private int compareByEnsemble(CSpellScore o1, CSpellScore o2)
{
int @out = 0;
OrthographicScore oScore1 = ((CSpellScore)o1).GetOScore();
OrthographicScore oScore2 = ((CSpellScore)o2).GetOScore();
FrequencyScore fScore1 = ((CSpellScore)o1).GetFScore();
FrequencyScore fScore2 = ((CSpellScore)o2).GetFScore();
ContextScore cScore1 = ((CSpellScore)o1).GetCScore();
ContextScore cScore2 = ((CSpellScore)o2).GetCScore();
double score1 = 0.6 * oScore1.GetScore() + 0.25 * fScore1.GetScore() + 0.15 * cScore1.GetScore();
double score2 = 0.6 * oScore2.GetScore() + 0.25 * fScore2.GetScore() + 0.15 * cScore2.GetScore();
// 1. compared by orthographic score, best
// SCR-2: use a fixed number to ensure result is not 0.
if (score2 > score1)
{
// from high to low
@out = 1;
}
else if (score2 < score1)
{
@out = -1;
}
// 2. alphabetic order
else
{
string cand1 = ((CSpellScore)o1).GetCandStr();
string cand2 = ((CSpellScore)o2).GetCandStr();
@out = cand2.CompareTo(cand1);
}
return(@out);
}
/// <summary>
/// Compare two object o1 and o2. Both objects o1 and o2 are
/// CSpellScore. The compare algorithm:
/// </summary>
/// <param name="o1"> first object to be compared </param>
/// <param name="o2"> second object to be compared
/// </param>
/// <returns> a negative integer, 0, or positive integer to represent the
/// object o1 is less, equals, or greater than object 02. </returns>
public virtual int Compare(CSpellScore o1, CSpellScore o2)
{
OrthographicScoreComparator <OrthographicScore> osc = new OrthographicScoreComparator <OrthographicScore>();
OrthographicScore oScore1 = ((CSpellScore)o1).GetOScore();
OrthographicScore oScore2 = ((CSpellScore)o2).GetOScore();
int @out = osc.Compare(oScore1, oScore2);
// the following code are simplfied by above
// We still keep this java file in case we need better implementation
// in the future
/// <summary>
/// int out = 0;
/// // 1. compared by orthographic score, best
/// OrthographicScore oScore1 = ((CSpellScore) o1).GetOScore();
/// OrthographicScore oScore2 = ((CSpellScore) o2).GetOScore();
/// if(oScore1.GetScore() != oScore2.GetScore())
/// {
/// OrthographicScoreComparator<OrthographicScore> osc
/// = new OrthographicScoreComparator<OrthographicScore>();
/// out = osc.compare(oScore1, oScore2);
/// }
/// else
/// {
/// String cand1 = ((CSpellScore) o1).GetCandStr();
/// String cand2 = ((CSpellScore) o2).GetCandStr();
/// out = cand2.compareTo(cand1);
/// }
///
/// </summary>
return(@out);
}
/// <summary>
/// Compare two object o1 and o2. Both objects o1 and o2 are
/// NoisyChannelScore. The compare algorithm:
/// </summary>
/// <param name="o1"> first object to be compared </param>
/// <param name="o2"> second object to be compared
/// </param>
/// <returns> a negative integer, 0, or positive integer to represent the
/// object o1 is less, equals, or greater than object 02. </returns>
public virtual int Compare(NoisyChannelScore o1, NoisyChannelScore o2)
{
// 1. compare how many words for the candidates
// for now, we assume less word is better,
// i.e. whatever is better than "what ever"
int @out = 0;
string cand1 = ((NoisyChannelScore)o1).GetCandStr();
string cand2 = ((NoisyChannelScore)o2).GetCandStr();
int wordNo1 = TermUtil.GetWordNo(cand1);
int wordNo2 = TermUtil.GetWordNo(cand2);
if (wordNo1 != wordNo2)
{
@out = wordNo1 - wordNo2; // less wordNo has higher rank
}
else
{
// 2. compare noisy Channel score
double score1 = ((NoisyChannelScore)o1).GetScore();
double score2 = ((NoisyChannelScore)o2).GetScore();
// SCR-2: use a fixed number to ensure result is not 0.
