def _compute_fst_results_for_comparison(comparison_key, cohort_a, cohort_b,
otu_defs, geno_samples, data_dir,
file_owner):
noprompt = True
fst_input_etl.dump_comparison_classification(comparison_key,
cohort_a,
cohort_b,
geno_samples,
otu_defs,
data_dir,
file_owner=file_owner)
fst_input_etl.dump_feature_metadata(otu_defs,
data_dir,
file_owner=file_owner)
if DEBUGGING_CONFIG:
fst_config = fst_input_etl._make_debugging_config(
comparison_key, data_dir, otu_defs)
else:
fst_config = fst_input_etl._make_fst_config(comparison_key, data_dir,
otu_defs)
scratch_dirname = dirname_of_fst_scratch(data_dir)
file_utils.ensure_directory(scratch_dirname, file_owner=file_owner)
os.environ['JOBLIB_TEMP_FOLDER'] = scratch_dirname
fst_runner = FST(fst_config, noprompt)
print('begin FST.refuse(): ' + comparison_key)
fst_runner.refuse()
print('end FST.refuse(): ' + comparison_key)
return
def fstDistance(self, data):
fst = FST()
sampleCount = data.shape[1]
dm = np.zeros((sampleCount, sampleCount))
for i in range(0, sampleCount):
for j in range(0, sampleCount):
#extract the whole copy number profile, so the whole column
sample1Profile = data[:, i]
sample2Profile = data[:, j]
dm[i, j] = fst.computeDistance(sample1Profile, sample2Profile)
return dm
def french_count():
f = FST('french')
f.add_state('start')
f.initial_state = 'start'
for ii in xrange(10):
f.add_arc('start', 'start', [str(ii)], [kFRENCH_TRANS[ii]])
f.set_final('start')
return f
def computeCTreeError(cMatrix, realTree):
sampleNum = cMatrix.shape[1]
#Compute the distance pairwise between samples
distanceMatrix = np.empty([sampleNum, sampleNum], dtype=float)
for sample1 in range(0, sampleNum):
for sample2 in range(0, sampleNum):
#The distance can be computed for the entire column at once using the FST
dist = FST().computeDistance(cMatrix[:, sample1], cMatrix[:,
sample2])
distanceMatrix[sample1, sample2] = dist
#Compute the MST
fullGraph = generateInitialTree(distanceMatrix, realTree.vertices)
inferredTree = computeMST(fullGraph, realTree.vertices)
[
ancestrySwapErrorAbsentInInferred, ancestrySwapErrorPresentInInferred,
noOfSamplePairs
] = computeAncestrySwapError(realTree, inferredTree)
summedError = (ancestrySwapErrorAbsentInInferred +
ancestrySwapErrorPresentInInferred)
averagedAncestrySwapError = summedError / float(noOfSamplePairs)
#simulationErrorHandler = SimulationErrorHandler()
#treeScore = simulationErrorHandler.computeTreeError([mst], realTree)
return averagedAncestrySwapError
def run_experiment():
score = open("score.csv", "wt")
for k in range(1, 21):
k *= 10
if k == 0:
alphabet, states, init_state, accept_states, transitions = \
SIMPLIFIED_JSON_ALPHABET, SIMPLIFIED_JSON_STATES, SIMPLIFIED_JSON_INIT_STATE, \
SIMPLIFIED_JSON_ACCEPT_STATES, SIMPLIFIED_JSON_TRANSITIONS
else:
alphabet, states, init_state, accept_states, transitions = load_fst_by_nodes_to_add(k)
init_state = init_state[0]
fst = FST(alphabet, states, init_state, accept_states, transitions)
fst_dataset = FstDataset(BinaryFSTParams(), fst=fst)
activator_params = BinaryActivatorParams()
activator_params.EPOCHS = 100
activator = binaryActivator(BinaryModule(BinaryModuleParams(alphabet_size=len(fst_dataset.chr_embed))),
activator_params, fst_dataset, split_fst_dataset)
activator.train(validate_rate=10)
score.write(str(k) + "train_loss," + ",".join([str(v) for v in activator.loss_train_vec]) + "\n")
score.write(str(k) + "train_acc," + ",".join([str(v) for v in activator.accuracy_train_vec]) + "\n")
score.write(str(k) + "train_auc," + ",".join([str(v) for v in activator.auc_train_vec]) + "\n")
score.write(str(k) + "dev_loss," + ",".join([str(v) for v in activator.loss_dev_vec]) + "\n")
score.write(str(k) + "dev_acc," + ",".join([str(v) for v in activator.accuracy_dev_vec]) + "\n")
score.write(str(k) + "dev_auc," + ",".join([str(v) for v in activator.auc_dev_vec]) + "\n")
def fst_test2(test_string):
# print "\nThis FST removes all b's from string\n"
fst_states = ["q0"]
fst_in_alph = ["a", "b"]
fst_out_alph = ["a", "b"]
fst_start = "q0"
fst_final = ["q0"]
fst_trans = {
"q0": {
"a": ["q0", "a"],
"b": ["q0", ""]
},
}
test_fst = FST(fst_states, fst_in_alph, fst_out_alph, fst_start, fst_final,
fst_trans)
