self.directory_results[resources] = {}
self.directory_results[resources][
'mean_reciprocal_rank'] = mean_reciprocal_rank(
directory_results_mrr)
self.directory_results[resources][
'mean_average_precision'] = mean_average_precision(
directory_results_mrr)
all_results_mrr += directory_results_mrr
return self.inner_results, self.directory_results, {
"mean_reciprocal_rank": mean_reciprocal_rank(all_results_mrr),
"mean_average_precision": mean_average_precision(all_results_mrr)
}
finalEvaluator = FinalEvaluator("output/results.data")
inner_results, directory_results, outer_results = finalEvaluator.final_evaluate(
)
print_model_into_file(inner_results, "output/basic_model/0_inner_results.json")
print_model_into_file(directory_results,
"output/basic_model/1_directory_results.json")
print_model_into_file(outer_results, "output/basic_model/2_outer_results.json")
finalEvaluator = FinalEvaluator("output/results_bm25.data")
inner_results, directory_results, outer_results = finalEvaluator.final_evaluate(
)
print_model_into_file(inner_results, "output/bm25_model/0_inner_results.json")
print_model_into_file(directory_results,
"output/bm25_model/1_directory_results.json")
print_model_into_file(outer_results, "output/bm25_model/2_outer_results.json")
from utils import print_model_into_pickle_file, print_model_into_file
class EvaluatorBM25:
def __init__(self):
self.modelQuery = ModelQueryBM25(modelfile="model_files/output_bm25.data")
self.results = {}
def evaluate(self, path):
for root, dirs, files in os.walk(path):
self.results[root] = {}
if root != 'news_resources/':
print('files in ' + root + ': ' + ', '.join(files))
for file in files:
if file.endswith('.json'):
fullpath = root + '/' + file
article = json.load(open(fullpath), object_hook=json_util.object_hook) # load article.
results = self.modelQuery.search_for_query(article['title'])
self.results[root][article['article_id']] = {"results": results, "article": article}
return self.results
start = datetime.now()
evaluator = EvaluatorBM25()
results = evaluator.evaluate("news_resources/")
print_model_into_pickle_file(results, "output/results_bm25.data")
end = datetime.now()
print_model_into_file({"time_start": start, "time_end": end, "diff": str(end - start)}, "output/timings/evaluation_bm25.json")
self.json_files[root])
self.total_statistics['total_number_of_documents'] += len(
self.json_files[root])
self.total_statistics[
'total_avarage_query_length'] += self.statistics[root][
'avarage_query_length']
self.total_statistics[
'total_avarage_text_length'] += self.statistics[root][
'avarage_text_length']
self.total_statistics['total_avarage_query_length'] /= len(
self.json_files)
self.total_statistics['total_avarage_text_length'] /= len(
self.json_files)
return self.json_files, self.statistics, self.total_statistics
INPUT_DIRECTORY = "news_resources/"
OUTPUT_FILEPATH = "output/"
datasetAnalyzer = DatasetAnalyzer()
json_files, statistics, total_statistics = datasetAnalyzer.analyze(
INPUT_DIRECTORY)
print_model_into_file(statistics,
OUTPUT_FILEPATH + "/dataset_statistics/statistics.json")
print_model_into_file(
total_statistics,
OUTPUT_FILEPATH + "/dataset_statistics/total_statistics.json")
def append_inverse_document_frequency(self):
docs, words = self.bag_of_words.shape
for i in range(words):
frequency_per_doc = self.bag_of_words[:, i]
nw = 0
for j in range(len(frequency_per_doc)):
if frequency_per_doc[j] != 0:
nw += 1
idf = np.log(docs / nw)
self.bag_of_words[:, i] *= idf
start = datetime.now()
indexer = Indexer()
indexer.fill_terms("news_resources/")
indexer.fill_bags_of_words()
indexer.append_inverse_document_frequency()
print_model_into_pickle_file(
{
'matrix': scipy.sparse.csr_matrix(indexer.bag_of_words),
'articles': indexer.article_list,
'terms': indexer.terms
}, "model_files/output.data")
end = datetime.now()
print_model_into_file(
{
"time_start": start,
"time_end": end,
"diff": str(end - start)
}, "output/timings/index_creation.json")