def test_dbpedia_loading():
adjpath = abspath(expanduser('~/Projects/truthy_data/dbpedia/2016-04/processed/kg/adjacency.npy'))
shape = (6060993, 6060993, 663)
dirpath = join(dirname(adjpath), '_undir')
G = Graph.reconstruct(dirpath, shape, sym=True)
assert np.all(G.csr.indices >= 0)
# reverse graph
dirpath = join(dirname(adjpath), '_revundir')
revG = Graph.reconstruct(dirpath, shape, sym=True)
assert np.all(revG.csr.indices >= 0)
def test_graph1_creation():
shape = np.asarray([4, 4, 2], dtype=np.int32)
adj = np.asarray([
[0, 1, 0],
[0, 2, 1],
[1, 2, 0],
[1, 2, 1],
[1, 3, 1],
[2, 3, 0],
[2, 1, 1],
[2, 3, 1],
],
dtype=np.int32)
values = np.arange(adj.shape[0]) + 10.
# create graph
expect_G = np.asarray([[0., 1., 0., 0., 0., 0., 1., 0.],
[1., 0., 1., 0., 0., 0., 1., 1.],
[0., 1., 0., 1., 1., 1., 0., 1.],
[0., 0., 1., 0., 0., 1., 1., 0.]])
G = make_graph(adj, shape, sym=True, save_csc=True)
assert np.array_equal(G.csr.toarray(), G.csc.toarray())
dirpath = join(abspath(expanduser(os.curdir)), '_undir')
if not exists(dirpath):
os.mkdir(dirpath)
G.save_graph(dirpath)
assert np.array_equal(G.indeg_vec, np.asarray([2, 3, 3, 2]))
assert np.array_equal(expect_G, G.csr.toarray())
# rebuild graph
G = Graph.reconstruct(dirpath, shape, sym=True, save_csc=True)
assert np.array_equal(expect_G, G.csr.toarray())
if exists(dirpath):
shutil.rmtree(dirpath)
print('Removed: %s' % dirpath)
def test_dbpedia():
adjpath = abspath(expanduser('./data/kg/adjacency.npy'))
shape = (6060993, 6060993, 663)
adj = np.load(adjpath)
adj = adj.astype(np.int32)
T = Graph(adj, shape, sym=True)
# save graph
print('Saving graph..')
t1 = time()
dirpath = join(dirname(adjpath), '_undir')
if not exists(dirpath):
os.makedirs(dirpath)
print('* Created: %s' % dirpath)
T.save_graph(dirpath)
print('Graph saved in {:.4f} secs at: {} '.format(time() - t1, dirpath))
def test_dbpedia():
dirpath = abspath(expanduser('./data/kg/_undir/'))
shape = (6060993, 6060993, 663)
G = Graph.reconstruct(dirpath, shape, sym=True)
cost_vec = np.log(G.indeg_vec)
s, p, o = 2145431, 178, 459128 # Gravity, Alfonso Cuarón
mincostflow = succ_shortest_path(G, cost_vec, s, p, o)
print mincostflow
def main(args=None):
# parse arguments
parser = argparse.ArgumentParser(
description=__doc__,
formatter_class=argparse.ArgumentDefaultsHelpFormatter)
parser.add_argument('-m',
type=str,
required=True,
dest='method',
help='Method to use: stream, relklinker, klinker, \
predpath, pra, katz, pathent, simrank, adamic_adar, jaccard, degree_product.'
)
parser.add_argument('-d',
type=str,
required=True,
dest='dataset',
help='Dataset to test on.')
parser.add_argument('-o',
type=str,
required=True,
dest='outdir',
help='Path to the output directory.')
args = parser.parse_args()
# logging
disable_logging(log.DEBUG)
if args.method not in ('stream', 'relklinker', 'klinker', 'predpath',
'pra', 'katz', 'pathent', 'simrank', 'adamic_adar',
'jaccard', 'degree_product'):
raise Exception('Invalid method specified.')
