public Lemmatizer(LemmatizerSettings lsett)
{
this.Lsett = lsett;
this.ElExamples = new ExampleList(lsett);
this.LtnRootNode = null;
this.LtnRootNodeFront = null;
}
/// <summary>
///
/// </summary>
/// <param name="lsett"></param>
/// <param name="elExamples"></param>
/// <param name="iStart">Index of the first word of the current group</param>
/// <param name="iEnd">Index of the last word of the current group</param>
/// <param name="ltnParentNode"></param>
private LemmaTreeNode(LemmatizerSettings lsett, ExampleList elExamples, int iStart, int iEnd, LemmaTreeNode ltnParentNode) : this(lsett) {
this.ltnParentNode = ltnParentNode;
this.dictSubNodes = null;
this.iStart = iStart;
this.iEnd = iEnd;
this.elExamples = elExamples;
if (iStart >= elExamples.Count || iEnd >= elExamples.Count || iStart > iEnd)
{
lrBestRule = elExamples.Rules.DefaultRule;
aBestRules = new RuleWeighted[1];
aBestRules[0] = new RuleWeighted(lrBestRule, 0);
dWeight = 0;
return;
}
int iConditionLength = Math.Min(ltnParentNode == null ? 0 : ltnParentNode.iSimilarity + 1, elExamples[iStart].Word.Length);
this.sCondition = elExamples[iStart].Word.Substring(elExamples[iStart].Word.Length - iConditionLength);
this.iSimilarity = elExamples[iStart].Similarity(elExamples[iEnd]);
this.bWholeWord = ltnParentNode == null ? false : elExamples[iEnd].Word.Length == ltnParentNode.iSimilarity;
FindBestRules();
AddSubAll();
//TODO check this heuristics, can be problematic when there are more applicable rules
if (dictSubNodes != null)
{
List <KeyValuePair <char, LemmaTreeNode> > lReplaceNodes = new List <KeyValuePair <char, LemmaTreeNode> >();
foreach (KeyValuePair <char, LemmaTreeNode> kvpChild in dictSubNodes)
{
if (kvpChild.Value.dictSubNodes != null && kvpChild.Value.dictSubNodes.Count == 1)
{
IEnumerator <LemmaTreeNode> enumChildChild = kvpChild.Value.dictSubNodes.Values.GetEnumerator();
enumChildChild.MoveNext();
LemmaTreeNode ltrChildChild = enumChildChild.Current;
if (kvpChild.Value.lrBestRule == lrBestRule)
{
lReplaceNodes.Add(new KeyValuePair <char, LemmaTreeNode>(kvpChild.Key, ltrChildChild));
}
}
}
foreach (KeyValuePair <char, LemmaTreeNode> kvpChild in lReplaceNodes)
{
dictSubNodes[kvpChild.Key] = kvpChild.Value;
kvpChild.Value.ltnParentNode = this;
}
}
}
public void Load(Latino.BinarySerializer binRead)
{
lsett = new LemmatizerSettings(binRead);
elExamples = new ExampleList(binRead, lsett);
if (!lsett.bBuildFrontLemmatizer)
{
ltnRootNode = new LemmaTreeNode(binRead, lsett, elExamples, null);
}
else
{
ltnRootNode = new LemmaTreeNode(binRead, lsett, elExamples.GetFrontRearExampleList(false), null);
ltnRootNodeFront = new LemmaTreeNode(binRead, lsett, elExamples.GetFrontRearExampleList(true), null);
}
}
public void Deserialize(BinaryReader binRead)
{
using (binRead)
{
// settings
Lsett = new LemmatizerSettings(binRead);
// examples
bool bSerializeExamples = binRead.ReadBoolean();
ElExamples = new ExampleList(binRead, Lsett);
ExampleList elExamplesRear;
ExampleList elExamplesFront;
if (bSerializeExamples)
{
elExamplesRear = ElExamples.GetFrontRearExampleList(false);
elExamplesFront = ElExamples.GetFrontRearExampleList(true);
}
else
{
elExamplesRear = new ExampleList(binRead, Lsett);
elExamplesFront = new ExampleList(binRead, Lsett);
}
// root node
LtnRootNode = new LemmaTreeNode(binRead, Lsett, Lsett.bBuildFrontLemmatizer ? elExamplesRear : ElExamples, null);
// root node front
if (Lsett.bBuildFrontLemmatizer)
{
LtnRootNodeFront = new LemmaTreeNode(binRead, Lsett, elExamplesFront, null);
}
// exceptions - use try catch for retro compatibility
// --> this section is missing in the old lemmatizer files
try
{
var nbOfExceptions = binRead.ReadInt32();
for (var i = 0; i < nbOfExceptions; i++)
{
var exception = binRead.ReadString();
var parts = exception.Split(' ');
this.AddException(parts[0], parts[1]);
}
}
catch (Exception)
{
Trace.WriteLine("Couldn't deserialize exceptions in Lemmatizer file");
}
}
}
private void AddSub(int iStart, int iEnd, char chChar)
{
var ltnSub = new LemmaTreeNode(lsett, elExamples, iStart, iEnd, this);
//TODO - maybe not realy appropriate because loosing statisitcs from multiple possible rules
if (ltnSub.lrBestRule == lrBestRule && ltnSub.dictSubNodes == null)
{
return;
}
if (dictSubNodes == null)
{
dictSubNodes = new Dictionary <char, LemmaTreeNode>();
}
dictSubNodes.Add(chChar, ltnSub);
}
public void Deserialize(BinaryReader binRead, LemmatizerSettings lsett, ExampleList elExamples, LemmaTreeNode ltnParentNode)
{
this.lsett = lsett;
// read is not null?
