def ParseParaStr(self, ParaStr):
lPara = ParaStr.split(',')
lParaScore = [Para.split('=') for Para in lPara]
hPara = dict(lParaScore)
conf = cxConfC()
conf.hConf = hPara
return conf
def ParseParaStr(self,ParaStr):
lPara = ParaStr.split(',')
lParaScore = [Para.split('=') for Para in lPara]
hPara = dict(lParaScore)
conf = cxConfC()
conf.hConf = hPara
return conf
def SetConf(self,ConfIn):
conf = cxConfC(ConfIn)
self.InQuery = conf.GetConf('in')
self.InStemQuery = conf.GetConf('stemmedq',self.InQuery)
self.MinPhraseLen = int(conf.GetConf('minphraselen',self.MinPhraseLen))
self.OutName = conf.GetConf('out')
self.NumOfObjPerQ = int(conf.GetConf('numofobj',self.NumOfObjPerQ))
def SetConf(self,ConfIn):
cxBaseC.SetConf(self, ConfIn)
conf = cxConfC(ConfIn)
self.lQRelIn = conf.GetConf("qrel",self.lQRelIn)
if type(self.lQRelIn) != list:
self.lQRelIn = [self.lQRelIn]
self.Depth = int(conf.GetConf("evadepth",self.Depth))
# self.IndriSearcher.SetConf(ConfIn)
self.Prepare()
return True
def SetConf(self,ConfIn):
conf = cxConfC(ConfIn)
self.DocTextDir = conf.GetConf('doctextdir')
self.LoadDocText()
self.ObjCenter.SetConf(ConfIn)
CtfInName = conf.GetConf('objctf')
self.CtfCenter.Load(CtfInName)
def Init(self):
cxBaseC.Init(self)
self.Evaluator = AdhocEvaC()
self.Searcher = IndriSearchCenterC()
self.Word2VecInName = ""
self.Word2VecModel = None
self.cLmName = "kde"
self.LmClass = KernelDensityLmC
#lm conf input
self.ParaConf = cxConfC()
def Process(self,TrainQueryIn,TestQueryIn,ParaStr, EvaOutName):
logging.info('training using [%s] testing using [%s] eva out to [%s]',TrainQueryIn,TestQueryIn,EvaOutName)
conf = cxConfC()
conf.ParseParaStr(ParaStr)
self.DataDir = conf.GetConf('datadir',self.DataDir)
OptMethod = conf.GetConf('optmethod','BFGS')
ConvergeThreshold = conf.GetConf('convergethreshold',1e-05)
logging.info('pipe start training')
llTrainQDocData = self.ReadTargetQDocData(TrainQueryIn,self.DataDir)
llTestQDocData = self.ReadTargetQDocData(TestQueryIn,self.DataDir)
w = self.Learner.Train(llTrainQDocData, OptMethod,ConvergeThreshold)
logging.info('trained w:\n%s',np.array2string(w,precision=6))
logging.info('pipe start testing')
lQid = [line.split('\t')[0] for line in open(TestQueryIn).read().splitlines()]
llDocScore = [ [[data.DocNo, data.X.dot(w)] for data in lTestQDocData] for lTestQDocData in llTestQDocData]
logging.info('pipe start evaluating')
lEvaRes = []
for qid,lDocScore in zip(lQid,llDocScore):
lDocScore.sort(key=lambda item:item[1],reverse = True)
lDocNo = [item[0] for item in lDocScore]
EvaRes = self.Evaluator.EvaluatePerQ(qid, "", lDocNo)
lEvaRes.append(EvaRes)
MeanEvaRes = AdhocMeasureC.AdhocMeasureMean(lEvaRes)
lEvaRes.append(MeanEvaRes)
lQid.append('mean')
out = open(EvaOutName,'w')
for qid,EvaRes in zip(lQid,lEvaRes):
print >>out, qid + '\t' + EvaRes.dumps()
out.close()
logging.info('finished, eva res [%s]',lEvaRes[-1].dumps())
return True
def SetConf(self,ConfIn):
conf = cxConfC(ConfIn)
self.CtfCenter.Load(conf.GetConf('termctf'))
self.ObjCenter.SetConf(ConfIn)