# ensure input file and output directory is valid.
outdir = abspath(expanduser(args.outdir))
assert exists(outdir)
args.outdir = outdir
datafile = abspath(expanduser(args.dataset))
assert exists(datafile)
args.dataset = datafile
log.info('Launching {}..'.format(args.method))
log.info('Dataset: {}'.format(basename(args.dataset)))
log.info('Output dir: {}'.format(args.outdir))
# read data
df = pd.read_table(args.dataset, sep=',', header=0)
log.info('Read data: {} {}'.format(df.shape, basename(args.dataset)))
spo_df = df.dropna(axis=0, subset=['sid', 'pid', 'oid'])
log.info('Note: Found non-NA records: {}'.format(spo_df.shape))
df = spo_df[['sid', 'pid', 'oid']].values
subs, preds, objs = df[:, 0].astype(_int), df[:, 1].astype(
_int), df[:, 2].astype(_int)
# load knowledge graph
G = Graph.reconstruct(PATH, SHAPE, sym=True) # undirected
assert np.all(G.csr.indices >= 0)
# relational similarity
relsim = np.load(RELSIMPATH)
# execute
base = splitext(basename(args.dataset))[0]
t1 = time()
if args.method == 'stream': # KNOWLEDGE STREAM (KS)
# compute min. cost flow
log.info('Computing KS for {} triples..'.format(spo_df.shape[0]))
with warnings.catch_warnings():
warnings.simplefilter("ignore")
outjson = join(args.outdir,
'out_kstream_{}_{}.json'.format(base, DATE))
outcsv = join(args.outdir,
'out_kstream_{}_{}.csv'.format(base, DATE))
mincostflows, times = compute_mincostflow(G, relsim, subs, preds,
objs, outjson)
# save the results
spo_df['score'] = mincostflows
spo_df['time'] = times
spo_df = normalize(spo_df)
spo_df.to_csv(outcsv, sep=',', header=True, index=False)
log.info('* Saved results: %s' % outcsv)
log.info(
'Mincostflow computation complete. Time taken: {:.2f} secs.\n'.
format(time() - t1))
elif args.method == 'relklinker': # RELATIONAL KNOWLEDGE LINKER (KL-REL)
log.info('Computing KL-REL for {} triples..'.format(spo_df.shape[0]))
scores, paths, rpaths, times = compute_relklinker(
G, relsim, subs, preds, objs)
# save the results
spo_df['score'] = scores
spo_df['path'] = paths
spo_df['rpath'] = rpaths
spo_df['time'] = times
spo_df = normalize(spo_df)
outcsv = join(args.outdir,
'out_relklinker_{}_{}.csv'.format(base, DATE))
spo_df.to_csv(outcsv, sep=',', header=True, index=False)
log.info('* Saved results: %s' % outcsv)
log.info(
'Relatioanal KL computation complete. Time taken: {:.2f} secs.\n'.
format(time() - t1))
elif args.method == 'klinker':
log.info('Computing KL for {} triples..'.format(spo_df.shape[0]))
scores, paths, rpaths, times = compute_klinker(G, subs, preds, objs)
# save the results
spo_df['score'] = scores
spo_df['path'] = paths
spo_df['rpath'] = rpaths
spo_df['time'] = times
spo_df = normalize(spo_df)
outcsv = join(args.outdir, 'out_klinker_{}_{}.csv'.format(base, DATE))
spo_df.to_csv(outcsv, sep=',', header=True, index=False)
log.info('* Saved results: %s' % outcsv)
log.info('KL computation complete. Time taken: {:.2f} secs.\n'.format(
time() - t1))
elif args.method == 'predpath': # PREDPATH
vec, model = predpath_train_model(G, spo_df) # train
print 'Time taken: {:.2f}s\n'.format(time() - t1)
# save model
predictor = {'dictvectorizer': vec, 'model': model}
try:
outpkl = join(args.outdir,
'out_predpath_{}_{}.pkl'.format(base, DATE))
with open(outpkl, 'wb') as g:
pkl.dump(predictor, g, protocol=pkl.HIGHEST_PROTOCOL)
print 'Saved: {}'.format(outpkl)
except IOError, e:
raise e
class Pra(object):
name = 'pra'
HOME = abspath(expanduser('./data/'))
if not exists(HOME):
print 'Data directory not found: %s' % HOME
print 'Download data per instructions on:'
print '\thttps://github.com/shiralkarprashant/knowledgestream#data'
print 'and enter the directory path below.'
