/// <summary>
/// Constructor
/// </summary>
/// <param name="wordNetEngine">WordNet engine to use</param>
public WordNetSimilarityModel(WordNetEngine wordNetEngine)
{
_wordNetEngine = wordNetEngine;
}
/// <summary>
/// Instantiates the current synset. If idSynset is non-null, related synsets references are set to those from
/// idSynset; otherwise, related synsets are created as shells.
/// </summary>
/// <param name="definition">Definition line of synset from data file</param>
/// <param name="idSynset">Lookup for related synsets. If null, all related synsets will be created as shells.</param>
internal void Instantiate(string definition, Dictionary <string, SynSet> idSynset)
{
// don't re-instantiate
if (_instantiated)
{
throw new Exception("Synset has already been instantiated");
}
/* get lexicographer file name...the enumeration lines up precisely with the wordnet spec (see the lexnames file) except that
* it starts with None, so we need to add 1 to the definition line's value to get the correct file name */
int lexicographerFileNumber = int.Parse(GetField(definition, 1)) + 1;
if (lexicographerFileNumber <= 0)
{
throw new Exception("Invalid lexicographer file name number. Should be >= 1.");
}
_lexicographerFileName = (WordNetEngine.LexicographerFileName)lexicographerFileNumber;
// get number of words in the synset and the start character of the word list
int wordStart;
int numWords = int.Parse(GetField(definition, 3, out wordStart), NumberStyles.HexNumber);
wordStart = definition.IndexOf(' ', wordStart) + 1;
// get words in synset
_words = new List <string>(numWords);
for (int i = 0; i < numWords; ++i)
{
int wordEnd = definition.IndexOf(' ', wordStart + 1) - 1;
int wordLen = wordEnd - wordStart + 1;
string word = definition.Substring(wordStart, wordLen);
if (Enumerable.Contains(word, ' ')) //ZK changed to static method
{
throw new Exception("Unexpected space in word: " + word);
}
_words.Add(word);
// skip lex_id field
wordStart = definition.IndexOf(' ', wordEnd + 2) + 1;
}
// get gloss
_gloss = definition.Substring(definition.IndexOf('|') + 1).Trim();
if (Enumerable.Contains(_gloss, '|')) //ZK changed to static method
{
throw new Exception("Unexpected pipe in gloss");
}
// get number and start of relations
int relationCountField = 3 + (_words.Count * 2) + 1;
int relationFieldStart;
int numRelations = int.Parse(GetField(definition, relationCountField, out relationFieldStart));
relationFieldStart = definition.IndexOf(' ', relationFieldStart) + 1;
// grab each related synset
_relationSynSets = new Dictionary <WordNetEngine.SynSetRelation, Set <SynSet> >();
_lexicalRelations = new Dictionary <WordNetEngine.SynSetRelation, Dictionary <SynSet, Dictionary <int, Set <int> > > >();
for (int relationNum = 0; relationNum < numRelations; ++relationNum)
{
string relationSymbol = null;
int relatedSynSetOffset = -1;
WordNetEngine.POS relatedSynSetPOS = WordNetEngine.POS.None;
int sourceWordIndex = -1;
int targetWordIndex = -1;
// each relation has four columns
for (int relationField = 0; relationField <= 3; ++relationField)
{
int fieldEnd = definition.IndexOf(' ', relationFieldStart + 1) - 1;
int fieldLen = fieldEnd - relationFieldStart + 1;
string fieldValue = definition.Substring(relationFieldStart, fieldLen);
// relation symbol
if (relationField == 0)
{
relationSymbol = fieldValue;
}
// related synset offset
else if (relationField == 1)
{
relatedSynSetOffset = int.Parse(fieldValue);
}
// related synset POS
else if (relationField == 2)
{
relatedSynSetPOS = GetPOS(fieldValue);
}
// source/target word for lexical relation
else if (relationField == 3)
