def runLocalSearch(self, alpha, iterations):
state = self.getInitialState()
print "Initial State Obtained!"
print state
print "Running Local Search for " + str(iterations) + " iterations..."
for i in range(iterations):
for keyword in self.keywords:
print
print "Analyzing the keyword " + keyword
candidateLinks = [k[0] for k in self.cachedPages[keyword]]
assignedLink = state[keyword][0]
for candidateLink in candidateLinks:
# Use Cache. duh.
assignedPage = self.retrieveCachedPage(assignedLink)
candidatePage = self.retrieveCachedPage(candidateLink)
candidateDocumentRelevances = []
currentLinkDocumentRelevances = []
context = [v[0] for k, v in state.items() if k != keyword]
for otherAssignedLink in context:
# get cached page
otherAssignedPage = self.retrieveCachedPage(candidateLink)
# Obtain the Document Relevances as a list
candidateScore = RelevanceModel.documentRelevance(candidatePage, otherAssignedPage)
candidateDocumentRelevances.append(candidateScore)
currentLinkScore = RelevanceModel.documentRelevance(candidatePage, assignedPage)
currentLinkDocumentRelevances.append(currentLinkScore)
# Obtain the LinkRelevances
assignedPage = self.retrieveCachedPage(assignedLink)
candidatePage = self.retrieveCachedPage(candidateLink)
currentLinkRelevance = RelevanceModel.linkRelevance(self.keywords, assignedPage)
candidateLinkRelevance = RelevanceModel.linkRelevance(self.keywords, candidatePage)
# Obtain a convex combination of the link relevances and the
# sum of the document relevances
candidatePsi = (1 - alpha)*candidateLinkRelevance + alpha*sum(candidateDocumentRelevances)
currentPsi = (1 - alpha)*currentLinkRelevance + alpha*sum(currentLinkDocumentRelevances)
print
print "Current Psi Value " + str(currentPsi) + " For " + assignedLink
print "Candidate Link Psi Value " + str(candidatePsi) + " For " + candidateLink
# If the candidate link's convex combination is greater than the current
# link's convex combination, we replace that assignment
if candidatePsi < currentPsi:
state[keyword] = (candidateLink, candidateLinkRelevance)
print "Replaced Link"
print str(i + 1) + "/" + str(iterations) + " iterations complete"
return state
def getMaxRelevance(self, keyword):
"""For every possible candidate link for a keyword, obtain the link's
relevance score to the keyword's context and return the highest
scoring link along with its relevance score"""
relevances = []
candidateLinks = [k[0] for k in self.cachedPages[keyword]]
for candidateLink in candidateLinks:
score = RelevanceModel.linkRelevance(self.keywords, candidateLink)
relevances.append((candidateLink, score))
# Returns the tuple with the highest relevance score in the array
return max(relevances, key=lambda x:x[1])