private static WSDWordInfo lookupWord(string word)
{
// OVERVIEW: For each part of speech, look for the word.
// Compare relative strengths of the synsets in each category
// to determine the most probable part of speech.
//
// PROBLEM: Word definitions are often context-based. It would be better
// to find a way to search in-context in stead of just singling
// out an individual word.
//
// SOLUTION: Modify FindPartOfSpeech to include a second argument, string
// context. The pass the entire sentence as the context for part
// of speech determination.
//
// PROBLEM: That's difficult to do so I'm going to keep this simple for now.
int maxCount = 0;
WSDWordInfo wordinfo = new WSDWordInfo();
wordinfo.partOfSpeech = Wnlib.PartsOfSpeech.Unknown;
// for each part of speech...
Wnlib.PartsOfSpeech[] enums = (Wnlib.PartsOfSpeech[])Enum.GetValues(typeof(Wnlib.PartsOfSpeech));
wordinfo.senseCounts = new int[enums.Length];
for (int i = 0; i < enums.Length; i++)
{
// get a valid part of speech
Wnlib.PartsOfSpeech pos = enums[i];
if (pos == Wnlib.PartsOfSpeech.Unknown)
{
continue;
}
// get an index to a synset collection
Wnlib.Index index = Wnlib.Index.lookup(word, Wnlib.PartOfSpeech.of(pos));
// none found?
if (index == null)
{
continue;
}
// does this part of speech have a higher sense count?
wordinfo.senseCounts[i] = index.sense_cnt;
if (wordinfo.senseCounts[i] > maxCount)
{
maxCount = wordinfo.senseCounts[i];
wordinfo.partOfSpeech = pos;
}
}
return(wordinfo);
}
/*--------
* Methods
*--------*/
/// <summary>Finds the part of speech for a given single word</summary>
/// <param name="word">the word</param>
/// <param name="includeMorphs">include morphology? (fuzzy matching)</param>
/// <returns>a structure containing information about the word</returns>
/// <remarks>
/// This function is designed to determine the part of speech of a word. Since all
/// of the WordNet search functions require the part of speech, this function will be useful
/// in cases when the part of speech of a word is not known. It is not 100% correct
/// because WordNet was most likely not intended to be used this way. However, it is
/// accurate enough for most applications.
/// </remarks>
public static WSDWordInfo FindWordInfo(string word, bool includeMorphs)
{
word = word.ToLower();
WSDWordInfo wordinfo = lookupWord(word);
// include morphology if nothing was found on the original word
if (wordinfo.Strength == 0 && includeMorphs)
{
wordinfo = lookupWordMorphs(word);
}
return(wordinfo);
}
private static WSDWordInfo getMorphInfo(ArrayList morphinfos, string morph)
{
// Attempt to find the morph string in the list.
// NOTE: Since the list should never get very large, a selection search will work just fine
foreach (WSDWordInfo morphinfo in morphinfos)
{
if (morphinfo.text == morph)
{
return(morphinfo);
}
}
// if not found, create a new one
WSDWordInfo wordinfo = new WSDWordInfo();
wordinfo.text = morph;
wordinfo.senseCounts = new int[enums.Length];
return((WSDWordInfo)morphinfos[morphinfos.Add(wordinfo)]);
}
private static WSDWordInfo lookupWordMorphs(string word)
{
// OVERVIEW: This functions only gets called when the word was not found with
// an exact match. So, enumerate all the parts of speech, then enumerate
// all of the word's morphs in each category. Perform a lookup on each
// morph and save the morph/strength/part-of-speech data sets. Finally,
// loop over all the data sets and then pick the strongest one.
ArrayList wordinfos = new ArrayList();
// for each part of speech...
for (int i = 0; i < enums.Length; i++)
{
// get a valid part of speech
Wnlib.PartsOfSpeech pos = enums[i];
if (pos == Wnlib.PartsOfSpeech.Unknown)
{
continue;
}
// generate morph list
Wnlib.MorphStr morphs = new Wnlib.MorphStr(word, Wnlib.PartOfSpeech.of(pos));
string morph = "";
while ((morph = morphs.next()) != null)
{
// get an index to a synset collection
Wnlib.Index index = Wnlib.Index.lookup(morph, Wnlib.PartOfSpeech.of(pos));
// none found?
if (index == null)
{
continue;
}
// save the wordinfo
WSDWordInfo wordinfo = getMorphInfo(wordinfos, morph);
wordinfo.senseCounts[i] = index.sense_cnt;
}
}
// search the wordinfo list for the best match
WSDWordInfo bestWordInfo = new WSDWordInfo();
int maxStrength = 0;
foreach (WSDWordInfo wordinfo in wordinfos)
{
// for each part of speech...
int maxSenseCount = 0;
int strength = 0;
for (int i = 0; i < enums.Length; i++)
{
// get a valid part of speech
Wnlib.PartsOfSpeech pos = enums[i];
if (pos == Wnlib.PartsOfSpeech.Unknown)
{
continue;
}
// determine part of speech and strength
strength += wordinfo.senseCounts[i];
if (wordinfo.senseCounts[i] > maxSenseCount)
{
maxSenseCount = wordinfo.senseCounts[i];
wordinfo.partOfSpeech = pos;
}
}
// best match?
if (strength > maxStrength)
{
maxStrength = strength;
bestWordInfo = wordinfo;
}
}
return(bestWordInfo);
}
