Exemple #1
0
        public void AddCorrection(MisspelledWord error)
        {
            string context = ExportContext ? ";" + error.GetLeftContext().ToStringRepresentation() + ";" + error.GetRightContext().ToStringRepresentation() : "";
            writer.WriteLine(error.WrongWord + ";" + error.CorrectWord + ";" + error.RevokedByLm.ToString() + context);

            if (!String.IsNullOrEmpty(error.CorrectWord))
            {
                writerCorrected.WriteLine(error.WrongWord + ";" + error.CorrectWord + ";" + Math.Round(error.Accuracy,1).ToString() + ";" + error.CorrectedBy.ToString() +";" + error.IsName().ToString());
            }
        }
Exemple #2
0
        public LanguageModelEvaluation EvaluateCandidates(MisspelledWord word, Dictionary<string, double> candidates)
        {
            foundInNgrams = false;
            List<string> leftContext = word.GetLeftContext();

            NgramType type = this.dictionary.GetHighestAvailableNgramCollection(leftContext.Count);

            Dictionary<string, double> probability = new Dictionary<string, double>();
            string[] lcArray = this.GetLeftContext(leftContext, type);
            NgramEvaluation evaluation;
            foreach (KeyValuePair<string, double> option in candidates)
            {
                lcArray[leftContext.Count - 1] = option.Key.Contains(' ') ? option.Key.Split(space).First() : option.Key;

                evaluation = this.dictionary.GetNgramCollection(type).GetProbability(lcArray);
                probability.Add(option.Key, evaluation.Probability);

                if (!foundInNgrams && evaluation.Occurence > 0)
                {
                    foundInNgrams = true;
                }
            }

            List<string> rightContext = word.GetRightContext();
            NgramType secType = this.dictionary.GetHighestAvailableNgramCollection(rightContext.Count);

            if (type == NgramType.Unigram && type == NgramType.Unigram)
            {
                // do nothing
            }
            else
            {
                string[] rcArray = this.GetRightContext(rightContext, secType);
                foreach (KeyValuePair<string, double> option in candidates)
                {
                    rcArray[0] = option.Key.Contains(' ') ? option.Key.Split(space).Last() : option.Key;

                    evaluation = this.dictionary.GetNgramCollection(secType).GetProbability(rcArray);
                    probability[option.Key] *= evaluation.Probability;

                    if (!foundInNgrams && evaluation.Occurence > 0)
                    {
                        foundInNgrams = true;
                    }
                }
            }

            return new LanguageModelEvaluation(probability, foundInNgrams);
        }