data_dir = raw_input('\nPlease enter data directory path: ')
if data_dir != '':
data_dir = abspath(expanduser(data_dir))
if not os.path.isdir(data_dir):
raise Exception('Entered path "%s" not a directory.' % data_dir)
if not exists(data_dir):
raise Exception('Directory does not exist: %s' % data_dir)
HOME = data_dir
# raise Exception('Please set HOME to data directory in algorithms/__main__.py')
PATH = join(HOME, 'kg/_undir/')
assert exists(PATH)
SHAPE = (6060993, 6060993, 663)
WTFN = 'logdegree'
# relational similarity using TF-IDF representation and cosine similarity
RELSIMPATH = join(HOME, 'relsim/coo_mat_sym_2016-10-24_log-tf_tfidf.npy')
assert exists(RELSIMPATH)
# Date
DATE = '{}'.format(date.today())
# data types for int and float
_short = np.int16
_int = np.int32
_int64 = np.int64
_float = np.float
# load knowledge graph
G = Graph.reconstruct(PATH, SHAPE, sym=True) # undirected
assert np.all(G.csr.indices >= 0)
# relational similarity
relsim = np.load(RELSIMPATH)
# ================= PRA ALGORITHM IMPLEMENTATION ============
@rpc # Methods are exposed to the outside world with entrypoint decorators (RPC in our case)
def stream(self, data, args=None):
print('\nThe following request in RDF format was passed:')
print(data)
identification, theDate, suri, puri, ouri = extract.getValues(data)
print('\nSURI, PURI and OURI are:')
print(suri)
print(puri)
print(ouri)
print('\n')
# sid, pid, oid = self.uriToId(suri, puri, ouri)
sid, pid, oid = mapping.convert(suri, puri, ouri)
# required for passing it to compute_mincostflow
sid, pid, oid = np.array([sid]), np.array([pid]), np.array([oid])
t1 = time()
print('\nTheir corresponding IDs are:')
print(sid)
print(pid)
print(oid)
print('\n')
log.info('Computing Predpath for triple')
int_sid = int(sid)
int_pid = int(pid)
int_oid = int(oid)
print("The subject id is: %s " % int_sid)
print("The predicate id is: %s" % int_pid)
print("The object id is: %s" % int_oid)
# Creating a dataframe
data = {
'sid': [int_sid],
'pid': [int_pid],
'oid': [int_oid],
'class': [0]
}
#__________________test________________________
dfObj = pd.DataFrame(data)
test_spo_df = dfObj.dropna(axis=0,
subset=['sid', 'pid', 'oid', 'class'])
test_model_pkl = open("./output/trained_pra_model.pkl", "rb")
test_model = pkl.load(test_model_pkl)
test_features_pkl = open("./output/pra_features_file.pkl", "rb")
test_features = pkl.load(test_features_pkl)
with warnings.catch_warnings():
try:
warnings.simplefilter("ignore")
# pra_predict() function is used to predict the triple's veracity
array_value = pra_predict(self.G, test_features, test_model,
test_spo_df) # test
val = str(array_value)[1:-1]
log.info(
'Predpath computation complete. Time taken: {:.2f} secs.\n'
.format(time() - t1))
result = '<http://swc2017.aksw.org/task2/dataset/s-' + str(
identification
) + '> <http://swc2017.aksw.org/hasTruthValue>\"' + str(
val) + '\"<http://www.w3.org/2001/XMLSchema#double> .'
print('The result in RDF format is:')
print(result)
except MemoryError:
print('\nA MemoryError is successfully caught.')
result = 'MemoryError'
return result
def main(args=None):
parser = argparse.ArgumentParser(
description=__doc__,
formatter_class=argparse.ArgumentDefaultsHelpFormatter)
parser.add_argument('-d',
type=str,
required=True,
dest='dataset',
help='Dataset to test on.')
parser.add_argument('-o',
type=str,
required=True,
dest='outdir',
help='Path to the output directory.')
