search_engine.py
No OneTemporary
Actions

Subscribers

None

File Metadata

Created: Tue, Dec 3, 13:04

search_engine.py
View Options

	# -- coding: utf-8 --
	##
	## This file is part of Invenio.
	## Copyright (C) 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011, 2012, 2014 CERN.
	##
	## Invenio is free software; you can redistribute it and/or
	## modify it under the terms of the GNU General Public License as
	## published by the Free Software Foundation; either version 2 of the
	## License, or (at your option) any later version.
	##
	## Invenio is distributed in the hope that it will be useful, but
	## WITHOUT ANY WARRANTY; without even the implied warranty of
	## MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	## General Public License for more details.
	##
	## You should have received a copy of the GNU General Public License
	## along with Invenio; if not, write to the Free Software Foundation, Inc.,
	## 59 Temple Place, Suite 330, Boston, MA 02111-1307, USA.

	""" Author search engine. """

	from .config import QGRAM_LEN, MATCHING_QGRAMS_PERCENTAGE, \
	MAX_T_OCCURANCE_RESULT_LIST_CARDINALITY, MIN_T_OCCURANCE_RESULT_LIST_CARDINALITY, \
	NAME_SCORE_COEFFICIENT

	from Queue import Queue
	from threading import Thread
	from operator import itemgetter
	from msgpack import packb as serialize
	from msgpack import unpackb as deserialize

	from invenio.utils.text import translate_to_ascii
	from intbitset import intbitset
	from .name_utils import create_indexable_name, distance, split_name_parts
	from .dbinterface import get_confirmed_name_to_authors_mapping, get_authors_data_from_indexable_name_ids, get_inverted_lists, \
	set_inverted_lists_ready, set_dense_index_ready, populate_table, search_engine_is_operating


	def get_qgrams_from_string(string, q):
	'''
	It decomposes the given string to its qgrams. The qgrams of a string are its substrings of length q.
	For example the 2-grams (q=2) of string cathey are (ca,at,th,he,ey).

	@param string: the string to be decomposed
	@type string: str
	@param q: the length of the grams
	@type q: int

	@return: the string qgrams ordered accordingly to the position they withhold in the string
	@rtype: list
	'''
	qgrams = list()

	for i in range(len(string)+1-q):
	qgrams.append(string[i:i+q])

	return qgrams


	def create_dense_index(name_pids_dict, names_list, q):
	'''
	It builds the dense index which maps a name to the set of personids whi withhold that name.
	Each entry in the dense index is identified by a unique id called name id.

	@param name_pids_dict:
	@type name_pids_dict: dict
	@param names_list: the names to be indexed
	@type names_list: list
	'''
	def _create_dense_index(name_pids_dict, names_list):
	name_id = 0
	args = list()

	for name in names_list:
	person_name, personids = name_pids_dict[name]
	args += [name_id, person_name, serialize(list(personids))]
	name_id += 1

	populate_table('aidDENSEINDEX', ['name_id','person_name','personids'], args)
	set_dense_index_ready()


	result = (True, None)

	try:
	_create_dense_index(name_pids_dict, names_list)
	except Exception as e:
	result = (False, e)

	q.put(result)


	def create_inverted_lists(names_list, q):
	'''
	It builds the inverted index which maps a qgram to the set of name ids that share that qgram.
	To construct the index it decomposes each name string into its qgrams and adds its id to the
	corresponding inverted list.

	@param names_list: the names to be indexed
	@type names_list: list
	'''
	def create_inverted_lists_worker(names_list):
	name_id = 0
	inverted_lists = dict()

	for name in names_list:
	qgrams = set(get_qgrams_from_string(name, QGRAM_LEN))
	for qgram in qgrams:
	try:
	inverted_list, cardinality = inverted_lists[qgram]
	inverted_list.add(name_id)
	inverted_lists[qgram][1] = cardinality + 1
	except KeyError:
	inverted_lists[qgram] = [set([name_id]), 1]
	name_id += 1

	args = list()

	for qgram in inverted_lists.keys():
	inverted_list, cardinality = inverted_lists[qgram]
	args += [qgram, serialize(list(inverted_list)), cardinality]

	populate_table('aidINVERTEDLISTS', ['qgram','inverted_list','list_cardinality'], args)
	set_inverted_lists_ready()


	result = (True, None)

	try:
	create_inverted_lists_worker(names_list)
	except Exception as e:
	result = (False, e)

	q.put(result)