print test_string + " : " + test_fst.transduce_string(test_string)
def french_count():
f = FST('french')
f.add_state('start')
f.initial_state = 'start'
for ii in xrange(10):
f.add_arc('start', 'start', str(ii), [kFRENCH_TRANS[ii]])
f.set_final('start')
return f
def french_count():
f = FST("french")
f.add_state("start")
f.initial_state = "start"
for ii in xrange(10):
f.add_arc("start", "start", str(ii), [kFRENCH_TRANS[ii]])
f.set_final("start")
return f
def rand_fst(self, size_states, size_alphabet, num_accept_states):
alphabet, states, start_state, accept_states, transitions = [], [], "q0", [], []
for i in range(size_alphabet):
alphabet.append("sym" + str(i))
for i in range(size_states):
states.append("q" + str(i))
accept_states = [(q, 1) for q in sample(states, num_accept_states)]
for q in states:
for symbol in alphabet:
transitions.append((q, symbol, choice(states)))
return FST(alphabet, set(states), start_state, accept_states, transitions)
def fstAlleleDistance(
self, data, samples
): #this could have been more efficient when we pass objects
fst = FST()
sampleCount = data.shape[1]
dm = np.zeros((sampleCount, sampleCount))
messages = dict()
for i in range(0, sampleCount):
for j in range(0, sampleCount):
#extract the whole copy number profile, so the whole column
sample1Profile = data[:, i]
sample2Profile = data[:, j]
returnValue = fst.computeAlleleDistance(
sample1Profile, sample2Profile, samples[i], samples[j])
messages[(i, j)] = returnValue[0]
dm[i, j] = returnValue[1]
print "distances: "
print dm
return [messages, dm]
def letters_to_numbers():
"""
Returns an FST that converts letters to numbers as specified by
the soundex algorithm
"""
# Let's define our first FST
f1 = FST('soundex-generate')
# Indicate that '1' is the initial state
f1.add_state('start')
f1.add_state('next')
f1.initial_state = 'start'
# Set all the final states
f1.set_final('next')
return f1
def fst_test(test_string):
# print "\nThis FST replaces the first 'a' in a string with an 'ba'"
fst_states = ["q0", "q1"]
fst_in_alph = ["a", "b"]
fst_out_alph = ["a", "b"]
fst_start = "q0"
fst_final = ["q0", "q1"]
fst_trans = {
"q0": {
"a": ["q1", "ba"],
"b": ["q0", "b"]
},
"q1": {
"a": ["q1", "a"],
"b": ["q1", "b"]
},
}
test_fst = FST(fst_states, fst_in_alph, fst_out_alph, fst_start, fst_final,
fst_trans)
print test_string + " : " + test_fst.transduce_string(test_string)
def add_zero_padding():
# Now, the third fst - the zero-padding fst
f3 = FST('soundex-padzero')
f3.add_state('1')
f3.add_state('2')
f3.initial_state = '1'
f3.set_final('2')
return f3
def truncate_to_three_digits():
"""
Create an FST that will truncate a soundex string to three digits
"""
# Ok so now let's do the second FST, the one that will truncate
# the number of digits to 3
f2 = FST('soundex-truncate')
# Indicate initial and final states
f2.add_state('1')
f2.initial_state = '1'
f2.set_final('1')
# Add the arcs
for letter in string.letters:
f2.add_arc('1', '1', (letter), (letter))
for n in range(10):
f2.add_arc('1', '1', str(n), str(n))
return f2
def fst_from_prohibited_string(input_alphabet, output_alphabet, banned_string,
violation_name):
length = len(banned_string)
fst = FST(set([""]), "", set(), set(), "")
# Add arcs
if length > 1:
for i in range(1, length):
fst.states.add(banned_string[0:i])
add_alternation_arcs(fst, banned_string[0:i - 1],
banned_string[0:i], '_', banned_string[i - 1])
# Send a penalty arc to the longest valid suffix
fst.arcs.add(
Arc(banned_string[0:-1], longest_suffix(banned_string, fst.states),
Label('_', banned_string[-1], Counter({violation_name: 1}))))
# Add loopback arcs and return
for state in fst.states:
for char in input_alphabet:
add_elision_arc(fst, state, char)
for char in fst.chars_not_leaving(state, output_alphabet):
add_alternation_arcs(fst, state,
longest_suffix(state + char, fst.states), '_',
char)
return fst
def generate(self, analysis):
"""Generate the morphologically correct word
e.g.
p = Parser()
analysis = ['p','a','n','i','c','+past form']
p.generate(analysis)
---> 'panicked'
"""
# Let's define our first FST
f1 = FST('morphology-generate')
output = ['p', 'a', 'n', 'i', 'c', 'k', 'e', 'd']
return ''.join(output)
def parse(self, word):
"""Parse a word morphologically
e.g.
