def ucto(string):
import terminal
"""Tokenizes a string with Ucto"""
open('uctofile','w').write(string)
result = terminal.command('ucto -L nl uctofile',True)
os.remove('uctofile')
return result.replace('<utt>','')
def startserver(self, model, port=None):
""" Start a new Timbl server"""
if not port:
port = self._getnewport()
terminal.command(self.cmdstring % (model, port))
def train_model(self, filename):
"""Train a model on the basis of these ngrams"""
terminal.command("timbl -f " + filename + " -I " + filename + ".IGTree +D +vdb -a1 -p 1000000", False)
return filename.replace(".training.txt", "")
def start_servers(self, model, look_for_existing):
"""Starts the necessary servers and connects to them"""
if self.approach == "w":
foldername = "wordmodels"
backupname = "nlsave"
elif self.approach == "l":
foldername = "lettermodels"
backupname = "nlsmall"
# Personal model
succeeded = False
while not succeeded:
if look_for_existing:
port = get_port_for_timblserver(model + ".training.txt.IGTree")
else:
port = False
if not port:
port = get_free_port()
terminal.command(
"timblserver -i " + model + ".training.txt.IGTree +vdb +D -a1 -G +vcf -S " + str(port) + " -C 1000"
)
s1 = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
try:
s1.connect(("", port))
s1.recv(1024)
succeeded = True
except socket.error:
pass
# General model
succeeded = False
while not succeeded:
if look_for_existing:
port = get_port_for_timblserver(foldername + "/" + backupname + ".training.txt.IGTree")
else:
port = False
if not port:
port = get_free_port()
terminal.command(
"timblserver -i "
+ foldername
+ "/"
+ backupname
+ ".training.txt.IGTree +vdb +D -a1 -G +vcf -S "
+ str(port)
+ " -C 1000"
)
s2 = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
try:
s2.connect(("", port))
s2.recv(1024)
succeeded = True
except socket.error:
pass
self.sockets = [s1, s2]
except IOError:
print("Model not found. Prepare data to create a new one:")
training_file, testfile, lexicon = ss.prepare_training_data(inp)
print("Training model")
model = ss.train_model(training_file)
finally:
ss.model = model
# If the user has his own testfile, abandon the automatically generated one
if testfile_preset:
testfile = testfile_preset
# Go do the prediction in one of the modes, with the model
if settings["mode"] == "d":
demo_mode(ss, lexicon)
elif settings["mode"] == "s":
simulation_mode(model, lexicon, testfile, settings)
elif settings["mode"] == "server":
server_mode(settings)
# Close everything
if settings["close_server"]:
terminal.command("killall timblserver")
# TODO
# Server modus
# Mogelijk: werkt het beter als je na lang doorgaan stopt met suggereren?
# Lettermodus: gaat er ergens iets fout?
# Haakje sluiten in simulatiemodus
# Backspace: fout als je over de kleurgrenzen heen backspacet gaan de kleuren raar doen