if (binRead.ReadBoolean())
{
// read all dictionary (key + value)
dictSubNodes = new ConcurrentDictionary <char, LemmaTreeNode>();
int iCount = binRead.ReadInt32();
for (int i = 0; i < iCount; i++)
{
char cKey = binRead.ReadChar();
var ltrSub = new LemmaTreeNode(binRead, this.lsett, elExamples, this);
dictSubNodes.TryAdd(cKey, ltrSub);
}
}
else
{
dictSubNodes = null;
}
this.ltnParentNode = ltnParentNode;
// read similarity, condition and wholeword?
iSimilarity = binRead.ReadInt32();
sCondition = binRead.ReadString();
bWholeWord = binRead.ReadBoolean();
// best rule signature
lrBestRule = elExamples.Rules[binRead.ReadString()];
// best rules
int iCountBest = binRead.ReadInt32();
aBestRules = new RuleWeighted[iCountBest];
for (int i = 0; i < iCountBest; i++)
{
aBestRules[i] = new RuleWeighted(elExamples.Rules[binRead.ReadString()], binRead.ReadDouble());
}
// weight, start, end
dWeight = binRead.ReadDouble();
iStart = binRead.ReadInt32();
iEnd = binRead.ReadInt32();
this.elExamples = elExamples;
}
// Essential Class Functions (building model & lemmatizing) ----------
public void BuildModel()
{
if (LtnRootNode != null)
{
return;
}
if (!Lsett.bBuildFrontLemmatizer)
{
//TODO remove: elExamples.FinalizeAdditions();
ElExamples.FinalizeAdditions();
LtnRootNode = new LemmaTreeNode(Lsett, ElExamples);
}
else
{
LtnRootNode = new LemmaTreeNode(Lsett, ElExamples.GetFrontRearExampleList(false));
LtnRootNodeFront = new LemmaTreeNode(Lsett, ElExamples.GetFrontRearExampleList(true));
}
}
public void Load(Latino.BinarySerializer binRead, LemmatizerSettings lsett, ExampleList elExamples, LemmaTreeNode ltnParentNode)
{
this.lsett = lsett;
if (binRead.ReadBool())
{
dictSubNodes = new Dictionary <char, LemmaTreeNode>();
int iCount = binRead.ReadInt();
for (int i = 0; i < iCount; i++)
{
char cKey = binRead.ReadChar();
LemmaTreeNode ltrSub = new LemmaTreeNode(binRead, this.lsett, elExamples, this);
dictSubNodes.Add(cKey, ltrSub);
}
}
else
{
dictSubNodes = null;
}
this.ltnParentNode = ltnParentNode;
iSimilarity = binRead.ReadInt();
sCondition = binRead.ReadString();
bWholeWord = binRead.ReadBool();
lrBestRule = elExamples.Rules[binRead.ReadString()];
int iCountBest = binRead.ReadInt();
aBestRules = new RuleWeighted[iCountBest];
for (int i = 0; i < iCountBest; i++)
{
aBestRules[i] = new RuleWeighted(elExamples.Rules[binRead.ReadString()], binRead.ReadDouble());
}
dWeight = binRead.ReadDouble();
iStart = binRead.ReadInt();
iEnd = binRead.ReadInt();
this.elExamples = elExamples;
}
public void AddExample(string sWord, string sLemma, double dWeight, string sMsd)
{
ElExamples.AddExample(sWord, sLemma, dWeight, sMsd);
LtnRootNode = null;
}
// Essential Class Functions (adding examples to repository) ----------
public void AddMultextFile(StreamReader srIn, string sFormat)
{
this.ElExamples.AddMultextFile(srIn, sFormat);
LtnRootNode = null;
}
public LemmaTreeNode(Latino.BinarySerializer binRead, LemmatizerSettings lsett, ExampleList elExamples, LemmaTreeNode ltnParentNode)
{
Load(binRead, lsett, elExamples, ltnParentNode);
}
public LemmaTreeNode(BinaryReader binRead, LemmatizerSettings lsett, ExampleList elExamples, LemmaTreeNode ltnParentNode)
{
Deserialize(binRead, lsett, elExamples, ltnParentNode);
}