parser.add_argument('-m',
type=str,
required=True,
dest='method',
help='Method to use: stream, relklinker, klinker, \
predpath, sm')
args = parser.parse_args()
relsim = np.load(RELSIMPATH)
outdir = abspath(expanduser(args.outdir))
assert exists(outdir)
args.outdir = outdir
datafile = abspath(expanduser(args.dataset))
assert exists(datafile)
args.dataset = datafile
LOGPATH = join(HOME, '../logs')
assert exists(LOGPATH)
base = splitext(basename(args.dataset))[0]
log_file = join('logs/', 'log_{}_{}_{}.log'.format(args.method, base,
DATE))
log.basicConfig(format='[%(asctime)s] %(message)s',
datefmt='%m/%d/%Y %H:%M:%S %p',
filename=log_file,
level=log.DEBUG)
log.getLogger().addHandler(log.StreamHandler())
log.info('Launching {}..'.format(args.method))
log.info('Dataset: {}'.format(basename(args.dataset)))
log.info('Output dir: {}'.format(args.outdir))
# read data
df = pd.read_table(args.dataset, sep=',', header=0)
log.info('Read data: {} {}'.format(df.shape, basename(args.dataset)))
spo_df = df.dropna(axis=0, subset=['sid', 'pid', 'oid'])
log.info('Note: Found non-NA records: {}'.format(spo_df.shape))
df = spo_df[['sid', 'pid', 'oid']].values
subs, preds, objs = df[:, 0].astype(_int), df[:, 1].astype(
_int), df[:, 2].astype(_int)
# load knowledge graph
G = Graph.reconstruct(PATH, SHAPE, sym=True) # undirected
assert np.all(G.csr.indices >= 0)
t1 = time()
if args.method == 'stream': # KNOWLEDGE STREAM (KS)
# compute min. cost flow
log.info('Computing KS for {} triples..'.format(spo_df.shape[0]))
with warnings.catch_warnings():
warnings.simplefilter("ignore")
outjson = join(args.outdir,
'out_kstream_{}_{}.json'.format(base, DATE))
outcsv = join(args.outdir,
'out_kstream_{}_{}.csv'.format(base, DATE))
mincostflows, times = compute_mincostflow(G, relsim, subs, preds,
objs, outjson)
# save the results
spo_df['score'] = mincostflows
spo_df['time'] = times
spo_df = normalize(spo_df)
spo_df.to_csv(outcsv, sep=',', header=True, index=False)
log.info('* Saved results: %s' % outcsv)
log.info(
'Mincostflow computation complete. Time taken: {:.2f} secs.\n'.
format(time() - t1))
elif args.method == 'relklinker': # RELATIONAL KNOWLEDGE LINKER (KL-REL)
log.info('Computing KL-REL for {} triples..'.format(spo_df.shape[0]))
scores, paths, rpaths, times = compute_relklinker(
G, relsim, subs, preds, objs)
# save the results
spo_df['score'] = scores
spo_df['path'] = paths
spo_df['rpath'] = rpaths
spo_df['time'] = times
spo_df = normalize(spo_df)
outcsv = join(args.outdir,
'out_relklinker_{}_{}.csv'.format(base, DATE))
spo_df.to_csv(outcsv, sep=',', header=True, index=False)
log.info('* Saved results: %s' % outcsv)
log.info(
'Relatioanal KL computation complete. Time taken: {:.2f} secs.\n'.
format(time() - t1))
elif args.method == 'klinker':
log.info('Computing KL for {} triples..'.format(spo_df.shape[0]))
scores, paths, rpaths, times = compute_klinker(G, subs, preds, objs)
# save the results
spo_df['score'] = scores
spo_df['path'] = paths
spo_df['rpath'] = rpaths
spo_df['time'] = times
spo_df = normalize(spo_df)
outcsv = join(args.outdir, 'out_klinker_{}_{}.csv'.format(base, DATE))
spo_df.to_csv(outcsv, sep=',', header=True, index=False)
log.info('* Saved results: %s' % outcsv)
log.info('KL computation complete. Time taken: {:.2f} secs.\n'.format(
time() - t1))
elif args.method == 'predpath': # PREDPATH
vec, model = predpath_train_model(G, spo_df) # train
# vec, model = predpath_train_model(G, spo_df, relsim)
print 'Time taken: {:.2f}s\n'.format(time() - t1)
# save model
predictor = {'dictvectorizer': vec, 'model': model}
try:
outpkl = join(args.outdir,
'out_predpath_{}_{}.pkl'.format(base, DATE))
with open(outpkl, 'wb') as g:
pkl.dump(predictor, g, protocol=pkl.HIGHEST_PROTOCOL)
print 'Saved: {}'.format(outpkl)
except IOError, e:
raise e
class KnowledgeLinker(object):
name = 'klinker'
HOME = abspath(expanduser('./data/'))
if not exists(HOME):
print 'Data directory not found: %s' % HOME
print 'Download data per instructions on:'
print '\thttps://github.com/shiralkarprashant/knowledgestream#data'
print 'and enter the directory path below.'
