public int Classify(string review)
{
var tokenizer = new HappyFunTokenizer(_preserveCase);
var tokens = this.AddNegationAugments(tokenizer.Tokenize(review));
var scores = new Dictionary <int, double>();
for (var i = 1; i < 6; i++)
{
scores[i] = this.GetEmptyScoreForSentiment(i);
}
foreach (var token in tokens)
{
if (!Utilities.CheckUtf8(token))
{
continue;
}
if (!this._wordList.ContainsKey(token))
{
continue;
}
var index = this._wordList[token];
for (var i = 1; i < 6; i++)
{
var score = this.GetProbabilityOfWordGivenSentimentFast(index, i);
scores[i] *= score / (1.0 - score);
}
}
return(scores.Aggregate((l, r) => l.Value > r.Value ? l : r).Key);
}
protected int AddTrainingEntry(HappyFunTokenizer tokenizer, int score, string review)
{
var addedEntries = new HashSet <int>();
var tokens = this.AddNegationAugments(tokenizer.Tokenize(review));
var count = 0;
foreach (var token in tokens)
{
if (!Utilities.CheckUtf8(token))
{
continue;
}
if (this._wordList.ContainsKey(token))
{
var index = this._wordList[token];
if (addedEntries.Contains(index))
{
continue;
}
count++;
this._sentimentWordCounts[score][index]++;
addedEntries.Add(index);
}
else
{
var index = this._wordList.Count;
this._wordList[token] = index;
count++;
for (var i = 0; i < 6; i++)
{
this._sentimentWordCounts[i][index] = 0;
}
this._sentimentWordCounts[score][index]++;
addedEntries.Add(index);
}
}
return(count);
}
protected bool ProcessTrainingFile(StreamReader sr)
{
var tokenizer = new HappyFunTokenizer(_preserveCase);
var score = 0;
var count = 0;
var maxCount = 0;
var totalCount = 0;
string line;
string review = null;
/*
* Read training data, tokenize and store feature vectors in a sparse index to
* limit the memory usage.
*/
this._wordList = new Dictionary <string, int>();
this._sentimentCounts = new Dictionary <int, int>();
this._sentimentWordCounts = new Dictionary <int, Dictionary <int, int> >();
for (var i = 0; i < 6; i++)
{
this._sentimentCounts[i] = 0;
this._sentimentWordCounts[i] = new Dictionary <int, int>();
}
while ((line = sr.ReadLine()) != null)
{
line = line.Trim();
if (line.Length == 0)
{
if (!string.IsNullOrEmpty(review) && score > 0)
{
var tokenCount = this.AddTrainingEntry(tokenizer, score, review);
this._sentimentCounts[score]++;
count++;
totalCount += tokenCount;
if (tokenCount > maxCount)
{
maxCount = tokenCount;
}
score = 0;
review = null;
}
continue;
}
var sTest = this._regexes["score"].Match(line);
if (sTest.Success)
{
score = int.Parse(sTest.Groups["score"].Value);
continue;
}
var rTest = this._regexes["review"].Match(line);
if (!rTest.Success)
{
continue;
}
review = rTest.Groups["review"].Value;
}
if (!string.IsNullOrEmpty(review) && score > 0)
{
var tokenCount = this.AddTrainingEntry(tokenizer, score, review);
this._sentimentCounts[score]++;
count++;
totalCount += tokenCount;
if (tokenCount > maxCount)
{
maxCount = tokenCount;
}
}
this._entryCount = count;
Console.WriteLine("Bytes: {0} (Average: {1:F2})", maxCount * sizeof(int), ((double)totalCount / count) * sizeof(int));
Console.WriteLine("Found {0} words and {1} entries.", this._wordList.Count, count);
return(true);
}