if (score2 > score1)
{
// from high to low
@out = 1;
}
else if (score2 < score1)
{
@out = -1;
}
else
{
// 3. compare by orthographic score
OrthographicScore oScore1 = ((NoisyChannelScore)o1).GetOScore();
OrthographicScore oScore2 = ((NoisyChannelScore)o2).GetOScore();
if (oScore1.GetScore() != oScore2.GetScore())
{
OrthographicScoreComparator <OrthographicScore> osc = new OrthographicScoreComparator <OrthographicScore>();
@out = osc.Compare(oScore1, oScore2);
}
else // 4. hannelScore
{
FrequencyScore fScore1 = ((NoisyChannelScore)o1).GetFScore();
FrequencyScore fScore2 = ((NoisyChannelScore)o2).GetFScore();
if (fScore1.GetScore() != fScore2.GetScore())
{
FrequencyScoreComparator <FrequencyScore> fsc = new FrequencyScoreComparator <FrequencyScore>();
@out = fsc.Compare(fScore1, fScore2);
}
else // 4. alphabetic order
{
@out = cand2.CompareTo(cand1);
}
}
}
}
return(@out);
}
// private method
private static void Test(string srcStr, string tarStr)
{
double wf1 = 1.00; // weighting factor
double wf2 = 0.70;
double wf3 = 0.80;
OrthographicScore os = new OrthographicScore(srcStr, tarStr, wf1, wf2, wf3);
Console.WriteLine(os.ToString());
}
// private constructor
public NoisyChannelScore(string wordStr, string candStr, WordWcMap wordWcMap, double wf1, double wf2, double wf3)
{
wordStr_ = wordStr;
candStr_ = candStr;
// calculate score
oScore_ = new OrthographicScore(wordStr_, candStr_, wf1, wf2, wf3);
fScore_ = new FrequencyScore(candStr_, wordWcMap);
score_ = oScore_.GetScore() * fScore_.GetScore();
}
// private constructor
public CSpellScore(string wordStr, string candStr, WordWcMap wordWcMap, DoubleVec contextVec, Word2Vec word2Vec, double wf1, double wf2, double wf3)
{
wordStr_ = wordStr;
candStr_ = candStr;
// calculate score
oScore_ = new OrthographicScore(wordStr_, candStr_, wf1, wf2, wf3);
fScore_ = new FrequencyScore(candStr_, wordWcMap);
nScore_ = new NoisyChannelScore(wordStr_, candStr_, wordWcMap, wf1, wf2, wf3);
cScore_ = new ContextScore(candStr_, contextVec, word2Vec);
}
// return candidate set with orthographic score
// inStr is the srcTxt used to calculate the score between it and cand
public static HashSet <OrthographicScore> GetCandidateScoreSet(string inStr, HashSet <string> candidates, double wf1, double wf2, double wf3)
{
HashSet <OrthographicScore> candScoreSet = new HashSet <OrthographicScore>();
foreach (string cand in candidates)
{
OrthographicScore os = new OrthographicScore(inStr, cand, wf1, wf2, wf3);
candScoreSet.Add(os);
}
return(candScoreSet);
}
// by combination, O, N, F, C
private int compareByCombo(CSpellScore o1, CSpellScore o2)
{
int @out = 0;
OrthographicScore oScore1 = ((CSpellScore)o1).GetOScore();
OrthographicScore oScore2 = ((CSpellScore)o2).GetOScore();
NoisyChannelScore nScore1 = ((CSpellScore)o1).GetNScore();
NoisyChannelScore nScore2 = ((CSpellScore)o2).GetNScore();
FrequencyScore fScore1 = ((CSpellScore)o1).GetFScore();
FrequencyScore fScore2 = ((CSpellScore)o2).GetFScore();
ContextScore cScore1 = ((CSpellScore)o1).GetCScore();
ContextScore cScore2 = ((CSpellScore)o2).GetCScore();
// 1. compared by orthographic score, best
if (oScore1.GetScore() != oScore2.GetScore())
{
OrthographicScoreComparator <OrthographicScore> osc = new OrthographicScoreComparator <OrthographicScore>();
@out = osc.Compare(oScore1, oScore2);
}
// 2. compared by noise channel score, 2nd best