data_dir = raw_input('\nPlease enter data directory path: ')
if data_dir != '':
data_dir = abspath(expanduser(data_dir))
if not os.path.isdir(data_dir):
raise Exception('Entered path "%s" not a directory.' % data_dir)
if not exists(data_dir):
raise Exception('Directory does not exist: %s' % data_dir)
HOME = data_dir
# raise Exception('Please set HOME to data directory in algorithms/__main__.py')
PATH = join(HOME, 'kg/_undir/')
assert exists(PATH)
SHAPE = (6060993, 6060993, 663)
WTFN = 'logdegree'
# relational similarity using TF-IDF representation and cosine similarity
RELSIMPATH = join(HOME, 'relsim/coo_mat_sym_2016-10-24_log-tf_tfidf.npy')
assert exists(RELSIMPATH)
# Date
DATE = '{}'.format(date.today())
# data types for int and float
_short = np.int16
_int = np.int32
_int64 = np.int64
_float = np.float
# load knowledge graph
G = Graph.reconstruct(PATH, SHAPE, sym=True) # undirected
assert np.all(G.csr.indices >= 0)
# relational similarity
relsim = np.load(RELSIMPATH)
# ================= KNOWLEDGE LINKER ALGORITHM ============
def compute_klinker(self, G, sid, pid, oid):
"""
Parameters:
-----------
G: rgraph
See `datastructures`.
subs, preds, objs: sequence
Sequences representing the subject, predicate and object of
input triples.
Returns:
--------
scores, paths, rpaths, times: sequence
One sequence each for the proximity scores, shortest path in terms of
nodes, shortest path in terms of relation sequence, and times taken.
"""
# set weights
indegsim = weighted_degree(G.indeg_vec, weight=self.WTFN).reshape(
(1, G.N))
indegsim = indegsim.ravel()
targets = G.csr.indices % G.N
specificity_wt = indegsim[targets] # specificity
G.csr.data = specificity_wt.copy()
# back up
data = G.csr.data.copy()
indices = G.csr.indices.copy()
indptr = G.csr.indptr.copy()
# compute closure
scores, paths, rpaths, times = [], [], [], []
for idx, (s, p, o) in enumerate(zip(sid, pid, oid)):
print '{}. Working on {}..'.format(idx + 1, (s, p, o)),
ts = time()
rp = closure(G, s, p, o, kind='metric', linkpred=True)
tend = time()
print 'time: {:.2f}s'.format(tend - ts)
times.append(tend - ts)
scores.append(rp.score)
paths.append(rp.path)
rpaths.append(rp.relational_path)
# reset graph
G.csr.data = data.copy()
G.csr.indices = indices.copy()
G.csr.indptr = indptr.copy()
sys.stdout.flush()
log.info('')
return scores, paths, rpaths, times
@rpc # Methods are exposed to the outside world with entrypoint decorators (RPC in our case)
def stream(self, data):
print('\nThe following request in RDF format was passed:')
print(data)
identification, theDate, suri, puri, ouri = extract.getValues(data)
print('\nSURI, PURI and OURI are:')
print(suri)
print(puri)
print(ouri)
print('\n')
# sid, pid, oid = self.uriToId(suri, puri, ouri)
sid, pid, oid = mapping.convert(suri, puri, ouri)
# required for passing it to compute_mincostflow
sid, pid, oid = np.array([sid]), np.array([pid]), np.array([oid])
t1 = time()
print('\nTheir corresponding IDs are:')
print(sid)
print(pid)
print(oid)
print('\n')
log.info('Computing KL for triple')
with warnings.catch_warnings():
try:
warnings.simplefilter("ignore")
# compute klinker
scores, paths, rpaths, times = self.compute_klinker(
self.G, sid, pid, oid)
normalizedScore = normalization.score(scores[0])
log.info(
'KLinker computation complete. Time taken: {:.2f} secs.\n'.
format(time() - t1))
result = '<http://swc2017.aksw.org/task2/dataset/s-' + str(
identification
) + '> <http://swc2017.aksw.org/hasTruthValue>\"' + str(
normalizedScore
) + '\"<http://www.w3.org/2001/XMLSchema#double> .'
