private void reportTarget(spiderTarget t, folderNode fn, int c)
{
string pageFolder = "P" + c.ToString("D3") + "_" + t.IsRelevant.ToString();
folderNode pfn = fn.Add(pageFolder, "Page " + c.ToString(), "Report on page " + t.url + " crawled by " + name + ". Target.IsRelevant: " + t.IsRelevant + ".".addLine(pageDescription));
fileunit content = new fileunit(pfn.pathFor("content.txt"), false);
fileunit links = new fileunit(pfn.pathFor("links.txt"), false);
if (t.evaluation != null)
{
t.evaluation.saveObjectToXML(pfn.pathFor("relevance.xml"));
}
content.setContent(t.pageText);
//t.page.relationship.outflowLinks
if (t.page != null)
{
foreach (spiderLink ln in t.page.relationship.outflowLinks.items.Values)
{
string rl = ln.url;
links.Append(ln.url);
}
//t.page.webpage.links.ForEach(x => links.Append(x.nature + " | " + x.name + " | " + x.url));
}
content.Save();
links.Save();
// marks.Save();
}
public override spiderEvalRuleResult evaluate(spiderLink link)
{
spiderTarget target = wRecord.context.targets.GetOrCreateTarget(link, false, false);
spiderEvalRuleResult output = new spiderEvalRuleResult(this);
double weight = 0;
foreach (IWeightTableTerm term in target.tokens)
{
if (language.isKnownWord(term.nominalForm))
{
weight += target.tokens.GetTF_IDF(term);
}
else
{
weight -= target.tokens.GetTF_IDF(term);
}
}
if (weight > 0)
{
output.layer = layerID;
}
else
{
output.layer = layer2ID;
}
output.weightScore = weight;
return(output);
}
public override spiderEvalRuleResult evaluate(spiderLink link)
{
spiderEvalRuleResult result = new spiderEvalRuleResult(this);
spiderTarget target = wRecord.context.targets.GetOrCreateTarget(link, false, false);
termQueryDocument query = target.getQuery(expansionSteps, wRecord.logBuilder);
wRecord.logBuilder.AppendLine("Target [" + link.url + "] query => [" + query.GetAllTermString().toCsvInLine(",") + "]");
weightTableMatchCollection <termSpark, termSpark> matchLinks = query.GetSparkMatchAgainst <termSpark>((termDocument)wRecord.context.targets.dlTargetLinkTokens.AggregateDocument);
weightTableMatchCollection <termSpark, termSpark> matchPage = query.GetSparkMatchAgainst <termSpark>((termDocument)wRecord.context.targets.dlTargetPageTokens.AggregateDocument);
if ((!matchLinks.Any()) && (!matchPage.Any()))
{
result.score = scoreUnit;
wRecord.logBuilder.AppendLine("D[" + link.url + "][" + target.tokens.GetAllTermString().toCsvInLine(",") + "] = no matches with query");
return(result);
}
else
{
wRecord.logBuilder.AppendLine("matchLinks => " + matchLinks.ToString());
wRecord.logBuilder.AppendLine("matchPage => " + matchPage.ToString());
}
double pLSim = matchLinks.GetSemanticSimilarity() * target_sd;
double pPSim = matchPage.GetSemanticSimilarity() * page_sd;
double sim = (pLSim + pPSim);
double sc = sim * (double)scoreUnit;
double score = ((double)scoreUnit) - sc;
if (doAdjustScoreByLanguageDetection)
{
// < ---- modification of diversity score
List <string> tkns = new List <string>();
foreach (IWeightTableTerm spark in query)
{
tkns.Add(spark.nominalForm);