p = Parser()
word = ['p','a','n','i','c','k','i','n','g']
p.parse(word)
---> 'panic+present participle form'
"""
# Ok so now let's do the second FST
f2 = FST('morphology-parse')
output = ['p', 'a', 'n', 'i', 'c', '+present participle form']
return ''.join(output)
def computeATreeError(aMatrix, lafMatrix, afMatrix, realTree):
sampleNum = aMatrix.shape[1]
aObjMatrix = np.empty(aMatrix.shape, dtype=object)
#Convert the a matrix to an actual allele matrix
for row in range(0, aMatrix.shape[0]):
for col in range(0, aMatrix.shape[1]):
allele = aMatrix[row][col]
AOccurrences = [m.start() for m in re.finditer('A', allele)]
ACount = len(AOccurrences)
BOccurrences = [m.start() for m in re.finditer('B', allele)]
BCount = len(BOccurrences)
alleleObj = Alleles(ACount, BCount)
aObjMatrix[row][col] = alleleObj
#Compute the distance pairwise between samples
distanceMatrix = np.empty([sampleNum, sampleNum], dtype=float)
[chromosomes, positions, segmentation,
chromosomeArms] = parseReferenceFile()
for sample1 in range(0, sampleNum):
for sample2 in range(0, sampleNum):
#make a dummy sample object for the FST function
sample1Obj = Sample(None, None)
sample1Obj.measurements = LAF(lafMatrix[:, sample1], chromosomes,
positions, positions)
sample1Obj.measurements.segmentation = segmentation
sample1Obj.afMeasurements = afMatrix[:, sample1]
sample2Obj = Sample(None, None)
sample2Obj.measurements = LAF(lafMatrix[:, sample2], chromosomes,
positions, positions)
sample2Obj.measurements.segmentation = segmentation
sample2Obj.afMeasurements = afMatrix[:, sample2]
#The distance can be computed for the entire column at once using the FST
[messages,
dist] = FST().computeAlleleDistance(aObjMatrix[:, sample1],
aObjMatrix[:, sample2],
sample1Obj, sample2Obj)
distanceMatrix[sample1, sample2] = dist
#print distanceMatrix
#exit()
#Compute the MST
fullGraph = generateInitialTree(distanceMatrix, realTree.vertices)
mst = computeMST(fullGraph, realTree.vertices)
simulationErrorHandler = SimulationErrorHandler()
treeScore = simulationErrorHandler.computeTreeError([mst], realTree)
return treeScore
def letters_to_numbers():
"""
Returns an FST that converts letters to numbers as specified by
the soundex algorithm
"""
# Let's define our first FST
f1 = FST('soundex-generate')
# Indicate that '1' is the initial state
f1.add_state('start')
f1.add_state('next')
f1.initial_state = 'start'
# Set all the final states
f1.set_final('next')
return f1
def add_zero_padding():
# Now, the third fst - the zero-padding fst
f3 = FST('soundex-padzero')
f3.add_state('1')
f3.add_state('2')
f3.initial_state = '1'
f3.set_final('2')
return f3
def truncate_to_three_digits():
"""
Create an FST that will truncate a soundex string to three digits
"""
# Ok so now let's do the second FST, the one that will truncate
# the number of digits to 3
f2 = FST('soundex-truncate')
# Indicate initial and final states
f2.add_state('1')
f2.initial_state = '1'
f2.set_final('1')
return f2
def add_zero_padding():
# Now, the third fst - the zero-padding fst
#Variable aliases
start_state = 'start'
just_numbers = 'just_numbers'
letter_first = 'letter_first'
epsilons = ['e0', 'e1', 'e2', 'e3', 'e4', 'e5']
#Initialization
f3 = FST('soundex-padzero')
f3.add_state(start_state)
f3.add_state(just_numbers)
f3.add_state(letter_first)
f3.initial_state = start_state
add_numbers(f3, start_state, just_numbers)
for letter in string.ascii_letters:
f3.add_arc(start_state, letter_first, letter, letter)
build_letter_first(f3, epsilons, letter_first)
build_number_first(f3, epsilons, just_numbers)
return f3
def fst_intersect(m, n):
arcs = set()
state_lookup = dict(
(product_state(p, q), set()) for p, q in states_set_product(m, n))
start = product_state(m.start, n.start)
# Compute arcs for each state pair
for ((x, y), (z, w)) in states_mega_product(m, n):
labels_lists = similar_labels_between(m, x, y, n, z, w)
elision_arcs = set()
for labels_list in labels_lists:
arcs_by_input = set()
for (k, l) in labels_list:
add_arc = False
seg = ''
if k.output == '':
# Faithfulness constraint; cares about input
if k.input == l.input:
if k.input not in elision_arcs:
add_arc = True
seg = k.input
elision_arcs.add(seg)
elif ((k.input == '_') or
(k.input == l.input)) and (l.input not in arcs_by_input):
# Markedness constraint
add_arc = True
seg = l.input
arcs_by_input.add(seg)
elif (l.input == '_') and (k.input not in arcs_by_input):
# Markedness constraint
add_arc = True
seg = k.input
arcs_by_input.add(seg)
if add_arc:
intersection_arc = Arc(
product_state(x, z), product_state(y, w),
Label(seg, k.output, otimes(k.violation, l.violation)))
arcs.add(intersection_arc)
state_lookup[intersection_arc.start].add(intersection_arc)
# Figure out the states reachable from the start
fst_states = traverse_states(state_lookup, start)
fst = FST(fst_states, start, fst_states,
filter((lambda arc: arc.start in fst_states), arcs), 1)
return fst
def truncate_to_three_digits():
"""
Create an FST that will truncate a soundex string to three digits
"""
# Ok so now let's do the second FST, the one that will truncate
# the number of digits to 3
f2 = FST('soundex-truncate')
# Indicate initial and final states
f2.add_state('1')
f2.initial_state = '1'
f2.set_final('1')
return f2
def computeCTreeError(cMatrix, realTree):
sampleNum = cMatrix.shape[1]
#Compute the distance pairwise between samples
distanceMatrix = np.empty([sampleNum, sampleNum], dtype=float)
for sample1 in range(0, sampleNum):
for sample2 in range(0, sampleNum):
#The distance can be computed for the entire column at once using the FST
dist = FST().computeDistance(cMatrix[:, sample1], cMatrix[:,
sample2])
distanceMatrix[sample1, sample2] = dist
#Compute the MST
fullGraph = generateInitialTree(distanceMatrix, realTree.vertices)
mst = computeMST(fullGraph, realTree.vertices)
simulationErrorHandler = SimulationErrorHandler()
treeScore = simulationErrorHandler.computeTreeError([mst], realTree)
return treeScore
def generate(self, analysis):
"""Generate the morphologically correct word
e.g.
