def xQuAD(basepath,topicID,cutoff= 50):
"""
return a ranking list using the framework of xQuAD.
"""
document_selected =[] #docs ranking list
out_file = open(basepath+"rank/"+topicID+"/xquad_result","w")
runlist_file = open(basepath+"rank/"+topicID+"/runlist","w")
print >>runlist_file,"xquad_result"
runlist_file.close()
documents,subtopicID_word,query,document_term_frequency,average_document_len,word2id_weight = data_preprocessing_xquad(basepath,topicID)
sid_doc_related_probility = similarity_computing(documents,subtopicID_word,average_document_len,query,document_term_frequency)
#iteratively selected the best document to build ranking list
for i in range(cutoff):
print "%s-th best document."%i
best_document = best_document_select(documents,subtopicID_word,query,document_term_frequency,average_document_len,word2id_weight,document_selected,sid_doc_related_probility)
document_selected.append(best_document)
if len(document_selected) == len(documents):
# all document in ranking list
break
# print the ranking result to the evaluation files
for index,document in enumerate(document_selected):
documentID = document.get_id()
document_rank = index+1
document_score = document.get_ranking_score()
print >>out_file,topicID +" Q0 "+documentID+" "+str(document_rank) +" "+str(document_score)+ " TESTRUN"
out_file.close()
# call the evaluation function to evaluate the run result
r = call_eval_for_result(basepath,topicID,10)
def run_ranking(basepath,topics,sparse_parameters):
for topicID in topics:
if topicID == "10001":continue
runlist = open(basepath+"rank/"+topicID+"/runlist","w")
sparse_base_ranking(basepath,topicID,sparse_parameters,runlist)
runlist.close()
r = call_eval_for_result(basepath,topicID,10)
def run_ranking(basepath, topics, sparse_parameters):
for topicID in topics:
if topicID == "10001": continue
runlist = open(basepath + "rank/" + topicID + "/runlist", "w")
sparse_base_ranking(basepath, topicID, sparse_parameters, runlist)
runlist.close()
r = call_eval_for_result(basepath, topicID, 10)
def MMR_ranking(basepath, topicID, cutoff):
""
docs, subtopicID_word, topic_words, document_term_frequency, average_document_len, word2id_weight = data_preprocessing_mmr(
basepath, topicID)
query = topic_words
first_score = -1000.0
first_doc_id = -1
document_collection = []
doc_selected = []
#prepare the out files for evaluation
out_file = open(basepath + "rank/" + topicID + "/mmr_result", "w")
runlist_file = open(basepath + "rank/" + topicID + "/runlist", "w")
print >> runlist_file, "mmr_result"
runlist_file.close()
#select the first document by relevance
for dindex, doc in enumerate(docs):
sim = similarity(query, doc.get_term_vec(), document_term_frequency,
average_document_len, len(docs))
sum_weight = 0.0
vec = [0.0 for i in range(len(word2id_weight))]
for word in doc.get_term_vec():
wid = word2id_weight[word][0]
vec[wid] = word2id_weight[word][1] * 1.0
sum_weight += word2id_weight[word][1] * 1.0
if sum_weight > 0:
doc.set_tfidf_vec([vec[i] / sum_weight for i in range(len(vec))])
else:
doc.set_tfidf_vec(vec)
#doc.set_ranking_score(0.0)
if first_score < sim:
first_doc_id = dindex
first_score = sim
docs[first_doc_id].set_ranking_score(sim)
doc_selected.append(docs[first_doc_id])
sim_dict = similar_documents(docs)
#iteratively selected the best document to build ranking list
for i in range(cutoff):
print "%2d-th document selecting for ranking" % (i + 1)
doc_selected.append(select_best_documnt(docs, doc_selected, sim_dict))
if len(doc_selected) == len(docs):
# all document in ranking list
break
# print the ranking result to the evaluation files
for index, document in enumerate(doc_selected):
documentID = document.get_id()
document_rank = index + 1
document_score = document.get_ranking_score()
print >> out_file, topicID + " Q0 " + documentID + " " + str(
document_rank) + " " + str(document_score) + " TESTRUN"
out_file.close()
# call the evaluation function to evaluate the run result
r = call_eval_for_result(basepath, topicID, 10)
def MMR_ranking(basepath,topicID,cutoff):
""
docs,subtopicID_word,topic_words,document_term_frequency,average_document_len,word2id_weight= data_preprocessing_mmr(basepath,topicID)
query = topic_words
first_score = -1000.0
first_doc_id = -1
document_collection = []
doc_selected = []
#prepare the out files for evaluation
out_file = open(basepath+"rank/"+topicID+"/mmr_result","w")
runlist_file = open(basepath+"rank/"+topicID+"/runlist","w")
print >>runlist_file,"mmr_result"
runlist_file.close()
#select the first document by relevance
