class LexRank: """ Constructs a summary of the input document by extracting most informative sentences. Arguments: directory - A corpus of text files to be summarized. """ def __init__(self, directory): self.graph = TextGraph(directory) def lexR(self, graph): """ Compute the LexRank of the sentences. LexRank of a sentence in the sentence graph is the PageRank of the node representing the sentence. It is a measure of the importance and influence of the sentence in the corpus. Arguments: graph - A networkx graph or digraph. Returns: A dictionary of all the nodes with their PageRank scores. """ pr = nx.pagerank_numpy(graph, alpha=0.85) return pr def summary(self, compression = 0.25): """ Builds the summary based on the LexRank scores of the sentences. Arguments: compression : A number in [0,1] which is equal to the fraction of total sentences to be included in the summary. Default value is 0.25 Returns: Summary of the input document. """ g = self.graph.sentenceGraph() total_sentences = len(g.nodes()) n_sentences = int(total_sentences * compression) rankings = self.lexR(g) ranked_sentences = sorted(rankings.iteritems(), key=operator.itemgetter(1), reverse=True) summary_sentences = "" i = 0 while (i < n_sentences): u,v = ranked_sentences[i] summary_sentences += u i = i + 1 return summary_sentences
class LexRank:
"""
Constructs a summary of the input document by extracting most informative sentences.
Arguments:
directory - A corpus of text files to be summarized.
"""
def __init__(self, directory):
self.graph = TextGraph(directory)
def lexR(self, graph):
"""
Compute the LexRank of the sentences.
LexRank of a sentence in the sentence graph is the PageRank of the node
representing the sentence. It is a measure of the importance and influence
of the sentence in the corpus.
Arguments:
graph - A networkx graph or digraph.
Returns:
A dictionary of all the nodes with their PageRank scores.
"""
pr = nx.pagerank_numpy(graph, alpha=0.85)
return pr
def summary(self, compression=0.25):
"""
Builds the summary based on the LexRank scores of the sentences.
Arguments:
compression : A number in [0,1] which is equal to the fraction of total
sentences to be included in the summary.
Default value is 0.25
Returns:
Summary of the input document.
"""
g = self.graph.sentenceGraph()
total_sentences = len(g.nodes())
n_sentences = int(total_sentences * compression)
rankings = self.lexR(g)
ranked_sentences = sorted(rankings.iteritems(),
key=operator.itemgetter(1),
reverse=True)
summary_sentences = ""
i = 0
while (i < n_sentences):
u, v = ranked_sentences[i]
summary_sentences += u
i = i + 1
return summary_sentences
def __init__(self, directory, cooccuranceThrehsold=1):
self.corpus = directory
self.model = TextGraph(directory)
self.graph = self.model.keywordGraph(cooccuranceThrehsold)
self.docFreqDict = self.model.docFreq()
class KeyGraph:
"""
Implements the KeyGraph from the keywordGraph.
KeyGraph is a slight modification of the keywordGraph as defined in the paper ...
KeyGraph is used in extracting topics for the corpus of the documents.
Arguments:
directory : a corpus of text files.
cooccuranceThrehsold : If the cooccurances of two keywords is above
coocuranceThreshold, there is an edge between the nodes represented
by the keywords.
Default value is 1
"""
def __init__(self, directory, cooccuranceThrehsold=1):
self.corpus = directory
self.model = TextGraph(directory)
self.graph = self.model.keywordGraph(cooccuranceThrehsold)
self.docFreqDict = self.model.docFreq()
def probCoocurance(self, keyword1, keyword2):
"""
Compute the probability of cooourance of a keyword in a document
given a given keyword occurs in the document.
Arguments:
keyword1 : first keyword.
keyword2 : second keyword.
Returns:
Numerical value of the probability.
"""
documentFrequencyDict = self.docFreqDict
docsContainingKeyword1 = set(documentFrequencyDict[keyword1])
docsContainingKeyword2 = set(documentFrequencyDict[keyword2])
commonDocs = len(docsContainingKeyword1 & docsContainingKeyword2)
prob = commonDocs / len(docsContainingKeyword1)
return prob
def network(self, docFreqThreshold=2, probThreshold=0.01):
"""
Compute the KeyGraph.
Arguments:
docFreqThreshold : If the documnet frequency of a keyword is below
the docFreqThreshold, then the node corresponding
to the keyword is removed from the graph.
Default value is 1.