p = Parser()
analysis = ['p','a','n','i','c','+past form']
p.generate(analysis)
---> 'panicked'
"""
start_state = 'start'
f = FST('generator')
f.add_state(start_state)
f.initial_state = start_state
self._build_generator_fst(f, analysis, start_state)
return ''.join(f.transduce(analysis)[0])
def add_zero_padding():
# Now, the third fst - the zero-padding fst
f3 = FST('soundex-padzero')
states = ['1', '2', '3', '4']
for state in states:
f3.add_state(state)
f3.initial_state = '1'
f3.set_final('4')
for letter in string.letters:
f3.add_arc('1', '1', letter, letter)
for number in range(1, 10):
f3.add_arc('1', '2', str(number), str(number))
f3.add_arc('2', '3', str(number), str(number))
f3.add_arc('3', '4', str(number), str(number))
f3.add_arc('2', '4', (), '00')
f3.add_arc('3', '4', (), '0')
return f3
def french_count():
f = FST('french')
f.add_state('start')
f.add_state('z')
for i in range(30):
f.add_state(str(i))
f.initial_state = ('start')
for i in range(20, 30):
f.set_final(str(i))
f.set_final('z')
f.add_arc('start', 'z', ['z'], [kFRENCH_TRANS[0]])
for i in range(10):
f.add_arc('start', str(i), [str(i)], [])
for j in range(10, 20):
if i is 0:
f.add_arc(str(i), str(j), [str(j - 10)], [])
elif i is 1:
f.add_arc(str(i), str(j), [str(j - 10)], [kFRENCH_TRANS[100]])
elif i in range(2, 10):
f.add_arc(str(i), str(j), [str(j - 10)],
[kFRENCH_TRANS[i], kFRENCH_TRANS[100]])
for i in range(10, 20):
for j in range(20, 30):
if i is 10:
if j != 20:
f.add_arc(str(i), str(j), [str(j - 20)],
[kFRENCH_TRANS[j - 20]])
else:
f.add_arc(str(i), str(j), [str(j - 20)], [])
elif i is 11 and j in range(20, 27):
f.add_arc(str(i), str(j), [str(j - 20)],
[kFRENCH_TRANS[j - 10]])
elif i is 11 and j in range(27, 30):
f.add_arc(str(i), str(j), [str(j - 20)],
[kFRENCH_TRANS[10], kFRENCH_TRANS[j - 20]])
elif i in range(12, 17):
if j is 20:
f.add_arc(str(i), str(j), [str(j - 20)],
[kFRENCH_TRANS[int(i % 10) * 10]])
elif j is 21:
f.add_arc(str(i), str(j), [str(j - 20)], [
kFRENCH_TRANS[int(i % 10) * 10], kFRENCH_AND,
kFRENCH_TRANS[1]
])
else:
f.add_arc(str(i), str(j), [str(j - 20)], [
kFRENCH_TRANS[int(i % 10) * 10], kFRENCH_TRANS[j - 20]
])
elif i is 17:
if j is 20:
f.add_arc(str(i), str(j), [str(j - 20)],
[kFRENCH_TRANS[60], kFRENCH_TRANS[10]])
elif j is 21:
f.add_arc(
str(i), str(j), [str(j - 20)],
[kFRENCH_TRANS[60], kFRENCH_AND, kFRENCH_TRANS[11]])
elif j in range(22, 27):
f.add_arc(str(i), str(j), [str(j - 20)],
[kFRENCH_TRANS[60], kFRENCH_TRANS[j - 10]])
elif j in range(27, 30):
f.add_arc(str(i), str(j), [str(j - 20)], [
kFRENCH_TRANS[60], kFRENCH_TRANS[10],
kFRENCH_TRANS[j - 20]
])
elif i is 18:
if j is 20:
f.add_arc(str(i), str(j), [str(j - 20)],
[kFRENCH_TRANS[4], kFRENCH_TRANS[20]])
elif j in range(21, 30):
f.add_arc(str(i), str(j), [str(j - 20)], [
kFRENCH_TRANS[4], kFRENCH_TRANS[20],
kFRENCH_TRANS[j - 20]
])
elif i is 19:
if j in range(20, 27):
f.add_arc(str(i), str(j), [str(j - 20)], [
kFRENCH_TRANS[4], kFRENCH_TRANS[20],
kFRENCH_TRANS[j - 10]
])
elif j in range(27, 30):
f.add_arc(str(i), str(j), [str(j - 20)], [
kFRENCH_TRANS[4], kFRENCH_TRANS[20], kFRENCH_TRANS[10],
kFRENCH_TRANS[j - 20]
])
return f
def union_fst(self, list_fst):
new_alphabet, new_states, new_start_state, new_accept_states, new_transitions =\
self._merge_fst(list_fst, op=UNION)
return FST(new_alphabet, new_states, new_start_state, new_accept_states, new_transitions)
_states_L1 = {"s0", "s1", "s2", "s3", "s4"}
_init_state_L1 = "s0"
_accept_states_L1 = [("s2", 1), ("s4", 1)]
_transitions_L1 = [
("s0", "a", "s1"),
("s0", "b", "s3"),
("s1", "a", "s1"),
("s1", "b", "s2"),
("s2", "a", "s1"),