for dindex,doc in enumerate(docs):
sim = similarity(query,doc.get_term_vec(),document_term_frequency,average_document_len,len(docs))
sum_weight = 0.0
vec = [0.0 for i in range(len(word2id_weight))]
for word in doc.get_term_vec():
wid = word2id_weight[word][0]
vec[wid] = word2id_weight[word][1]*1.0
sum_weight += word2id_weight[word][1]*1.0
if sum_weight>0:
doc.set_tfidf_vec([vec[i]/sum_weight for i in range(len(vec))])
else:
doc.set_tfidf_vec(vec)
#doc.set_ranking_score(0.0)
if first_score< sim:
first_doc_id = dindex
first_score = sim
docs[first_doc_id].set_ranking_score(sim)
doc_selected.append(docs[first_doc_id])
sim_dict = similar_documents(docs)
#iteratively selected the best document to build ranking list
for i in range(cutoff):
print "%2d-th document selecting for ranking"%(i+1)
doc_selected.append(select_best_documnt(docs,doc_selected,sim_dict))
if len(doc_selected) == len(docs):
# all document in ranking list
break
# print the ranking result to the evaluation files
for index,document in enumerate(doc_selected):
documentID = document.get_id()
document_rank = index+1
document_score = document.get_ranking_score()
print >>out_file,topicID +" Q0 "+documentID+" "+str(document_rank) +" "+str(document_score)+ " TESTRUN"
out_file.close()
# call the evaluation function to evaluate the run result
r = call_eval_for_result(basepath,topicID,10)
def run_ranking(basepath, topics, methods):
for topicID in topics:
for method in methods:
if method == "mine":
filename1 = "xquad_mine_less"
filename2 = "xquad_mine_all"
else:
filename1 = "xquad_standard_less"
filename2 = "xquad_standard_all"
xQuAD(basepath, topicID, filename1, 30, method)
xQuAD(basepath, topicID, filename2, 30, method)
r = call_eval_for_result(basepath, topicID, 10)
def run_ranking(basepath,topics,methods):
for topicID in topics:
for method in methods:
if method =="mine":
filename1= "xquad_mine_less"
filename2 = "xquad_mine_all"
else:
filename1= "xquad_standard_less"
filename2 = "xquad_standard_all"
xQuAD(basepath,topicID,filename1,30,method)
xQuAD(basepath,topicID,filename2,30,method)
r = call_eval_for_result(basepath,topicID,10)
def xQuAD(basepath, topicID, cutoff=50):
"""
return a ranking list using the framework of xQuAD.
"""
document_selected = [] #docs ranking list
out_file = open(basepath + "rank/" + topicID + "/xquad_result", "w")
runlist_file = open(basepath + "rank/" + topicID + "/runlist", "w")
print >> runlist_file, "xquad_result"
runlist_file.close()
documents, subtopicID_word, query, document_term_frequency, average_document_len, word2id_weight = data_preprocessing_xquad(
basepath, topicID)
sid_doc_related_probility = similarity_computing(documents,
subtopicID_word,
average_document_len,
query,
document_term_frequency)
#iteratively selected the best document to build ranking list
for i in range(cutoff):
print "%s-th best document." % i
best_document = best_document_select(documents, subtopicID_word, query,
document_term_frequency,
average_document_len,
word2id_weight, document_selected,
sid_doc_related_probility)
document_selected.append(best_document)
if len(document_selected) == len(documents):
# all document in ranking list
break
# print the ranking result to the evaluation files
for index, document in enumerate(document_selected):
documentID = document.get_id()
document_rank = index + 1
document_score = document.get_ranking_score()
print >> out_file, topicID + " Q0 " + documentID + " " + str(
document_rank) + " " + str(document_score) + " TESTRUN"
out_file.close()
# call the evaluation function to evaluate the run result
r = call_eval_for_result(basepath, topicID, 10)
#-*- coding:utf8 -*-
import os
import sys
sys.path.append("..")
from result.result_analysis import DR_result_analysis
from util.evaluation import call_eval_for_result
if __name__=="__main__":
print "start..."
files = os.listdir("/users/songwei/xuwenbin/diversification/ntcir09/rank/")
for file_name in files:
out1 = open("/users/songwei/xuwenbin/diversification/ntcir09/rank/"+file_name+"/runlist","w")
out = open("/users/songwei/xuwenbin/diversification/ntcir09/rank/"+file_name+"/ideal","w")
print >>out1,"ideal"
with open("/users/songwei/xuwenbin/diversification/ntcir09/rank/"+file_name+"/ideal.rank") as lines:
for line in lines:
if line.startswith("0"):
print>>out,line.replace("\n","")
out.close()
out1.close()
call_eval_for_result("/users/songwei/xuwenbin/diversification/ntcir09/",file_name,10)
DR_result_analysis("/users/songwei/xuwenbin/diversification/ntcir09/","","ideal")
