def run_validation(validation_file_path, w2vec_model_file_path, nn_model, result_path, threshold,
vector_dimension):
# loading the pre-trained model
w2vec_model = gensim.models.KeyedVectors.load_word2vec_format(w2vec_model_file_path)
lines = file_processor.read_lines(validation_file_path)
lines.pop(0) # removing column header
qid_queries = get_queries(lines)
counter = 0;
total_queries = len(qid_queries)
for qid in qid_queries:
counter += 1
print("Started->", counter)
query = qid_queries[qid]
pid_passage_relevance_tuple = file_processor.get_candidate_passages_relevance_by_qid(lines, qid)
scored_passage = []
for ppr in pid_passage_relevance_tuple:
passage = ppr["passage"]
relevancy = float(ppr["relevancy"])
feature_vector = word2vec.build_feature_vector(query, passage, w2vec_model, vector_dimension)
score = get_probability_score(nn_model, feature_vector)
pre_relevancy = 1.0 if score >= threshold else 0.0
scored_passage.append(
{"qid": qid, "pid": ppr["pid"], "rank": 0, "score": score, "relevancy": relevancy,
"pre_relevancy": pre_relevancy, "assigment_name": "A1", "algorithm_name": "NN"})
sorted_passage = set_rank_by_score(scored_passage)
file_processor.write_scored_passage(sorted_passage, result_path)
# select top 250 queries
if counter > 250:
break
def get_scored_passage_from_result(result_file_path):
result_lines = file_processor.read_lines(result_file_path)
scored_passage = []
for line in result_lines:
elements = line.split("\t")
qid = elements[0]
pid = elements[2]
rank = int(elements[3])
score = float(elements[4])
relevancy = float(elements[6])
scored_passage.append({
"qid": qid,
"pid": pid,
"rank": rank,
"score": score,
"relevancy": relevancy
})
return scored_passage
def run_bm25_model(validation_file_path, result_file_path):
lines = file_processor.read_lines(validation_file_path)
lines.pop(0) # removing column header
qid_queries = get_queries(lines)
counter = 0
total_queries = len(qid_queries)
for qid in qid_queries:
counter += 1
query = qid_queries[qid]
pid_passage_relevance_tuple = file_processor.get_candidate_passages_relevance_by_qid(
lines, qid)
all_pid_passage = get_pid_passage_pair(pid_passage_relevance_tuple)
relevant_pid_passage = get_relevant_pid_passage_pair(
pid_passage_relevance_tuple)
scored_passage = get_scored_passage_by_query(qid, query,
all_pid_passage,
relevant_pid_passage)
print("Completed-> " + str(counter) + " -> out of: " +
str(total_queries))
file_processor.write_ranked_passage(scored_passage, result_file_path)
import file_util.file_processor as file_processor
import index.index_builder as index_builder
file_path = "../dataset/candidate_passages_top1000.tsv"
# passage collection
lines = file_processor.read_lines(file_path)
pid_passage_pair = file_processor.get_candidate_passages_by_qid(
lines, "1113437")
index = index_builder.build_inverted_index(pid_passage_pair)