}
textEvaluation evaluation = new textEvaluation(wRecord.aJob.langTextEvaluator, null);
evaluation.evaluateTokens(tkns, null, false);
double evalAdj = Math.Pow(evaluation.ratioA, 2);
result.score = Convert.ToInt32((double)score * evalAdj); //Convert.ToInt32(sim_inv * (Double) scoreUnit);
wRecord.logBuilder.AppendLine();
wRecord.logBuilder.AppendLine("Score is adjusted by language evaluation ratioA ^ 2: " + evalAdj);
}
wRecord.logBuilder.AppendLine("D[" + link.url + "][" + target.tokens.GetAllTermString().toCsvInLine(",") + "]=[pL:" + pLSim.ToString("P2") + "][pP:" + pPSim.ToString("P2") + "]=" + sim.ToString("#0.0000") + " (" + result.score + ")");
return(result);
}
public override spiderEvalRuleResult evaluate(spiderLink link)
{
spiderEvalRuleResult output = new spiderEvalRuleResult(this);
spiderTarget target = wRecord.context.targets.GetOrCreateTarget(link, false, false);
double score = hits[target] * scoreUnit;
output.score = (int)score;
return(output);
}
public indexPage deployTarget(spiderTarget target, modelSpiderSiteRecord wRecord, indexDomain idomain)
{
indexPage page = pageIndexTable.GetPageForUrl(target.url); //.GetOrCreate(md5.GetMd5Hash(target.url));
if (idomain == null)
{
idomain = domainIndexTable.GetDomain(wRecord.domainInfo.domainName);
}
page.url = target.url;
page.tst = target.tokens.ToList().toCsvInLine(",");
page.domain = wRecord.domainInfo.domainName;
if (target.isLoaded)
{
if (target.evaluation != null)
{
if (target.evaluation.result_language != basicLanguageEnum.unknown)
{
page.langTestRatio = target.evaluation.result_ratio;
page.singleMatchTokens = target.evaluation.singleLanguageTokens.toCsvInLine(",");
page.multiMatchTokens = target.evaluation.multiLanguageTokens.toCsvInLine(",");
page.wordCount = target.evaluation.allContentTokens.Count();
page.AllWords = target.evaluation.allContentTokens.toCsvInLine();
page.language = target.evaluation.result_language.ToString();
}
}
if (target.IsRelevant)
{
page.relevancy = indexPageRelevancyEnum.isRelevant;
}
else if (target.evaluatedLanguage == basicLanguageEnum.unknown)
{
page.relevancy = indexPageRelevancyEnum.unknown;
}
else
{
page.relevancy = indexPageRelevancyEnum.notRelevant;
}
page.byteSize = target.page.spiderResult.page.result.byteSize;
}
pageIndexTable.AddOrUpdate(page);
return(page);
}
public modelSpiderSiteRecordEventArgs(ISpiderTarget __target, modelSpiderSiteRecordEventType __type = modelSpiderSiteRecordEventType.DLCTargetPageAttached)
{
Target = __target;
if (__target is spiderTarget)
{
spiderTarget target = (spiderTarget)__target;
sourceHtml = target.page.webpage.result.sourceCode;
htmlDoc = target.page.webpage.result.HtmlDocument; //(HtmlDocument)target.page.webpage.result.document.getDocument<HtmlDocument>();
if (htmlDoc != null)
{
if (htmlDoc.DocumentNode != null)
{
sourceXml = target.page.webpage.result.sourceCode;
}
}
//sourceXml = target.page.spiderResult.page.result.document.getDocument<HtmlDocument>()
}
type = __type;
}
/// <summary>
/// Evaluates the specified link.