{
sourceWordIndex = int.Parse(fieldValue.Substring(0, 2), NumberStyles.HexNumber);
targetWordIndex = int.Parse(fieldValue.Substring(2), NumberStyles.HexNumber);
}
else
{
throw new Exception();
}
relationFieldStart = definition.IndexOf(' ', relationFieldStart + 1) + 1;
}
// get related synset...create shell if we don't have a lookup
SynSet relatedSynSet;
if (idSynset == null)
{
relatedSynSet = new SynSet(relatedSynSetPOS, relatedSynSetOffset, _wordNetEngine);
}
// look up related synset directly
else
{
relatedSynSet = idSynset[relatedSynSetPOS + ":" + relatedSynSetOffset];
}
// get relation
WordNetEngine.SynSetRelation relation = WordNetEngine.GetSynSetRelation(_pos, relationSymbol);
// add semantic relation if we have neither a source nor a target word index
if (sourceWordIndex == 0 && targetWordIndex == 0)
{
DictionaryExtensions.EnsureContainsKey(_relationSynSets, relation, typeof(Set <SynSet>)); //ZK changed to static method
_relationSynSets[relation].Add(relatedSynSet);
}
// add lexical relation
else
{
DictionaryExtensions.EnsureContainsKey(_lexicalRelations, relation, typeof(Dictionary <SynSet, Dictionary <int, Set <int> > >)); //ZK changed to static method
DictionaryExtensions.EnsureContainsKey(_lexicalRelations[relation], relatedSynSet, typeof(Dictionary <int, Set <int> >)); //ZK changed to static method
DictionaryExtensions.EnsureContainsKey(_lexicalRelations[relation][relatedSynSet], sourceWordIndex, typeof(Set <int>)); //ZK changed to static method
if (!_lexicalRelations[relation][relatedSynSet][sourceWordIndex].Contains(targetWordIndex))
{
_lexicalRelations[relation][relatedSynSet][sourceWordIndex].Add(targetWordIndex);
}
}
}
// release the wordnet engine if we have one...don't need it anymore
if (_wordNetEngine != null)
{
_wordNetEngine = null;
}
_instantiated = true;
}
/// <summary>
/// Gets synset shells from a word index line. A synset shell is an instance of SynSet with only the POS and Offset
/// members initialized. These members are enough to look up the full synset within the corresponding data file. This
/// method is static to prevent inadvertent references to a current WordNetEngine, which should be passed via the
/// corresponding parameter.
/// </summary>
/// <param name="wordIndexLine">Word index line from which to get synset shells</param>
/// <param name="pos">POS of the given index line</param>
/// <param name="mostCommonSynSet">Returns the most common synset for the word</param>
/// <param name="wordNetEngine">WordNetEngine to pass to the constructor of each synset shell</param>
/// <returns>Synset shells for the given index line</returns>
private static Set <SynSet> GetSynSetShells(string wordIndexLine, POS pos, out SynSet mostCommonSynSet, WordNetEngine wordNetEngine)
{
Set <SynSet> synsets = new Set <SynSet>();
mostCommonSynSet = null;
// get number of synsets
string[] parts = wordIndexLine.Split(new char[] { ' ' }, StringSplitOptions.RemoveEmptyEntries);
int numSynSets = int.Parse(parts[2]);
// grab each synset shell, from last to first
int firstOffsetIndex = parts.Length - numSynSets;
for (int i = parts.Length - 1; i >= firstOffsetIndex; --i)
{
// create synset
int offset = int.Parse(parts[i]);
// add synset to collection
SynSet synset = new SynSet(pos, offset, wordNetEngine);
synsets.Add(synset);
// if this is the last synset offset to get (since we grabbed them in reverse order), record it as the most common synset
if (i == firstOffsetIndex)
{
mostCommonSynSet = synset;
}
}
if (mostCommonSynSet == null)
{
throw new Exception("Failed to get most common synset");
}
return(synsets);
}