("s2", "b", "s2"),
("s3", "a", "s4"),
("s3", "b", "s3"),
("s4", "a", "s4"),
("s4", "b", "s3")
]
_fst_L1 = FST(_alphabet_L1, _states_L1, _init_state_L1, _accept_states_L1, _transitions_L1)
print("check FST - L1")
print(_fst_L1)
assert _fst_L1.go("aaabbababaabbab")[1]
assert not _fst_L1.go("aaabbbbba")[1]
rand = "".join(_fst_L1.go())
print("sample:" + rand)
assert _fst_L1.go(rand)
# L2
_alphabet_L2 = ["a", "b"]
_states_L2 = {"q0", "q1", "q2", "q3"}
_init_state_L2 = "q0"
_accept_states_L2 = [("q2", 3)]
_transitions_L2 = [
("q0", "a", "q1"),
def truncate_to_three_digits():
"""
Create an FST that will truncate a soundex string to three digits
"""
# Ok so now let's do the second FST, the one that will truncate
# the number of digits to 3
f2 = FST('soundex-truncate')
# Indicate initial and final states
f2.add_state('ste')
f2.add_state('L1')
f2.add_state('N1')
f2.add_state('N2')
f2.add_state('N3')
f2.add_state('next1')
f2.initial_state = 'ste'
f2.set_final('next1')
for letter in string.letters:
f2.add_arc('ste', 'L1', (letter), (letter))
for n in range(10):
f2.add_arc('ste', 'N1', (str(n)), (str(n)))
f2.add_arc('L1', 'N1', (str(n)), (str(n)))
f2.add_arc('N1', 'N2', (str(n)), (str(n)))
f2.add_arc('N2', 'N3', (str(n)), (str(n)))
f2.add_arc('N3', 'N3', (str(n)), ())
# Add the arcs
"""for letter in string.letters:
f2.add_arc('1', '1', (letter), (letter))
for n in range(10):
f2.add_arc('1', '1', (str(n)), (str(n)))"""
f2.add_arc('L1', 'next1', (), ())
f2.add_arc('N1', 'next1', (), ())
f2.add_arc('N2', 'next1', (), ())
f2.add_arc('N3', 'next1', (), ())
return f2
fst_trans[char_list[i][0]] = {}
fst_trans["start"][char_list[i][0]] = [
char_list[i][:1], arp_list[i][0] + " "
]
for i in range(0, len(char_list)):
for j in range(1, len(char_list[i])):
# adding new states to fst_states
fst_states.append(char_list[i][:j + 1])
# adding state transitions, where state names are string of chars in the color
# so far so the 3rd state on the way to blue is named 'blu'
# example in dictionary: 'ru': {'b': ['rub', 'B ']}
fst_trans[char_list[i][:j + 1]] = {}
fst_trans[char_list[i][:j]][char_list[i][j]] = [char_list[i][:j + 1]]
# if/else accounting for silent letters at end of word
if j < len(arp_list[i]):
fst_trans[char_list[i][:j]][char_list[i][j]].append(
arp_list[i][j] + " ")
else:
fst_trans[char_list[i][:j]][char_list[i][j]].append("")
# Our final states are all of our input words
fst_final = char_list
# Declaring our fst
test_fst = FST(fst_states, fst_in_alph, fst_out_alph, fst_start, fst_final,
fst_trans)
# Calling transduce on each color in our input
for word in char_list:
print test_fst.transduce_string(word)
def french_count():
f = FST('french')
f.add_state('start')
# one number and two trailing unknowns
f.add_state('n**')
# exception from state n**
f.add_state('n**+')
# two numbers and one trailing unknown
f.add_state('nn*')
# zero and two uknown digits trailing and so on
f.add_state('0**')
f.add_state('00*')
f.add_state('00n')
f.add_state('0n*')
f.add_state('0n*+')
f.add_state('0nn')
f.add_state('n00')
f.add_state('nnn')
f.add_state('nnn*')
f.add_state('*et*')
# vegasimal counting for 7 in ((0/n)n*)
f.add_state('0n*Vega7+')
f.add_state('0n*Vega7')
f.add_state('0nnVega7')
# vegasimal counting for 8 in ((0/n)n*)
f.add_state('0n*Vega8')
f.add_state('0n*Vega8+')
f.add_state('0nnVega8')
# vegasimal counting for 9 in ((0/n)n*)
f.add_state('0n*Vega9')
f.add_state('0n*Vega9+')
f.add_state('0n*Vega9++')
f.add_state('0nnVega9')
# set final states
f.set_final('00n')
f.set_final('0nn')
f.set_final('nnn')
f.set_final('n00')
f.set_final('0nnVega7')
