = i + 2)
{
string word = parts[i];
int wordCount = Convert.ToInt32(parts[i + 1]);
// Add to word list
this.Documents[id].Words.Add(word, wordCount);
}
}
}
Trace.TraceInformation("Documents: {0} documents", this.Documents.Keys.Count);
Trace.TraceInformation("Done loading testing data");
Trace.TraceInformation("");
}
// Steps:
// 1. Read in the raw file
// 2. Look for extra features
// 3. Add the extra features in the words dictionary
// The dictionary key will start with a ! followed by the feature name
public static void AddExtraFeatures(string fileId, Dictionary <string, int> words)
{
string dataFolder = ConfigurationManager.AppSettings["DataFolder"];
if (String.IsNullOrWhiteSpace(dataFolder))
{
Trace.TraceWarning("DataFolder is not set in the configuration. Skipping extra features");
return;
}
string filePath = Path.Combine(dataFolder, fileId.Substring(1)); // Ignore the leading '/' on the fileId
if (!File.Exists(filePath))
{
Trace.TraceWarning("File {0} was not found. Skipping.", filePath);
return;
}
Dictionary <string, int> features = new Dictionary <string, int>();
StringBuilder fullText = new StringBuilder();
using (StreamReader sr = File.OpenText(filePath))
{
string s = String.Empty;
while ((s = sr.ReadLine()) != null)
{
fullText.AppendLine(s);
ExtraFeatures.TestForKnownSender(s, words);
ExtraFeatures.TestForIsReply(s, words);
ExtraFeatures.TestForIsEdu(s, words);
}
}
ExtraFeatures.TestForCatchPhrases(fullText.ToString(), words);
ExtraFeatures.TestForAttachments(fullText.ToString(), words);
}
private void LoadTrainingData(string trainingSetPath)
{
Trace.TraceInformation("Loading training data from {0}", trainingSetPath);
this.Vocabulary = new Vocabulary();
this.Targets = new Dictionary <string, Target>();
int progress = 0;
using (StreamReader sr = File.OpenText(trainingSetPath))
{
string s = String.Empty;
while ((s = sr.ReadLine()) != null)
{
string[] parts = s.Split(NB.Delimiter);
string id = parts[0];
string value = parts[1];
if (!this.Targets.ContainsKey(value))
{
this.Targets.Add(value, new Target(value));
}
if (this.UseExtraFeatures)
{
Trace.TraceInformation("Finding extra features on file {0}", progress++.ToString());
ExtraFeatures.AddExtraFeatures(id, this.Targets[value].Words);
ExtraFeatures.AddExtraFeatures(id, this.Vocabulary.Words);
}
this.Vocabulary.ExampleCount++;
this.Targets[value].DocumentCount++;
// Add the words to the Target
for (int i = 2; i < parts.Length; i = i + 2)
{
string word = parts[i];
int wordCount = Convert.ToInt32(parts[i + 1]);
// Add to the vocabulary list
if (this.Vocabulary.Words.ContainsKey(word))
{
this.Vocabulary.Words[word] += wordCount;
}
else
{
this.Vocabulary.Words.Add(word, wordCount);
}
// Add to the spam or ham list
if (this.Targets[value].Words.ContainsKey(word))
{
this.Targets[value].Words[word] += wordCount;
}
else
{
this.Targets[value].Words.Add(word, wordCount);
}
}
}
}
Trace.TraceInformation("Vocabulary: {0} unique words, {1} total words, {2} documents", this.Vocabulary.Words.Keys.Count, this.Vocabulary.Words.Values.Sum(), this.Vocabulary.ExampleCount);
Trace.TraceInformation("Spam: {0} unique words, {1} total words, {2} documents", this.Targets["spam"].Words.Keys.Count, this.Targets["spam"].Words.Values.Sum(), this.Targets["spam"].DocumentCount);
Trace.TraceInformation("Ham: {0} unique words, {1} total words, {2} documents", this.Targets["ham"].Words.Keys.Count, this.Targets["ham"].Words.Values.Sum(), this.Targets["ham"].DocumentCount);
Trace.TraceInformation("Done loading training data");
Trace.TraceInformation("");
}
