public static double GetWordLikelihood(Model model, Queue <string> evidence, string word)
{
List <double> weights = getWeights();
List <double> likelihoods = new List <double>();
nWeights = new List <double>();
while (evidence.Count >= 0)
{
Gram g = model.getGramFromChain(evidence);
double temp = 0;
int total = g.getCount();
if (total > 0)
{
int occurences = g[word].getCount();
temp = occurences / (double)total;
}
likelihoods.Add(temp);
nWeights.Add(weights[evidence.Count]);
if (evidence.Count == 0)
{
break;
}
evidence.Dequeue();
}
double likelihood = 0;
for (int i = 0; i < likelihoods.Count; i++)
{
likelihood += likelihoods[i] * nWeights[i];
}
return(likelihood * 100);
}
private static Dictionary <string, double> LaplacianSmoothing(Gram predicateState, List <string> dictionary) //this list dictionary may contain words not in the official dictionary such as a misspelling of currentword
{
Dictionary <string, double> probabilityDistribution = new Dictionary <string, double>();
int predicateFrequency = predicateState.getCount();
predicateFrequency += dictionary.Count; // the plus one smoothing
foreach (string word in dictionary)
{
int count = 1; // the plus one smoothing
count += predicateState.NextWordCount(word);
probabilityDistribution.Add(word, count / (double)predicateFrequency);
}
return(probabilityDistribution);
}