f.set_final('0nnVega8')
f.set_final('0nnVega9')
# initial state
f.initial_state = 'start'
# remove initial zeroes
f.add_arc('start', '0**', '0', ())
f.add_arc('0**', '00*', '0', ())
for ii in xrange(10):
#from '0n*Vega8' to '0nnVega8
if ii != 0:
f.add_arc('0n*Vega8+', '0nnVega8', str(ii), [kFRENCH_TRANS[ii]])
elif ii == 0:
f.add_arc('0n*Vega8+', '0nnVega8', str(ii), ())
#from '0n*Vega7' to '0nnVega7' 7-9
if ii == 0 or ii == 7 or ii ==8 or ii == 9:
f.add_arc('0n*Vega7', '0n*Vega7+', (), [kFRENCH_TRANS[10]])
f.add_arc('0n*Vega7+', '0n*Vega7+', str(ii), [kFRENCH_TRANS[ii]])
#
f.add_arc('0n*Vega9+', '0n*Vega9++', (), [kFRENCH_TRANS[10]])
f.add_arc('0n*Vega9++', '0nnVega9', str(ii), [kFRENCH_TRANS[ii]])
if ii == 0:
f.add_arc('0n*Vega7+', '0nnVega7', '0', ())
f.add_arc('0n*Vega9++', '0nnVega9', '0', ())
elif ii == 7 or ii == 8 or ii == 9:
f.add_arc('0n*Vega7+', '0nnVega7', str(ii), [kFRENCH_TRANS[ii]])
#from '0n*Vega' to '0nnVega' 2-6
if ii == 2 or ii == 3 or ii ==4 or ii == 5 or ii == 6:
f.add_arc('0n*Vega7', '0nnVega7', str(ii), [kFRENCH_TRANS[ii+10]])
f.add_arc('0n*Vega9+', '0nnVega9', str(ii), [kFRENCH_TRANS[ii+10]])
if ii == 1:
f.add_arc('0**','0n*', str(ii), [kFRENCH_TRANS[10]])
f.add_arc('n**','0n*', str(ii), [kFRENCH_TRANS[10]])
f.add_arc('0n*Vega7', '0n*Vega7+', str(ii), [kFRENCH_AND])
f.add_arc('0n*Vega7+', '0nnVega7', str(ii), [kFRENCH_TRANS[ii+10]])
f.add_arc('0n*Vega9+', '0nnVega9', str(ii), [kFRENCH_TRANS[ii+10]])
#from '00*' to '00n'
f.add_arc('00*', '00n', str(ii), [kFRENCH_TRANS[ii]])
#from '*n*' to '*nn' 2-9
if ii != 0 and ii !=9:
f.add_arc('0n*','0nn', str(ii+1), [kFRENCH_TRANS[ii+1]])
f.add_arc('0n*+','0nn', str(ii), [kFRENCH_TRANS[ii]])
#from 'start' to 'nnn' 200,300,...,900
if ii != 0 and ii !=1:
f.add_arc('start','n**+', str(ii), [kFRENCH_TRANS[ii]])
f.add_arc('n**+', 'n**', (), [kFRENCH_TRANS[100]])
#from 'n**' to 'n0*' 0
if ii == 0:
f.add_arc('n**', 'n00', '00', ())
if ii == 1:
f.add_arc('start', 'n**', '1', [kFRENCH_TRANS[100]])
#from '*n*' to '*et*' 1
f.add_arc('0n*','*et*', '1', [kFRENCH_AND])
#from '*et*' to '*nn' 1
f.add_arc('*et*','0nn', (), [kFRENCH_TRANS[1]])
#from '0**' to '*nn' 10-16
for ii in xrange(10,17):
f.add_arc('0**','0nn', str(ii), [kFRENCH_TRANS[ii]])
f.add_arc('n**','0nn', str(ii), [kFRENCH_TRANS[ii]])
#from '0**' to '*nn' 20-60
for ii in xrange(2,7):
f.add_arc('0**', '0nn', str(ii*10), [kFRENCH_TRANS[ii*10]])
f.add_arc('n**', '0nn', str(ii*10), [kFRENCH_TRANS[ii*10]])
#from '0**', to *n*
f.add_arc('0**','0n*', str(ii), [kFRENCH_TRANS[ii*10]])
#from 'n**' to '0n*'
f.add_arc('n**', '0n*+', str(ii), [kFRENCH_TRANS[ii*10]])
for ii in xrange(7,10):
if ii == 7:
f.add_arc('0**', '0n*Vega7', str(ii), [kFRENCH_TRANS[60]])
f.add_arc('n**', '0n*Vega7', str(ii), [kFRENCH_TRANS[60]])
elif ii == 8:
f.add_arc('0**', '0n*Vega8', str(ii), [kFRENCH_TRANS[4]])
f.add_arc('n**', '0n*Vega8', str(ii), [kFRENCH_TRANS[4]])
f.add_arc('0n*Vega8', '0n*Vega8+', (), [kFRENCH_TRANS[20]])
elif ii == 9:
f.add_arc('0**', '0n*Vega9', str(ii), [kFRENCH_TRANS[4]])
f.add_arc('n**', '0n*Vega9', str(ii), [kFRENCH_TRANS[4]])
f.add_arc('0n*Vega9', '0n*Vega9+', (), [kFRENCH_TRANS[20]])
f.add_arc('n**', '0n*+', '0', ())
return f
def truncate_to_three_digits():
"""
Create an FST that will truncate a soundex string to three digits
"""
# Ok so now let's do the second FST, the one that will truncate
# the number of digits to 3
f2 = FST('soundex-truncate')
# Indicate initial and final states
f2.add_state('1')
f2.add_state('2')
f2.add_state('3')
f2.add_state('4')
f2.initial_state = '1'
f2.set_final('1')
f2.set_final('2')
f2.set_final('3')
f2.set_final('4')