print('The result in RDF format is:')
print(result)
except MemoryError:
print('\nA MemoryError is successfully caught.')
result = 'MemoryError'
return result
class KnowledgeStream(object):
name = 'kstream'
HOME = abspath(expanduser('./data/'))
if not exists(HOME):
print 'Data directory not found: %s' % HOME
print 'Download data per instructions on:'
print '\thttps://github.com/shiralkarprashant/knowledgestream#data'
print 'and enter the directory path below.'
data_dir = raw_input('\nPlease enter data directory path: ')
if data_dir != '':
data_dir = abspath(expanduser(data_dir))
if not os.path.isdir(data_dir):
raise Exception('Entered path "%s" not a directory.' % data_dir)
if not exists(data_dir):
raise Exception('Directory does not exist: %s' % data_dir)
HOME = data_dir
# raise Exception('Please set HOME to data directory in algorithms/__main__.py')
PATH = join(HOME, 'kg/_undir/')
assert exists(PATH)
SHAPE = (6060993, 6060993, 663)
WTFN = 'logdegree'
# relational similarity using TF-IDF representation and cosine similarity
RELSIMPATH = join(HOME, 'relsim/coo_mat_sym_2016-10-24_log-tf_tfidf.npy')
assert exists(RELSIMPATH)
# Date
DATE = '{}'.format(date.today())
# data types for int and float
_short = np.int16
_int = np.int32
_int64 = np.int64
_float = np.float
# load knowledge graph
G = Graph.reconstruct(PATH, SHAPE, sym=True) # undirected
assert np.all(G.csr.indices >= 0)
# relational similarity
relsim = np.load(RELSIMPATH)
# ================= KNOWLEDGE STREAM ALGORITHM ============
def compute_mincostflow(self, G, relsim, sid, pid, oid):
"""
Parameters:
-----------
G: rgraph
See `datastructures`.
relsim: ndarray
A square matrix containing relational similarity scores.
subs, preds, objs: sequence
Sequences representing the subject, predicate and object of
input triples.
flowfile: str
Absolute path of the file where flow will be stored as JSON,
one line per triple.
Returns:
--------
mincostflows: sequence
A sequence containing total flow for each triple.
times: sequence
Times taken to compute stream of each triple.
"""
# take graph backup
G_bak = {
'data': G.csr.data.copy(),
'indices': G.csr.indices.copy(),
'indptr': G.csr.indptr.copy()
}
cost_vec_bak = np.log(G.indeg_vec).copy()
# some set up
G.sources = np.repeat(np.arange(G.N), np.diff(G.csr.indptr))
G.targets = G.csr.indices % G.N
cost_vec = cost_vec_bak.copy()
indegsim = weighted_degree(G.indeg_vec, weight=self.WTFN)
specificity_wt = indegsim[G.targets] # specificity
relations = (G.csr.indices - G.targets) / G.N
s, p, o = [int(x) for x in (sid, pid, oid)]
ts = time()
print '{}. Working on {} .. '.format(1, (s, p, o)),
sys.stdout.flush()
# set weights
relsimvec = np.array(relsim[p, :]) # specific to predicate p
relsim_wt = relsimvec[relations]
G.csr.data = np.multiply(relsim_wt, specificity_wt)
# compute
mcflow = succ_shortest_path(
G, cost_vec, s, p, o, return_flow=False, npaths=5
)
mincostflow = mcflow.flow
tend = time()
times = tend - ts
print 'mincostflow: {:.5f}, #paths: {}, time: {:.2f}s.'.format(
mcflow.flow, len(mcflow.stream['paths']), tend - ts
)
# reset state of the graph
np.copyto(G.csr.data, G_bak['data'])
np.copyto(G.csr.indices, G_bak['indices'])
np.copyto(G.csr.indptr, G_bak['indptr'])
np.copyto(cost_vec, cost_vec_bak)
return mincostflow, times
@rpc # Methods are exposed to the outside world with entrypoint decorators (RPC in our case)
def stream(self, data):
print('\nThe following request in RDF format was passed:')
print(data)
identification, theDate, suri, puri, ouri = extract.getValues(data)
print('\nSURI, PURI and OURI are:')
print(suri)
print(puri)
print(ouri)
print('\n')
# sid, pid, oid = self.uriToId(suri, puri, ouri)
sid, pid, oid = mapping.convert(suri, puri, ouri)
# required for passing it to compute_mincostflow
sid, pid, oid = np.array([sid]), np.array([pid]), np.array([oid])
t1 = time()
print('\nTheir corresponding IDs are:')
print(sid)
print(pid)
print(oid)
print('\n')
log.info('Computing KS for triple')
with warnings.catch_warnings():
try:
warnings.simplefilter("ignore")
# compute min. cost flow
mincostflows, times = self.compute_mincostflow(self.G, self.relsim, sid, pid, oid)
# spo_df = self.normalize(spo_df)
normalizedScore = normalization.score(mincostflows)
log.info('Mincostflow computation complete. Time taken: {:.2f} secs.\n'.format(time() - t1))
result = '<http://swc2017.aksw.org/task2/dataset/s-' + str(
identification) + '> <http://swc2017.aksw.org/hasTruthValue>\"' + str(
normalizedScore) + '\"<http://www.w3.org/2001/XMLSchema#double> .'
