private static string[] lookupSynonyms(Wnlib.Index index)
{
// OVERVIEW: For each sense, grab the synset associated with our index.
// Then, add the lexemes in the synset to a list.
var synonyms = new ArrayList(10);
// for each sense...
for (int s = 0; s < index.SynsetOffsets.Length; s++)
{
// read in the word and its pointers
var synset = new Wnlib.SynSet(index.SynsetOffsets[s], index.PartOfSpeech, index.Wd, null, s);
// build a string out of the words
for (int i = 0; i < synset.words.Length; i++)
{
string word = synset.words[i].word.Replace("_", " ");
// if the word is capitalized, that means it's a proper noun. We don't want those.
if (word[0] <= 'Z')
{
continue;
}
// add it to the list if it's a different word
if (string.Compare(word, index.Wd, true) != 0)
{
synonyms.Add(word);
}
}
}
return((string[])synonyms.ToArray(typeof(string)));
}
private static WordInfo 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;
WordInfo wordinfo = new WordInfo();
wordinfo.partOfSpeech = Wnlib.PartsOfSpeech.Unknown;
// for each part of speech...
var 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.SenseCnt;
if (wordinfo.SenseCounts[i] > maxCount)
{
maxCount = wordinfo.SenseCounts[i];
wordinfo.partOfSpeech = pos;
}
}
return(wordinfo);
}
/// <summary>Returns a list of Synonyms for a given word</summary>
/// <param name="word">the word</param>
/// <param name="pos">The Part of speech of a word</param>
/// <param name="includeMorphs">include morphology? (fuzzy matching)</param>
/// <returns>An array of strings containing the synonyms found</returns>
/// <remarks>
/// Note that my usage of 'Synonyms' here is not the same as hypernyms as defined by
/// WordNet. Synonyms in this sense are merely words in the same SynSet as the given
/// word. Hypernyms are found by tracing the pointers in a given synset.
/// </remarks>
public static string[] FindSynonyms(string word, Wnlib.PartsOfSpeech pos, bool includeMorphs)
{
// get an index to a synset collection
word = word.ToLower();
Wnlib.Index index = Wnlib.Index.lookup(word, Wnlib.PartOfSpeech.of(pos));
// none found?
if (index == null)
{
if (!includeMorphs)
{
return(null);
}
// check morphs
var morphs = new Wnlib.MorphStr(word, Wnlib.PartOfSpeech.of(pos));
string morph = "";
while ((morph = morphs.next()) != null)
{
index = Wnlib.Index.lookup(morph, Wnlib.PartOfSpeech.of(pos));
if (index != null)
{
break;
}
}
}
// still none found?
if (index == null)
{
return(null);
}
// at this point we will always have a valid index
return(lookupSynonyms(index));
}
private static WordInfo 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
WordInfo wordinfo = GetMorphInfo(wordinfos, morph);
wordinfo.SenseCounts[i] = index.SenseCnt;
}
}
// search the wordinfo list for the best match
WordInfo bestWordInfo = new WordInfo();
int maxStrength = 0;
foreach (WordInfo 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);
}
private static WordInfo lookupWordMorphs(string word, bool tagged_only)
{
// 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;
}
// none tagged
if (tagged_only && index.tagsense_cnt == 0)
{
continue;
}
// save the wordinfo
WordInfo wordinfo = getMorphInfo(wordinfos, morph);
if (tagged_only)
{
wordinfo.senseCounts[i] = index.tagsense_cnt;
}
else
{
wordinfo.senseCounts[i] = index.sense_cnt;
}
}
}
return(WordInfo.Compine(wordinfos));
/*
* // search the wordinfo list for the best match
* WordInfo bestWordInfo = new WordInfo();
* int maxStrength = 0;
* foreach( WordInfo 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;
*/
}