/// </summary>
/// <param name="link">The link.</param>
/// <returns></returns>
public override spiderEvalRuleResult evaluate(spiderLink link)
{
spiderTarget target = wRecord.context.targets.GetOrCreateTarget(link, false, false);
spiderEvalRuleResult output = new spiderEvalRuleResult(this);
output.layer = layer2ID;
foreach (IWeightTableTerm term in target.tokens)
{
if (needles.Contains(term.nominalForm))
{
//if (needles.Contains("ru"))
//{
// aceLog.log("Target [" + target.url + "] triggered by : " + needles.Join(","));
//}
output.layer = layerID;
break;
}
}
return(output);
}
public override spiderEvalRuleResult evaluate(spiderLink link)
{
spiderTarget target = wRecord.context.targets.GetOrCreateTarget(link, false, false);
spiderEvalRuleResult output = new spiderEvalRuleResult(this);
double weight = 0;
foreach (IWeightTableTerm term in target.tokens)
{
var response = semanticLexiconManager.lexiconCache.getLexiconItems(term.nominalForm, wRecord.logBuilder);
switch (response.type)
{
case lexiconResponse.responseType.cachedLexicon:
case lexiconResponse.responseType.lexicon:
case lexiconResponse.responseType.askingLexiconContext:
weight += target.tokens.GetTF_IDF(term);
break;
default:
weight -= target.tokens.GetTF_IDF(term);
break;
}
}
if (weight > 0)
{
output.layer = layerID;
}
else
{
output.layer = layer2ID;
}
output.weightScore = weight;
return(output);
}
/// <summary>
/// Evaluates the specified link.
/// </summary>
/// <param name="link">The link.</param>
/// <returns></returns>
public override spiderEvalRuleResult evaluate(spiderLink link)
{
try
{
spiderEvalRuleResult result = new spiderEvalRuleResult(this);
spiderTarget target = wRecord.context.targets.GetByOrigin(link);
if (target == null)
{
result.layer = layer2ID;
return(result);
}
nodeBlock bl = target.contentBlocks.GetBlockByXPath(link.link.xPath);
if (bl == null)
{
result.layer = layer2ID;
return(result);
}
if (bl.role == semanticRole)
{
result.layer = layerID;
}
else
{
result.layer = layer2ID;
}
return(result);
} catch (Exception ex)
{
throw new aceGeneralException(ex.Message, ex, this, "layerBlockRolePRule broken");
}
return(new spiderEvalRuleResult(this));
}
public override spiderEvalRuleResult evaluate(spiderLink link)
{
spiderEvalRuleResult output = new spiderEvalRuleResult(this);
spiderTarget target = wRecord.context.targets.GetOrCreateTarget(link, false, false);
if (!ranks.Any())
{
output.score = scoreUnit;
return(output);
}
if (ranks.ContainsKey(target))
{
output.score = ranks[target];
}
else
{
output.score = penaltyUnit;
}
return(output);
}
public List <string> GetTermsForPage(spiderTarget target, indexDomain idomain = null, indexPage ipage = null, multiLanguageEvaluator evaluator = null, builderForLog loger = null)
{
if (idomain == null)
{
idomain = imbWEMManager.index.domainIndexTable.GetOrCreate(target.parent.wRecord.domain);
}
if (ipage == null)
{
ipage = imbWEMManager.index.pageIndexTable.GetOrCreate(md5.GetMd5Hash(target.url));
}
List <string> output = new List <string>();
FileInfo file = GetWordList_File(idomain, ipage);
if (imbWEMManager.settings.TFIDF.doUseSavedPageWordlists && file.Exists)
{
output = file.FullName.openFileToList(true);
return(output);
}
string cont = target.pageText.transliterate();
// cont = cont.imbHtmlDecode();
termDocument pageTF = null;
if (evaluator == null)
{
evaluator = target.parent.wRecord.tRecord.evaluator;
}
multiLanguageEvaluation evaluation = evaluator.evaluate(cont);
if (evaluation.result_language == basicLanguageEnum.serbian)
{
List <string> pt = new List <string>();
pt.AddRange(evaluation.singleLanguageTokens);
if (!imbWEMManager.settings.TFIDF.doUseOnlySingleMatch)
{
pt.AddRange(evaluation.multiLanguageTokens);
}
pt.RemoveAll(x => !x.isCleanWord());
pt.RemoveAll(x => x.isSymbolicContentOnly());
var tkns = semanticLexiconManager.lexiconCache.encodeTwins(pt);
output.AddRange(tkns);
}
if (imbWEMManager.settings.TFIDF.doSavePageWordlist)
{
output.saveContentOnFilePath(file.FullName);
}
return(output);
}