print('The result in RDF format is:')
print(result)
except MemoryError:
print('\nA MemoryError is successfully caught.')
result = 'MemoryError'
return result
class DegreeProduct(object):
name = 'degree_product'
HOME = abspath(expanduser('./data/'))
if not exists(HOME):
print 'Data directory not found: %s' % HOME
print 'Download data per instructions on:'
print '\thttps://github.com/shiralkarprashant/knowledgestream#data'
print 'and enter the directory path below.'
data_dir = raw_input('\nPlease enter data directory path: ')
if data_dir != '':
data_dir = abspath(expanduser(data_dir))
if not os.path.isdir(data_dir):
raise Exception('Entered path "%s" not a directory.' % data_dir)
if not exists(data_dir):
raise Exception('Directory does not exist: %s' % data_dir)
HOME = data_dir
# raise Exception('Please set HOME to data directory in algorithms/__main__.py')
PATH = join(HOME, 'kg/_undir/')
assert exists(PATH)
SHAPE = (6060993, 6060993, 663)
WTFN = 'logdegree'
# relational similarity using TF-IDF representation and cosine similarity
RELSIMPATH = join(HOME, 'relsim/coo_mat_sym_2016-10-24_log-tf_tfidf.npy')
assert exists(RELSIMPATH)
# Date
DATE = '{}'.format(date.today())
# data types for int and float
_short = np.int16
_int = np.int32
_int64 = np.int64
_float = np.float
# load knowledge graph
G = Graph.reconstruct(PATH, SHAPE, sym=True) # undirected
assert np.all(G.csr.indices >= 0)
# relational similarity
relsim = np.load(RELSIMPATH)
# ================= DEGREE PRODUCT ALGORITHM ============
def compute_degree_product(self, G, subs, preds, objs):
"""
Performs link prediction using a specified measure, such as Katz or SimRank.
Parameters:
-----------
G: rgraph
See `datastructures`.
subs, preds, objs: sequence
Sequences representing the subject, predicate and object of
input triples.
Returns:
--------
scores, times: sequence
One sequence each for the proximity scores and times taken.
"""
measure_map = {
'degree_product': {
'measure': preferential_attachment,
'tag': 'PA'
}
}
selected_measure = 'degree_product'
# back up
data = G.csr.data.copy()
indices = G.csr.indices.copy()
indptr = G.csr.indptr.copy()
# compute closure
measure_name = measure_map[selected_measure]['tag']
measure = measure_map[selected_measure]['measure']
log.info('Computing {} for {} triples..'.format(
measure_name, len(subs)))
t1 = time()
scores, times = [], []
for idx, (s, p, o) in enumerate(zip(subs, preds, objs)):
print '{}. Working on {}..'.format(idx + 1, (s, p, o)),
sys.stdout.flush()
ts = time()
score = measure(G, s, p, o, linkpred=True)
tend = time()
print 'score: {:.5f}, time: {:.2f}s'.format(score, tend - ts)
times.append(tend - ts)
scores.append(score)
# reset graph
G.csr.data = data.copy()
G.csr.indices = indices.copy()
G.csr.indptr = indptr.copy()
sys.stdout.flush()
print ''
return scores, times
@rpc # Methods are exposed to the outside world with entrypoint decorators (RPC in our case)
def stream(self, data):
print('\nThe following request in RDF format was passed:')
print(data)
identification, theDate, suri, puri, ouri = extract.getValues(data)
print('\nSURI, PURI and OURI are:')
print(suri)
print(puri)
print(ouri)
print('\n')
# sid, pid, oid = self.uriToId(suri, puri, ouri)
sid, pid, oid = mapping.convert(suri, puri, ouri)
# required for passing it to compute_degree_product
sid, pid, oid = np.array([sid]), np.array([pid]), np.array([oid])
t1 = time()
print('\nTheir corresponding IDs are:')
print(sid)
print(pid)
print(oid)
print('\n')
log.info('Computing PA for triple')
with warnings.catch_warnings():
try:
warnings.simplefilter("ignore")
# compute degree_product
scores, times = self.compute_degree_product(
self.G, sid, pid, oid)
normalizedScore = normalization.score(scores[0])
log.info(
'DegreeProduct computation complete. Time taken: {:.2f} secs.\n'
.format(time() - t1))
result = '<http://swc2017.aksw.org/task2/dataset/s-' + str(
identification
) + '> <http://swc2017.aksw.org/hasTruthValue>\"' + str(
normalizedScore
) + '\"<http://www.w3.org/2001/XMLSchema#double> .'