else if (nScore1.GetScore() != nScore2.GetScore())
{
NoisyChannelScoreComparator <NoisyChannelScore> nsc = new NoisyChannelScoreComparator <NoisyChannelScore>();
@out = nsc.Compare(nScore1, nScore2);
}
// 3. compared by pure frequency score, 3rd best
else if (fScore1.GetScore() != fScore2.GetScore())
{
FrequencyScoreComparator <FrequencyScore> fsc = new FrequencyScoreComparator <FrequencyScore>();
@out = fsc.Compare(fScore1, fScore2);
}
// 4. compared by context score, 4 last
else if (cScore1.GetScore() != cScore2.GetScore())
{
ContextScoreComparator <ContextScore> csc = new ContextScoreComparator <ContextScore>();
@out = csc.Compare(cScore1, cScore2);
}
// 5. alphabetic order
else
{
string cand1 = ((CSpellScore)o1).GetCandStr();
string cand2 = ((CSpellScore)o2).GetCandStr();
@out = cand2.CompareTo(cand1);
}
return(@out);
}
private int compareByOrthographics(CSpellScore o1, CSpellScore o2)
{
int @out = 0;
OrthographicScore oScore1 = ((CSpellScore)o1).GetOScore();
OrthographicScore oScore2 = ((CSpellScore)o2).GetOScore();
// 1. compared by orthographic score, best
if (oScore1.GetScore() != oScore2.GetScore())
{
OrthographicScoreComparator <OrthographicScore> osc = new OrthographicScoreComparator <OrthographicScore>();
@out = osc.Compare(oScore1, oScore2);
}
// 2. alphabetic order
else
{
string cand1 = ((CSpellScore)o1).GetCandStr();
string cand2 = ((CSpellScore)o2).GetCandStr();
@out = cand2.CompareTo(cand1);
}
return(@out);
}
/// <summary>
/// Compare two object o1 and o2. Both objects o1 and o2 are
/// OrthographicScore. The compare algorithm:
/// </summary>
/// <param name="o1"> first object to be compared </param>
/// <param name="o2"> second object to be compared
/// </param>
/// <returns> a negative integer, 0, or positive integer to represent the
/// object o1 is less, equals, or greater than object 02. </returns>
public virtual int Compare(OrthographicScore o1, OrthographicScore o2)
{
// 1. compare total score first
double score1 = ((OrthographicScore)o1).GetScore();
double score2 = ((OrthographicScore)o2).GetScore();
int @out = 0;
// SCR-2: use a fixed number to ensure result is not 0.
if (score2 > score1)
{
// from high to low
//out = (int) (1000*(score2-score1)); might result 0
@out = 1;
}
else if (score2 < score1)
{
@out = -1;
}
else
{
// 2. compare edit score
double tokenScore1 = ((OrthographicScore)o1).GetTokenScore();
double tokenScore2 = ((OrthographicScore)o2).GetTokenScore();
// SCR-2: use a fixed number to ensure result is not 0.
if (tokenScore2 > tokenScore1)
{
@out = 1;
}
if (tokenScore2 < tokenScore1)
{
@out = -1;
}
else
{
// 3. compare phoneticScore
double pScore1 = ((OrthographicScore)o1).GetPhoneticScore();
double pScore2 = ((OrthographicScore)o2).GetPhoneticScore();
// SCR-2: use a fixed number to ensure result is not 0.
if (pScore2 > pScore1)
{
@out = 1;
}
else if (pScore2 < pScore1)
{
@out = -1;
}
else
{
double oScore1 = ((OrthographicScore)o1).GetOverlapScore();
double oScore2 = ((OrthographicScore)o2).GetOverlapScore();
// 4. compare overlap Score
// SCR-2: use a fixed number to ensure result is not 0.
if (oScore2 > oScore1)
{
@out = 1;
}
else if (oScore2 < oScore1)
{
@out = -1;
}
else // 5. Alphabetic order
{
string str1 = ((OrthographicScore)o1).GetTarStr();
string str2 = ((OrthographicScore)o2).GetTarStr();
@out = str2.CompareTo(str1);
}
}
}
}
return(@out);
}