# Adds letters from input string of 'A###0000'
for letter in string.letters:
f2.add_arc('1', '1', (letter), (letter))
# Adds numbers from first FST of range 0-9
for n in range(10):
f2.add_arc('1', '2', str(n), (str(n)))
f2.add_arc('2', '3', str(n), (str(n)))
f2.add_arc('3', '4', str(n), (str(n)))
f2.add_arc('4', '4', str(n), ())
return f2
def generate_control(self):
arguments = self.matchers.keys()
# this will be a hypercube
control = FST()
# zero state is for verb
control.add_state("0", is_init=True, is_final=False)
# inside states for the cube, except the last, accepting state
for i in xrange(1, pow(2, len(arguments))):
control.add_state(str(i), is_init=False, is_final=False)
# last node of the hypercube
control.add_state(
str(int(pow(2, len(arguments)))),
is_init=False, is_final=True)
# first transition
control.add_transition(KRPosMatcher("VERB"), [ExpandOperator(
self.lexicon, self.working_area)], "0", "1")
# count every transition as an increase in number of state
for path in permutations(arguments):
actual_state = 1
for arg in path:
increase = pow(2, arguments.index(arg))
new_state = actual_state + increase
control.add_transition(
self.matchers[arg],
[FillArgumentOperator(arg, self.working_area)],
str(actual_state), str(new_state))
actual_state = new_state
return control
from fst import FST
import argparse
parser = argparse.ArgumentParser()
parser.add_argument('--fst', action='store_true')
parser.add_argument('--isyms', action='store_true')
parser.add_argument('--osyms', action='store_true')
parser.add_argument('--name', type=str, required=True)
parser.add_argument('--file', type=str, required=True)
if __name__ == '__main__':
args = parser.parse_args()
fst = FST(args.name)
fst.initial_state = fst.new_state()
fst.final_states.append(fst.new_state())
with open(args.file, 'r') as f:
for line in f:
if fst.name == 'L':
word, phones = line.strip().split('\t')
tokens = phones.split()
elif fst.name == 'S':
word = line.strip().split('\t')[0]
tokens = list(word)
if len(tokens) == 1:
fst.add_arc(fst.initial_state, fst.final_states[0], word,
tokens[0])
else:
state = fst.new_state()
fst.add_arc(fst.initial_state, state, word, tokens[0])
for phone in tokens[1:-1]:
from fst import FST
# This function returns fn o ... o f3 o f2 o f1 (input)
# where ALL transducers use characters as input symbols
def compose(input, *fsts):
output_list = [input]
for fst in fsts:
next_output_list = []
for o in output_list:
new_output = ''.join(o)
next_output_list.extend(fst.transduce(new_output))
output_list = next_output_list
return output_list
if __name__ == '__main__':
f1 = FST('test-generate')
# Indicate that '1' is the initial state
f1.add_state('start')
f1.add_state('next')
f1.initial_state = 'start'
# Set all the final states
f1.set_final('next')
# Add the rest of the arcs
for letter in ['A','B','C','D']:
f1.add_arc('start', 'next', letter, '1')
f1.add_arc('next', 'next', letter, '0')
f2 = FST('test-generate')
def letters_to_numbers():
"""
Returns an FST that converts letters to numbers as specified by
the soundex algorithm
"""
vowels = [
'a', 'A', 'e', 'E', 'h', 'H', 'i', 'I', 'o', 'O', 'u', 'U', 'w', 'W',
'y', 'Y'
]
# Let's define our first FST
f1 = FST('soundex-generate')
# Indicate that '1' is the initial state
f1.add_state('start')
f1.add_state('s11')
f1.add_state('s22')
f1.add_state('s33')
f1.add_state('s44')
f1.add_state('s55')
f1.add_state('s66')
f1.add_state('s1')
f1.add_state('s2')
f1.add_state('s3')
f1.add_state('sv')
f1.add_state('s4')
f1.add_state('s5')
f1.add_state('s6')
f1.add_state('next')
f1.initial_state = 'start'
