// TBD, this file should be deleted by moving each method to
// the assocaited ranking class
// public method
public static void PrintContextScore(HashSet <string> candSet, int tarPos, int tarSize, List <TokenObj> inTextList, Word2Vec word2VecIm, Word2Vec word2VecOm, bool word2VecSkipWord, int contextRadius, int maxCandNo, bool debugFlag)
{
if (debugFlag == true)
{
ContextScoreComparator <ContextScore> csc = new ContextScoreComparator <ContextScore>();
HashSet <ContextScore> cScoreSet = RankByContext.GetCandidateScoreSet(candSet, tarPos, tarSize, inTextList, word2VecIm, word2VecOm, word2VecSkipWord, contextRadius, debugFlag);
var list = cScoreSet.OrderBy(x => x, csc).Take(maxCandNo).Select(obj => obj.ToString()).ToList();
foreach (var item in list)
{
DebugPrint.PrintCScore(item, debugFlag);
}
}
}
// return candidate scoreObj list sorted by score, higher first
public static List <ContextScore> GetCandidateScoreList(HashSet <MergeObj> candidates, List <TokenObj> nonSpaceTokenList, Word2Vec word2VecIm, Word2Vec word2VecOm, bool word2VecSkipWord, int contextRadius, bool debugFlag)
{
// find score object set for each candidates ...
HashSet <ContextScore> candScoreSet = GetCandidateScoreSet(candidates, nonSpaceTokenList, word2VecIm, word2VecOm, word2VecSkipWord, contextRadius, debugFlag);
// sorted by the score, higher go first
List <ContextScore> candScoreList = new List <ContextScore>(candScoreSet);
ContextScoreComparator <ContextScore> csc = new ContextScoreComparator <ContextScore>();
candScoreList.Sort(csc);
// print detail
foreach (ContextScore contextScore in candScoreList)
{
DebugPrint.PrintCScore(contextScore.ToString(), debugFlag);
}
return(candScoreList);
}
// return the best ranked str from candidates using word2Vec score
// inTokenList, includes space token, is not coreTerm.Lc
// return the orignal inStr if no candidate has score > 0.0d
public static string GetTopRankStr(string inStr, HashSet <string> candidates, int tarPos, int tarSize, List <TokenObj> nonSpaceTokenList, Word2Vec word2VecIm, Word2Vec word2VecOm, bool word2VecSkipWord, int contextRadius, int shortSplitWordLength, int maxShortSplitWordNo, double rwSplitFactor, int maxCandNo, bool debugFlag)
{
// init
string topRankStr = inStr;
// Find the correction str
if (candidates.Count > 0)
{
// 1. sorted score list for each candidates ...
// This ranking can be improved if n-gram model (frequecny) is used
List <ContextScore> candScoreList = RankByContext.GetCandidateScoreList(candidates, tarPos, tarSize, nonSpaceTokenList, word2VecIm, word2VecOm, word2VecSkipWord, contextRadius, debugFlag);
// 1.1 get the top tank candidate
ContextScore topContextScore = candScoreList[0];
// 2. validate the top rank
// 2.1 wordVec for context
DoubleVec contextVec = Word2VecContext.GetContextVec(tarPos, tarSize, nonSpaceTokenList, word2VecIm, contextRadius, word2VecSkipWord, debugFlag);
// 2.2 wordVec for the original words before split
ContextScore orgContextScore = new ContextScore(inStr, contextVec, word2VecOm);
// 2.3 compare the top rank split to the original string b4 split
if (IsTopCandValid(inStr, orgContextScore, topContextScore, rwSplitFactor, debugFlag) == true)
{
// no correction: if score is not good enough for corection
topRankStr = topContextScore.GetTerm();
}
// debug print
if (debugFlag == true)
{
// print focus token (original)
DebugPrint.PrintCScore(orgContextScore.ToString(), debugFlag);
// print candidates
ContextScoreComparator <ContextScore> csc = new ContextScoreComparator <ContextScore>();
var list = candScoreList.OrderBy(x => x, csc).Take(maxCandNo).Select(obj => obj.ToString()).ToList();
foreach (var item in list)
{
DebugPrint.PrintCScore(item, debugFlag);
}
}
}
return(topRankStr);
}
