def getResultsTimeSpan(topNResults, tweetsEpoch, queriesEpoch):
resultsTimeSpan = {}
for qid in topNResults.keys():
queryEpoch = queriesEpoch[qid]
resultsTimeSpan.setdefault(qid, list())
docDict = topNResults[qid]
for docId in docDict.keys():
tweetEpoch = tweetsEpoch[docId]
if tweetEpoch <= queryEpoch:
timeSpan = getTimeSpan(queryEpoch, tweetEpoch )
resultsTimeSpan[qid].append(timeSpan)
return resultsTimeSpan
def getQwordTimeSpan(qid, qword, topNResults, wordsIndex, tweetsEpoch, queriesEpoch):
qwordTimeSpans = list()
queryEpoch = queriesEpoch[qid]
qidTopNDocs = topNResults[qid].keys()
qwordAllDocs = wordsIndex[qword]
for tweetId in qidTopNDocs:
if qwordAllDocs.has_key(tweetId):
tweetEpoch = tweetsEpoch[tweetId]
if tweetEpoch <= queryEpoch:
timeSpan = getTimeSpan(queryEpoch, tweetEpoch )
key = qid + '_' + qword
qwordTimeSpans.append(timeSpan)
return qwordTimeSpans
def dtfWeightingT(queriesEpoch, tweetsEpoch, kdeQwordDict, qid, qword, docId, wordsIndex, docsLength, avgDocsLength, k1, b):
kde = kdeQwordDict[qid + '_' + qword]
queryEpoch = queriesEpoch[qid]
tweetEpoch = tweetsEpoch[docId]
if queryEpoch >= tweetEpoch:
timeSpan = getTimeSpan(queryEpoch, tweetEpoch )
probDens = prediction(kde, [timeSpan])
probDen = probDens[0]
count = wordsIndex[qword][docId]
docLength = docsLength[docId]
numerator = (k1 + 1) * count * probDen
denominator = k1 * (1 - b + b * docLength / avgDocsLength) + count * probDen
dtf = 1.0 * numerator / denominator
else:
dtf = 0
return dtf
def predictResultsProbDens(retrievalResults, queriesEpoch, tweetsEpoch, kdeDict):
probDens = {}
for qid in retrievalResults.keys():
probDens.setdefault(qid, {})
kde = kdeDict[qid]
queryEpoch = queriesEpoch[qid]
resultsList = retrievalResults[qid]
for result in resultsList:
docId = result.docId
tweetEpoch = tweetsEpoch[docId]
if queryEpoch >= tweetEpoch:
timeSpan = getTimeSpan(queryEpoch, tweetEpoch )
probDen = prediction(kde, [timeSpan])
probDens[qid][docId] = probDen[0]
else:
probDens[qid][docId] = 0
print qid
return probDens