print('The result in RDF format is:')
print(result)
except MemoryError:
print('\nA MemoryError is successfully caught.')
result = 'MemoryError'
return result
class Predpath(object):
name = 'pra'
HOME = abspath(expanduser('./data/'))
if not exists(HOME):
print 'Data directory not found: %s' % HOME
print 'Download data per instructions on:'
print '\thttps://github.com/shiralkarprashant/knowledgestream#data'
print 'and enter the directory path below.'
data_dir = raw_input('\nPlease enter data directory path: ')
if data_dir != '':
data_dir = abspath(expanduser(data_dir))
if not os.path.isdir(data_dir):
raise Exception('Entered path "%s" not a directory.' % data_dir)
if not exists(data_dir):
raise Exception('Directory does not exist: %s' % data_dir)
HOME = data_dir
# raise Exception('Please set HOME to data directory in algorithms/__main__.py')
PATH = join(HOME, 'kg/_undir/')
assert exists(PATH)
SHAPE = (6060993, 6060993, 663)
WTFN = 'logdegree'
# relational similarity using TF-IDF representation and cosine similarity
RELSIMPATH = join(HOME, 'relsim/coo_mat_sym_2016-10-24_log-tf_tfidf.npy')
assert exists(RELSIMPATH)
# Date
DATE = '{}'.format(date.today())
# data types for int and float
_short = np.int16
_int = np.int32
_int64 = np.int64
_float = np.float
# load knowledge graph
G = Graph.reconstruct(PATH, SHAPE, sym=True) # undirected
assert np.all(G.csr.indices >= 0)
# relational similarity
relsim = np.load(RELSIMPATH)
#__________________train_______________________
# ensure input file and output directory is valid.
outdir = abspath(expanduser('./output'))
assert exists(outdir)
#sample data file consisting of records that is used to train the model.
datafile = abspath(expanduser('./datasets/sample_data_pra.csv'))
assert exists(datafile)
log.info('Dataset: {}'.format(basename(datafile)))
# Date
DATE = '{}'.format(date.today())
# read data
df = pd.read_table(datafile, sep=',', header=0)
log.info('Read data: {} {}'.format(df.shape, basename(datafile)))
spo_df = df.dropna(axis=0, subset=['sid', 'pid', 'oid'])
log.info('Note: Found non-NA records: {}'.format(spo_df.shape))
# execute
base = splitext(basename(datafile))[0]
t1 = time()
log.info('Computing pra for {} triples..'.format(spo_df.shape[0]))
#function that trains the model
features, model = pra_train_model(G, spo_df) # train
print 'Time taken: {:.2f}s\n'.format(time() - t1)
# save model
predictor = {'dictvectorizer': features, 'model': model}
try:
outpkl = join(outdir, 'trained_pra_model.pkl')
with open(outpkl, 'wb') as g:
s = pkl.dump(model, g, protocol=pkl.HIGHEST_PROTOCOL)
outpkl_features = join(outdir, 'pra_features_file.pkl')
with open(outpkl_features, 'wb') as g:
s = pkl.dump(features, g, protocol=pkl.HIGHEST_PROTOCOL)
except IOError, e:
raise e
