Python Statistics.importantPara方法代码示例

# 需要导入模块: from Statistics import Statistics [as 别名]
# 或者: from Statistics.Statistics import importantPara [as 别名]

#.........这里部分代码省略.........
     #it may contain duplicates at the end of this loop
     #Duplicates are removed using an elegant method (See:self.removeDuplicates() )
     for paragraph in self.paragraphs:
         sentenses=nltk.sent_tokenize(paragraph)
         tags=[]
         for sentens in sentenses:
             tokens=nltk.word_tokenize(sentens)
             tags.extend(nltk.pos_tag(tokens))
         nouns=self.getNouns(tags)
         filteredNouns=self.removeStopWords(nouns)
         allNouns.extend(filteredNouns)
         nounCountDict=self.getNounsCounts(filteredNouns)
         nounCountDictList.append(nounCountDict)
     allNouns=self.removeDuplicates(allNouns)
     
     #Creates occurrenceVector. See createOccurenceVector() for more details
     occurenceVectorDict=self.createOccurenceVector(allNouns,nounCountDictList)
     weightVectorDict=self.createNounWeightDict(occurenceVectorDict)
     numberOfParagraphs=len(self.paragraphs)
     pointList=[]
     for key in weightVectorDict.keys():
         totalOccurrences=sum(weightVectorDict[key])
         averageCount=totalOccurrences/numberOfParagraphs
         variance=numpy.var(weightVectorDict[key])
         #TODO: have to replace the following line with a better formula that balances mean and variance 
         point=averageCount-variance
         pointList.append((key,point))
     
     #Sort keywords according to weight
     pointList.sort(key=itemgetter(1),reverse=True)
     print(pointList)
     
     #Take most important 10 words
     keyWords=[]
     print("Important words")
     if len(pointList)>10:
         for i in range(0,10):
             keyWords.append(pointList[i][0])
     else:
         for i in range(0,len(pointList)):
             keyWords.append(pointList[i][0])
     
     s=Statistics()
     s.keywords=keyWords
     s.title=keyWords[0]
     
     '''
     Following code implements the paragraph scoring algorithm based of Eigan vectors of similarity matrix
     '''
     #Creating the similarity vector to find rate paragraphs
     similarityMatrix=[];
     for i in range(0,numberOfParagraphs):
         currentRow=[]
         for j in range(0,numberOfParagraphs):
             freq=0
             nounsInThisPara=0
             for k in allNouns:
                 currentOccurenceVector=occurenceVectorDict[k]
                 if currentOccurenceVector[i]>0:
                     nounsInThisPara=nounsInThisPara+1
                 if currentOccurenceVector[i]>0 and currentOccurenceVector[j]>0:
                     freq=freq+1
             if nounsInThisPara==0:
                 similarity=0
             else:
                 similarity=float(freq)/float(nounsInThisPara)
             currentRow.append(similarity)
         similarityMatrix.append(currentRow)
     
     print("Similarity Matrix")
     self.printMatrix(similarityMatrix)
     
     similarityArray=numpy.array(similarityMatrix)
     
     #Calculating eigan values of similarity matrix
     eigenvalues, eigenvectors = numpy.linalg.eig(similarityArray)
     
     #Only for the purpose of printing Eigan values
     print("Eigan Vectors")
     for i in range(0,numberOfParagraphs):
         print(self.paragraphs[i])
         print(eigenvectors[i])
         
     paragraphRatings=[];
     k=0
     for i in eigenvectors:
         count=0
         for j in i:
             if j>0.001: #checking for positive value
                 count=count+1
         paragraphRatings.append((count,k))
         k=k+1
     #The following lines are to locate the most important paragraph
     
     #Sort paragraphs according to rating
     paragraphRatings.sort(key=itemgetter(0),reverse=True)
     
     s.importantPara=self.paragraphs[paragraphRatings[0][1]]
     s.importantParaRating=paragraphRatings[0][0]
     return s