"""
g = self.graph
allNodes = g.nodes()
allEdges = g.edges()
documentFrequency = self.docFreqDict
for x in allNodes:
if len(documentFrequency[x]) < docFreqThreshold:
g.remove_node(x)
for e in allEdges:
node1 = e[0]
node2 = e[1]
if (self.probCoocurance(node1, node2) < probThreshold) or (
self.probCoocurance(node2, node1) < probThreshold
):
g.remove_edge(node1, node2)
return g
from TextGraphics.src.graph import TextGraph from TextGraphics.Analysis.plotting import Pictures from TextGraphics.Analysis.properties import Analysis from TextGraphics.Applications.summary import LexRank directory = 'Data' g = TextGraph(directory) senGraph = g.sentenceGraph() node = senGraph.nodes()[0] out = Pictures(senGraph) out.graphPlot(0.3, labelingByNumbers=True) out1 = Analysis(senGraph) l = out1.nodeInConnectedComponent(node) print l.nodes() cS = out1.centralNodes(4) print cS lR = LexRank(directory) print lR.summary()
def __init__(self, directory, cooccuranceThrehsold=1):
self.corpus = directory
self.model = TextGraph(directory)
self.graph = self.model.keywordGraph(cooccuranceThrehsold)
self.docFreqDict = self.model.docFreq()
class KeyGraph:
"""
Implements the KeyGraph from the keywordGraph.
KeyGraph is a slight modification of the keywordGraph as defined in the paper ...
KeyGraph is used in extracting topics for the corpus of the documents.
Arguments:
directory : a corpus of text files.
cooccuranceThrehsold : If the cooccurances of two keywords is above
coocuranceThreshold, there is an edge between the nodes represented
by the keywords.
Default value is 1
"""
def __init__(self, directory, cooccuranceThrehsold=1):
self.corpus = directory
self.model = TextGraph(directory)
self.graph = self.model.keywordGraph(cooccuranceThrehsold)
self.docFreqDict = self.model.docFreq()
def probCoocurance(self, keyword1, keyword2):
"""
Compute the probability of cooourance of a keyword in a document
given a given keyword occurs in the document.
Arguments:
keyword1 : first keyword.
keyword2 : second keyword.
Returns:
Numerical value of the probability.
"""
documentFrequencyDict = self.docFreqDict
docsContainingKeyword1 = set(documentFrequencyDict[keyword1])
docsContainingKeyword2 = set(documentFrequencyDict[keyword2])
commonDocs = len(docsContainingKeyword1 & docsContainingKeyword2)
prob = commonDocs / len(docsContainingKeyword1)
return prob
def network(self, docFreqThreshold=2, probThreshold=0.01):
"""
Compute the KeyGraph.
Arguments:
docFreqThreshold : If the documnet frequency of a keyword is below
the docFreqThreshold, then the node corresponding
to the keyword is removed from the graph.
Default value is 1.
"""
g = self.graph
allNodes = g.nodes()
allEdges = g.edges()
documentFrequency = self.docFreqDict
for x in allNodes:
if (len(documentFrequency[x]) < docFreqThreshold):
g.remove_node(x)
for e in allEdges:
node1 = e[0]
node2 = e[1]
if (self.probCoocurance(node1, node2) < probThreshold) or (
self.probCoocurance(node2, node1) < probThreshold):
g.remove_edge(node1, node2)
return g
from TextGraphics.src.graph import TextGraph from TextGraphics.Analysis.plotting import Pictures from TextGraphics.Analysis.properties import Analysis from TextGraphics.Applications.summary import LexRank ## Import the corpus of text files. directory = 'Data' ## Create the sentence graph g = TextGraph(directory) senGraph = g.sentenceGraph() node = senGraph.nodes()[0] ## Plot the sentence graph out = Pictures(senGraph) out.graphPlot(0.3, labelingByNumbers=True) # Compute the keyword graph kwgraph = g.keywordGraph() print len(kwgraph.nodes()) ## Plot the keyword graph out = Pictures(kwgraph) out.graphPlot(0.3, labelingByNumbers=True) #### Analysis out1 = Analysis(senGraph) # Find the connected component of a node. l = out1.nodeInConnectedComponent(node) print l.nodes()
def __init__(self, directory):
self.graph = TextGraph(directory)
def __init__(self, directory): self.graph = TextGraph(directory)
from TextGraphics.src.graph import TextGraph from TextGraphics.Analysis.plotting import Pictures from TextGraphics.Analysis.properties import Analysis from TextGraphics.Applications.summary import LexRank ## Import the corpus of text files. directory = "Data" ## Create the sentence graph g = TextGraph(directory) senGraph = g.sentenceGraph() node = senGraph.nodes()[0] ## Plot the sentence graph out = Pictures(senGraph) out.graphPlot(0.3, labelingByNumbers=True) # Compute the keyword graph kwgraph = g.keywordGraph() print len(kwgraph.nodes()) ## Plot the keyword graph out = Pictures(kwgraph) out.graphPlot(0.3, labelingByNumbers=True) #### Analysis out1 = Analysis(senGraph) # Find the connected component of a node.
