-
Notifications
You must be signed in to change notification settings - Fork 0
/
main.py
301 lines (251 loc) · 9.23 KB
/
main.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
from sentenceSegmentation import SentenceSegmentation
from tokenization import Tokenization
from inflectionReduction import InflectionReduction
from stopwordRemoval import StopwordRemoval
from informationRetrieval_ESA import InformationRetrieval
from evaluation import Evaluation
import time
from sys import version_info
import argparse
import json
import matplotlib.pyplot as plt
import csv
# name of csv file
filename = "evalData_ESA.csv"
t1 = time.perf_counter()
# Input compatibility for Python 2 and Python 3
if version_info.major == 3:
pass
elif version_info.major == 2:
try:
input = raw_input
except NameError:
pass
else:
print ("Unknown python version - input function not safe")
class SearchEngine:
def __init__(self, args):
self.args = args
self.tokenizer = Tokenization()
self.sentenceSegmenter = SentenceSegmentation()
self.inflectionReducer = InflectionReduction()
self.stopwordRemover = StopwordRemoval()
self.informationRetriever = InformationRetrieval()
self.evaluator = Evaluation()
def segmentSentences(self, text):
"""
Call the required sentence segmenter
"""
if self.args.segmenter == "naive":
return self.sentenceSegmenter.naive(text)
elif self.args.segmenter == "punkt":
return self.sentenceSegmenter.punkt(text)
def tokenize(self, text):
"""
Call the required tokenizer
"""
if self.args.tokenizer == "naive":
return self.tokenizer.naive(text)
elif self.args.tokenizer == "ptb":
return self.tokenizer.pennTreeBank(text)
def reduceInflection(self, text):
"""
Call the required stemmer/lemmatizer
"""
return self.inflectionReducer.reduce(text)
def removeStopwords(self, text):
"""
Call the required stopword remover
"""
return self.stopwordRemover.fromList(text)
def preprocessQueries(self, queries):
"""
Preprocess the queries - segment, tokenize, stem/lemmatize and remove stopwords
"""
# Segment queries
segmentedQueries = []
for query in queries:
segmentedQuery = self.segmentSentences(query)
segmentedQueries.append(segmentedQuery)
json.dump(segmentedQueries, open(self.args.out_folder + "segmented_queries.txt", 'w'))
# Tokenize queries
tokenizedQueries = []
for query in segmentedQueries:
tokenizedQuery = self.tokenize(query)
tokenizedQueries.append(tokenizedQuery)
json.dump(tokenizedQueries, open(self.args.out_folder + "tokenized_queries.txt", 'w'))
# Stem/Lemmatize queries
reducedQueries = []
for query in tokenizedQueries:
reducedQuery = self.reduceInflection(query)
reducedQueries.append(reducedQuery)
json.dump(reducedQueries, open(self.args.out_folder + "reduced_queries.txt", 'w'))
# Remove stopwords from queries
stopwordRemovedQueries = []
for query in reducedQueries:
stopwordRemovedQuery = self.removeStopwords(query)
stopwordRemovedQueries.append(stopwordRemovedQuery)
json.dump(stopwordRemovedQueries, open(self.args.out_folder + "stopword_removed_queries.txt", 'w'))
preprocessedQueries = stopwordRemovedQueries
return preprocessedQueries
def preprocessDocs(self, docs):
"""
Preprocess the documents
"""
# Segment docs
segmentedDocs = []
for doc in docs:
segmentedDoc = self.segmentSentences(doc)
segmentedDocs.append(segmentedDoc)
json.dump(segmentedDocs, open(self.args.out_folder + "segmented_docs.txt", 'w'))
# Tokenize docs
tokenizedDocs = []
for doc in segmentedDocs:
tokenizedDoc = self.tokenize(doc)
tokenizedDocs.append(tokenizedDoc)
json.dump(tokenizedDocs, open(self.args.out_folder + "tokenized_docs.txt", 'w'))
# Stem/Lemmatize docs
reducedDocs = []
for doc in tokenizedDocs:
reducedDoc = self.reduceInflection(doc)
reducedDocs.append(reducedDoc)
json.dump(reducedDocs, open(self.args.out_folder + "reduced_docs.txt", 'w'))
# Remove stopwords from docs
stopwordRemovedDocs = []
for doc in reducedDocs:
stopwordRemovedDoc = self.removeStopwords(doc)
stopwordRemovedDocs.append(stopwordRemovedDoc)
json.dump(stopwordRemovedDocs, open(self.args.out_folder + "stopword_removed_docs.txt", 'w'))
preprocessedDocs = stopwordRemovedDocs
return preprocessedDocs
def evaluateDataset(self):
"""
- preprocesses the queries and documents, stores in output folder
- invokes the IR system
- evaluates precision, recall, fscore, nDCG and MAP
for all queries in the Cranfield dataset
- produces graphs of the evaluation metrics in the output folder
"""
# Read queries
queries_json = json.load(open(args.dataset + "cran_queries.json", 'r'))[:]