# self.QObjRankName = conf.GetConf('objrank')
self.QFaccObjRankName = conf.GetConf('faccrank')
self.QGoogleObjRankName = conf.GetConf('googlerank')
self.ObjRankDepth = int(conf.GetConf('objrankdepth',self.ObjRankDepth))
self.InName = conf.GetConf('in')
self.OutName = conf.GetConf('out')
def AdhocEvaUnitTest(ConfIn = ""):
#UnitTest add hoc eva
#input: trec type input + qrel
#output: evaluation result
if "" == ConfIn:
print "conf:\nin\nqrel\nevadepth\nout\n"
return False
conf = cxConfC(ConfIn)
InName = conf.GetConf('in')
OutName = conf.GetConf('out')
AdhocEva = AdhocEvaC(ConfIn)
MeanRes = AdhocMeasureC()
cnt = 0
CurrentQid = ""
lDocNo = []
out = open(OutName,'w')
for line in open(InName):
line = line.strip()
vCol = line.split()
ThisQid = vCol[0]
DocNo = vCol[2]
if CurrentQid == "":
CurrentQid = ThisQid
if CurrentQid != ThisQid:
EvaRes = AdhocEva.EvaluatePerQ(CurrentQid, lDocNo)
OutStr = CurrentQid + " %s" %(EvaRes.dumps())
#this is bad
MeanRes = MeanRes + EvaRes
cnt += 1
print >> out, OutStr
CurrentQid = ThisQid
lDocNo = []
lDocNo.append(DocNo)
EvaRes = AdhocEva.EvaluatePerQ(CurrentQid, lDocNo)
OutStr = CurrentQid + " %s" %(EvaRes.dumps())
MeanRes = MeanRes + EvaRes
cnt += 1
print >> out, OutStr
MeanRes = MeanRes / cnt
print >> out,"mean %s" %(MeanRes.dumps())
out.close()
return True
if __name__=='__main__':
import sys,os
from AdhocEva.RankerEvaluator import RankerEvaluatorC
if 2 != len(sys.argv):
print 'I evaluate Boe exp model '
print 'in\nout'
BoePRFRerankerC.ShowConf()
RankerEvaluatorC.ShowConf()
sys.exit()
root = logging.getLogger()
root.setLevel(logging.DEBUG)
ch = logging.StreamHandler(sys.stdout)
# ch.setLevel(logging.DEBUG)
formatter = logging.Formatter('%(asctime)s - %(name)s - %(levelname)s - %(message)s')
ch.setFormatter(formatter)
root.addHandler(ch)
conf = cxConfC(sys.argv[1])
QIn = conf.GetConf('in')
EvaOut = conf.GetConf('out')
Ranker = BoePRFRerankerC(sys.argv[1])
Evaluator = RankerEvaluatorC(sys.argv[1])
Evaluator.Evaluate(QIn, Ranker.Rank, EvaOut)
for line in open(SVMInName):
LeToRData = LeToRDataBaseC(line.strip())
LeToRData = self.ProcessOneInstance(LeToRData)
print >> out, LeToRData.dumps()
out.close()
print "finished"
if __name__ == '__main__':
import sys
if 2 != len(sys.argv):
ExtractDocVecFeatureToSVMDataC.ShowConf()
print "in\nout"
sys.exit()
root = logging.getLogger()
root.setLevel(logging.INFO)
ch = logging.StreamHandler(sys.stdout)
ch.setLevel(logging.DEBUG)
formatter = logging.Formatter(
'%(asctime)s - %(name)s - %(levelname)s - %(message)s')
ch.setFormatter(formatter)
root.addHandler(ch)
conf = cxConfC(sys.argv[1])
SVMInName = conf.GetConf('in')
OutName = conf.GetConf('out')
Extractor = ExtractDocVecFeatureToSVMDataC(sys.argv[1])
Extractor.Process(SVMInName, OutName)
def SetConf(self,ConfIn):
conf = cxConfC(ConfIn)
self.FaccDir = conf.GetConf('faccdir')
def SetConf(self,ConfIn):
conf = cxConfC(ConfIn)
# self.MaxP = float(conf.GetConf('maxp',self.MaxP))
self.lName = conf.GetConf('methodname',[])
self.lBaseName = conf.GetConf('baseline',[])
self.lResInName = conf.GetConf('methodevares',[])
