class Summarizer:
def __init__(self):
self.parser = Parser()
'''
test : it is catchphrase, one sentence.
title: sentence list.
'''
def summarize(self, text, title):
sentences = text
result = []
## step 1, get term list of catchphrase.
(catchphrase_keywords, catchphrase_wordCount) = self.parser.getKeywords(title)
result.append( (catchphrase_keywords, catchphrase_wordCount) )
catchword_list = [catchphrase_keywords[idx]['word'] for idx in range(len(catchphrase_keywords))]
#print("[*catchword_list*]",catchword_list)
## step 2, get top k word list in sentences.
## 2.1 get term list of detail.
#text_merged = " ".join(sentences)
#(detail_keywords, detail_wordCount) = self.parser.getKeywords(text_merged)
for idx in range(len(text)):
(sentence_keywords, sentence_wordCount) = self.parser.getKeywords(text[idx])
result.append( (sentence_keywords, sentence_wordCount) )
word_list = [sentence_keywords[idx]['word'] for idx in range(len(sentence_keywords))]
#print("\n[*word_list*]", word_list)
return result
def predictBayesianModel(sentenceList=[
'hello world occupation lease', 'machine learning board',
'machine learning lease occupation'
],
input_path="./model/train_model.npz",
word_index_file="./model/word_index.npz"):
#print("\n-----------------------------------------")
#print("Load model1: ./model/train_model.npz")
#print("-----------------------------------------")
#--------------------------------------------------
npzfile = np.load(input_path)
pi_bar = npzfile['arr_0']
theta_bar = npzfile['arr_1']
#print("\n[[pi]]:")
#print(pi_bar)
#print(pi_bar.shape)
#print("\n[[theta]]:")
#print(theta_bar)
#print(theta_bar.shape)
#print("\n-----------------------------------------")
#print("Load model2: ./model/word_index.npz")
#print("-----------------------------------------")
#--------------------------------------------------
npzfile2 = np.load(word_index_file)
catchword_index = npzfile2['arr_0']
bodyword_index = npzfile2['arr_1']
#print("\n[[catchword index]]:")
#print(catchword_index)
#print(catchword_index.shape)
#print("\n[[bodyword index]]:")
#print(bodyword_index)
#print(bodyword_index.shape)
#--------------------------------------------------
scoreRecord = []
parser = Parser()
catchword_list = catchword_index.tolist()
bodyword_list = bodyword_index.tolist()
#--------------------------------------------------
# Get catchword_positionList
#--------------------------------------------------
catchwords = sentenceList[0]
(keywords, wordCount) = parser.getKeywords(catchwords)
catchword_positionList = []
#print("keywords: ", keywords)
for elem in keywords:
word = elem['word']
count = elem['count']
idx = catchword_list.index(word) if word in catchword_list else -1
if (idx != -1):
#print("appending ", catchword_list[idx])
catchword_positionList.append(idx)
#Debug
#print("catchword_positionList:", catchword_positionList)
#for catchwordPos in catchword_positionList:
#print(catchword_list[catchwordPos])
#--------------------------------------------------
# Calculate score for each word in body sentence.
# The first sentence is catchphrases.
#--------------------------------------------------
for idx in range(1, len(sentenceList)):
(keywords, wordCount) = parser.getKeywords(sentenceList[idx])
sentence_score = 0
'''
1) get the position list of catch words in predicted case.
2) for each word in each sentence, find the scores for each catchword in theta_bar.
3) add these scores which will be the final for one word in this sentence.
4) evaluate next word... until the end of this sentence.
5) goto 2).
'''
## print("----------- sentence --------------")
for elem in keywords:
# Jeff: For each word in body sentence.
word = elem['word']
count = elem['count']
## print("sentence word :", word)
## print("sentence word count:", count)
## print(" ")
word_score = 0
wordInSentence_idx = bodyword_list.index(
word) if word in bodyword_list else -1
if (wordInSentence_idx != -1):
# Jeff: For each word in catchphrase
for catchwordIdx in catchword_positionList:
## print("* theta_bar[",idx, "][", catchword_list[catchwordIdx], "]")
## print("* score:", theta_bar[idx][catchwordIdx])
## print(" ")
word_score += theta_bar[idx][catchwordIdx]
sentence_score += word_score * count
scoreRecord.append(sentence_score)
# NB: sentence ith, from 1 to end.
#print("\nScore list for each sentence:")
#print([ float("%.2f" % elem) for elem in scoreRecord ])
#print("")
return scoreRecord