query_ids, queries = [item["query number"] for item in queries_json], \
[item["query"] for item in queries_json]
# Process queries
processedQueries = self.preprocessQueries(queries)
# Read documents
docs_json = json.load(open(args.dataset + "cran_docs.json", 'r'))[:]
doc_ids, docs, doc_titles = [item["id"] for item in docs_json], \
[item["body"] for item in docs_json], \
[item["title"] for item in docs_json]
# Process documents
processedDocs = self.preprocessDocs(docs)
# Build document index
self.informationRetriever.buildIndex(processedDocs, doc_ids,doc_titles)
# Rank the documents for each query
doc_IDs_ordered = self.informationRetriever.rank(processedQueries)
# Read relevance judements
qrels = json.load(open(args.dataset + "cran_qrels.json", 'r'))[:]
# Calculate precision, recall, f-score, MAP and nDCG for k = 1 to 10
precisions, recalls, fscores, MAPs, nDCGs = [], [], [], [], []
APdatas,nDCGdatas = {},{}
for k in range(1, 11):
precision = self.evaluator.meanPrecision(
doc_IDs_ordered, query_ids, qrels, k)
precisions.append(precision)
recall = self.evaluator.meanRecall(
doc_IDs_ordered, query_ids, qrels, k)
recalls.append(recall)
fscore = self.evaluator.meanFscore(
doc_IDs_ordered, query_ids, qrels, k)
fscores.append(fscore)
print("Precision, Recall and F-score @ " +
str(k) + " : " + str(precision) + ", " + str(recall) +
", " + str(fscore))
MAP,APdata = self.evaluator.meanAveragePrecision(
doc_IDs_ordered, query_ids, qrels, k)
MAPs.append(MAP)
APdatas[k] = APdata
nDCG,nDCGdata = self.evaluator.meanNDCG(
doc_IDs_ordered, query_ids, qrels, k)
nDCGs.append(nDCG)
nDCGdatas[k] = nDCGdata
print("MAP, nDCG @ " +
str(k) + " : " + str(MAP) + ", " + str(nDCG))
# Plot the metrics and save plot
plt.plot(range(1, 11), precisions, label="Precision")
plt.plot(range(1, 11), recalls, label="Recall")
plt.plot(range(1, 11), fscores, label="F-Score")
plt.plot(range(1, 11), MAPs, label="MAP")
plt.plot(range(1, 11), nDCGs, label="nDCG")
plt.legend()
plt.title("Evaluation Metrics - Cranfield Dataset")
plt.xlabel("k")
plt.savefig(args.out_folder + "eval_plot.png")
rows = []
for query_id in query_ids:
row = [query_id]
for k in range(1, 11):
row.append(nDCGdatas[k][query_id])
for k in range(1, 11):
row.append(APdatas[k][query_id])
rows.append(row)
# field names
fields = ['Doc ID']+['nDCG@{}'.format(k) for k in range(1,11)]+['AP@{}'.format(k) for k in range(1,11)]
# writing to csv file
with open(filename, 'w',newline='') as csvfile:
# creating a csv writer object
csvwriter = csv.writer(csvfile)
# writing the fields
csvwriter.writerow(fields)
# writing the data rows
csvwriter.writerows(rows)
def handleCustomQuery(self):
"""
Take a custom query as input and return top five relevant documents
"""
#Get query
print("Enter query below")
query = input()
# Process documents
processedQuery = self.preprocessQueries([query])[0]
# Read documents
docs_json = json.load(open(args.dataset + "cran_docs.json", 'r'))[:]
doc_ids, docs, doc_titles = [item["id"] for item in docs_json], \
[item["body"] for item in docs_json], \
[item["title"] for item in docs_json]
# Process documents
processedDocs = self.preprocessDocs(docs)
# Build document index
self.informationRetriever.buildIndex(processedDocs, doc_ids,doc_titles)
# Rank the documents for the query
doc_IDs_ordered = self.informationRetriever.rank([processedQuery])[0]
# Print the IDs of first five documents
print("\nTop five document IDs : ")
for id_ in doc_IDs_ordered[:5]:
print(id_,":",doc_titles[int(id_)-1])
print('start')
if __name__ == "__main__":
# Create an argument parser
parser = argparse.ArgumentParser(description='main.py')
# Tunable parameters as external arguments
parser.add_argument('-dataset', default = "cranfield/",
help = "Path to the dataset folder")
parser.add_argument('-out_folder', default = "output/",
help = "Path to output folder")
parser.add_argument('-segmenter', default = "punkt",
help = "Sentence Segmenter Type [naive|punkt]")
parser.add_argument('-tokenizer', default = "ptb",
help = "Tokenizer Type [naive|ptb]")
parser.add_argument('-custom', action = "store_true",
help = "Take custom query as input")
# Parse the input arguments
args = parser.parse_args()
# Create an instance of the Search Engine
searchEngine = SearchEngine(args)
# Either handle query from user or evaluate on the complete dataset
if args.custom:
searchEngine.handleCustomQuery()
else:
searchEngine.evaluateDataset()
t2 = time.perf_counter()
print('Runtime = {} minutes'.format((t2-t1)/60))