	def create_bibauthorid_indexer():
	'''
	It constructs the disk-based indexer. It consists of the dense index (which maps a name
	to the set of personids who withhold that name) and the inverted lists (which map a qgram
	to the set of name ids that share that qgram).
	'''
	name_pids_dict = get_confirmed_name_to_authors_mapping()
	if not name_pids_dict:
	return

	indexable_name_pids_dict = dict()

	for name in name_pids_dict.keys():
	asciified_name = translate_to_ascii(name)[0]
	indexable_name = create_indexable_name(asciified_name)
	if indexable_name:
	try:
	asciified_name, pids = indexable_name_pids_dict[indexable_name]
	updated_pids = pids \| name_pids_dict[name]
	indexable_name_pids_dict[indexable_name] = (asciified_name, updated_pids)
	except KeyError:
	indexable_name_pids_dict[indexable_name] = (asciified_name, name_pids_dict[name])

	surname = split_name_parts(name)[0]
	asciified_surname = translate_to_ascii(surname)[0]
	indexable_surname = create_indexable_name(asciified_surname)
	if indexable_surname:
	try:
	asciified_surname, pids = indexable_name_pids_dict[indexable_surname]
	updated_pids = pids \| name_pids_dict[name]
	indexable_name_pids_dict[indexable_surname] = (asciified_surname, updated_pids)
	except KeyError:
	indexable_name_pids_dict[indexable_surname] = (asciified_surname, name_pids_dict[name])

	indexable_names_list = indexable_name_pids_dict.keys()

	# If an exception/error occurs in any of the threads it is not detectable
	# so inter-thread communication is necessary to make it visible.
	q = Queue()
	threads = list()
	threads.append(Thread(target=create_dense_index, args=(indexable_name_pids_dict, indexable_names_list, q)))
	threads.append(Thread(target=create_inverted_lists, args=(indexable_names_list, q)))

	for t in threads:
	t.start()

	for t in threads:
	all_ok, error = q.get(block=True)
	if not all_ok:
	raise error
	q.task_done()

	for t in threads:
	t.join()


	def solve_T_occurence_problem(query_string):
	'''
	It solves a 'T-occurence problem' which is defined as follows: find the string ids
	that apper at least T times on the inverted lists of the query string qgrams. If the
	result dataset is bigger than a threshold it tries to limit it further.

	@param query_string:
	@type query_string: str

	@return: T_occurence_problem answers
	@rtype: list
	'''
	query_string_qgrams = get_qgrams_from_string(query_string, QGRAM_LEN)
	query_string_qgrams_set = set(query_string_qgrams)
	if not query_string_qgrams_set:
	return None

	inverted_lists = get_inverted_lists(query_string_qgrams_set)
	if not inverted_lists:
	return None

	inverted_lists = sorted(inverted_lists, key=itemgetter(1), reverse=True)
	T = int(MATCHING_QGRAMS_PERCENTAGE * len(inverted_lists))
	nameids = intbitset(deserialize(inverted_lists[0][0]))

	for i in range(1, T):
	inverted_list = intbitset(deserialize(inverted_lists[i][0]))
	nameids &= inverted_list

	for i in range(T, len(inverted_lists)):
	if len(nameids) < MAX_T_OCCURANCE_RESULT_LIST_CARDINALITY:
	break
	inverted_list = intbitset(deserialize(inverted_lists[i][0]))
	nameids_temp = inverted_list & nameids
	if len(nameids_temp) > MIN_T_OCCURANCE_RESULT_LIST_CARDINALITY:
	nameids = nameids_temp
	else:
	break

	return nameids


	def calculate_name_score1(query_string, nameids):
	'''
	docstring

	@param query_string:
	@type query_string:
	@param nameids:
	@type nameids:

	@return:
	@rtype:
	'''
	name_personids_list = get_authors_data_from_indexable_name_ids(nameids)
	query_last_name = split_name_parts(query_string)[0]
	query_last_name_len = len(query_last_name)
	name_score_list = list()

	for name, personids in name_personids_list:
	current_last_name = split_name_parts(name)[0]
	current_last_name_len = len(current_last_name)
	if abs(query_last_name_len - current_last_name_len) == 0:
	dist = distance(query_last_name, current_last_name)
	limit = min([query_last_name_len, current_last_name_len])
	name_score = sum([1/float(2**(i+1)) for i in range(limit) if query_last_name[i] == current_last_name[i]])/(dist + 1)
	if name_score > 0.5:
	name_score_list.append((name, name_score, deserialize(personids)))

	return name_score_list

	def calculate_name_score(query_string, nameids):
	'''
	docstring

	@param query_string:
	@type query_string:
	@param nameids:
	@type nameids:

	@return:
	@rtype:
	'''
	name_personids_list = get_authors_data_from_indexable_name_ids(nameids)
	query_last_name = split_name_parts(query_string)[0]
	query_last_name_len = len(query_last_name)
	name_score_list = list()

	for name, personids in name_personids_list:
	current_last_name = split_name_parts(name)[0]
	current_last_name_len = len(current_last_name)
	if abs(query_last_name_len - current_last_name_len) == 0:
	dist = distance(query_last_name, current_last_name)
	limit = min([query_last_name_len, current_last_name_len])
	name_score = sum([1/float(2**(i+1)) for i in range(limit) if query_last_name[i] == current_last_name[i]])/(dist + 1)
	if name_score > 0.5:
	name_score_list.append((name, name_score, deserialize(personids)))

	return name_score_list


	def calculate_pid_score(names_score_list):
	'''
	docstring

	@param names_score_list:
	@type names_score_list:

	@return:
	@rtype:
	'''
	max_appearances = 1
	pid_metrics_dict = dict()

	for name, name_score, personids in names_score_list:
	for pid in personids:
	try:
	appearances = pid_metrics_dict[pid][2]+1
	pid_metrics_dict[pid][2] = appearances
	if appearances > max_appearances:
	max_appearances = appearances
	except KeyError:
	pid_metrics_dict[pid] = [name, name_score, 1]

	pids_score_list = list()

	for pid in pid_metrics_dict.keys():
	name, name_score, appearances = pid_metrics_dict[pid]
	final_score = NAME_SCORE_COEFFICIENTname_score + (1-NAME_SCORE_COEFFICIENT)(appearances/float(max_appearances))
	pids_score_list.append((pid, name, final_score))

	return pids_score_list


	def find_personids_by_name1(query_string):
	'''
	It finds a collection of personids who own a signature that is similar to the given query string.
	Its approach is by solving a 'T-occurance problem' and then it applies some filters to the candidate
	answers so it can remove the false positives. In the end it sorts the result set based on the score
	they obtained.

	@param query_string:
	@type query_string: str

	@return: personids which own a signature similar to the query string
	@rtype: list
	'''
	search_engine_is_functioning = search_engine_is_operating()
	if not search_engine_is_functioning:
	return list()

	asciified_query_string = translate_to_ascii(query_string)[0]
	indexable_query_string = create_indexable_name(asciified_query_string)
	if not indexable_query_string:
	return list()

	#query_string_surname = split_name_parts(query_string)[0]
	#asciified_query_string_surname = translate_to_ascii(query_string_surname)[0]
	#indexable_query_string_surname = create_indexable_name(asciified_query_string_surname)

	#if not indexable_query_string and not indexable_query_string_surname:
	# return list()

	s1 = solve_T_occurence_problem(indexable_query_string)

	if not s1:
	s1 = intbitset()

	nameids = solve_T_occurence_problem(indexable_query_string)

	#s2 = solve_T_occurence_problem(indexable_query_string_surname)
	#if not s2:
	# s2 = intbitset()

	#nameids = s1 \| s2
	if not nameids:
	return list()

	name_score_list = calculate_name_score(asciified_query_string, nameids)

	return name_score_list
	#name_ranking_list = sorted(name_score_list, key=itemgetter(1), reverse=True)

	#pid_score_list = calculate_pid_score(name_ranking_list)
	#pids_ranking_list = sorted(pid_score_list, key=itemgetter(2), reverse=True)

	#ranked_pid_name_list = [pid for pid, name, final_score in pids_ranking_list]

	#return ranked_pid_name_list


	def find_personids_by_name(query_string):
	query_string_surname = split_name_parts(query_string)[0]

	name_score_list = set(find_personids_by_name1(query_string) + find_personids_by_name1(query_string_surname))
	name_ranking_list = sorted(name_score_list, key=itemgetter(1), reverse=True)

	pid_score_list = calculate_pid_score(name_ranking_list)
	pids_ranking_list = sorted(pid_score_list, key=itemgetter(2), reverse=True)

	ranked_pid_name_list = [pid for pid, name, final_score in pids_ranking_list]

	return ranked_pid_name_list

search_engine.pyNo OneTemporaryActions

File Metadata

search_engine.pyView Options

Event Timeline

search_engine.py
No OneTemporary
Actions

search_engine.py
View Options