# Set all the final states
f1.set_final('next')
# Add the rest of the arcs
for letter in string.ascii_letters:
#f1.add_arc('start', 'next', (letter), (letter))
#f1.add_arc('next', 'next', (letter), ('0'))
if letter in vowels:
f1.add_arc('start', 'sv', (letter), (letter))
f1.add_arc('s11', 'sv', (letter), ())
f1.add_arc('s33', 'sv', (letter), ())
f1.add_arc('s22', 'sv', (letter), ())
f1.add_arc('s44', 'sv', (letter), ())
f1.add_arc('s55', 'sv', (letter), ())
f1.add_arc('s66', 'sv', (letter), ())
f1.add_arc('sv', 'sv', (letter), ())
f1.add_arc('s1', 'sv', (letter), ())
f1.add_arc('s2', 'sv', (letter), ())
f1.add_arc('s3', 'sv', (letter), ())
f1.add_arc('s4', 'sv', (letter), ())
f1.add_arc('s6', 'sv', (letter), ())
f1.add_arc('s5', 'sv', (letter), ())
#f1.add_arc('s3','s4',(letter),(letter))
elif letter in "Ll":
f1.add_arc('start', 's44', (letter), (letter))
f1.add_arc('s44', 's4', (letter), ())
f1.add_arc('s11', 's4', (letter), ('4'))
f1.add_arc('s22', 's4', (letter), ('4'))
f1.add_arc('s33', 's4', (letter), ('4'))
f1.add_arc('s55', 's4', (letter), ('4'))
f1.add_arc('s66', 's4', (letter), ('4'))
f1.add_arc('s4', 's4', (letter), ())
f1.add_arc('s1', 's4', (letter), ('4'))
f1.add_arc('s2', 's4', (letter), ('4'))
f1.add_arc('s3', 's4', (letter), ('4'))
f1.add_arc('s5', 's4', (letter), ('4'))
f1.add_arc('s6', 's4', (letter), ('4'))
f1.add_arc('sv', 's4', (letter), ('4'))
elif letter in 'Rr':
f1.add_arc('start', 's66', (letter), (letter))
f1.add_arc('s66', 's6', (letter), ())
f1.add_arc('s22', 's6', (letter), ('6'))
f1.add_arc('s33', 's6', (letter), ('6'))
f1.add_arc('s44', 's6', (letter), ('6'))
f1.add_arc('s55', 's6', (letter), ('6'))
f1.add_arc('s11', 's6', (letter), ('6'))
f1.add_arc('s6', 's6', (letter), ())
f1.add_arc('s1', 's6', (letter), ('6'))
f1.add_arc('s2', 's6', (letter), ('6'))
f1.add_arc('s3', 's6', (letter), ('6'))
f1.add_arc('s5', 's6', (letter), ('6'))
f1.add_arc('s4', 's6', (letter), ('6'))
f1.add_arc('sv', 's6', (letter), ('6'))
elif letter in "bfpvBFPV":
f1.add_arc('start', 's11', (letter), (letter))
f1.add_arc('s11', 's1', (letter), ())
f1.add_arc('s22', 's1', (letter), ('1'))
f1.add_arc('s33', 's1', (letter), ('1'))
f1.add_arc('s44', 's1', (letter), ('1'))
f1.add_arc('s55', 's1', (letter), ('1'))
f1.add_arc('s66', 's1', (letter), ('1'))
f1.add_arc('s1', 's1', (letter), ())
f1.add_arc('s5', 's1', (letter), ('1'))
f1.add_arc('s2', 's1', (letter), ('1'))
f1.add_arc('s3', 's1', (letter), ('1'))
f1.add_arc('s4', 's1', (letter), ('1'))
f1.add_arc('sv', 's1', (letter), ('1'))
f1.add_arc('s6', 's1', (letter), ('1'))
elif letter in "cgjkqsxzCGJKQSXZ":
f1.add_arc('start', 's22', (letter), (letter))
f1.add_arc('s22', 's2', (letter), ())
f1.add_arc('s11', 's2', (letter), ('2'))
f1.add_arc('s33', 's2', (letter), ('2'))
f1.add_arc('s44', 's2', (letter), ('2'))
f1.add_arc('s55', 's2', (letter), ('2'))
f1.add_arc('s66', 's2', (letter), ('2'))
f1.add_arc('s2', 's2', (letter), ())
f1.add_arc('s5', 's2', (letter), ('2'))
f1.add_arc('s1', 's2', (letter), ('2'))
f1.add_arc('s3', 's2', (letter), ('2'))
f1.add_arc('s4', 's2', (letter), ('2'))
f1.add_arc('sv', 's2', (letter), ('2'))
f1.add_arc('s6', 's2', (letter), ('2'))
elif letter in "mnMN":
f1.add_arc('start', 's55', (letter), (letter))
f1.add_arc('s55', 's5', (letter), ())
f1.add_arc('s11', 's5', (letter), ('5'))
f1.add_arc('s44', 's5', (letter), ('5'))
f1.add_arc('s33', 's5', (letter), ('5'))
f1.add_arc('s22', 's5', (letter), ('5'))
f1.add_arc('s66', 's5', (letter), ('5'))
f1.add_arc('s5', 's5', (letter), ())
f1.add_arc('s2', 's5', (letter), ('5'))
f1.add_arc('s1', 's5', (letter), ('5'))
f1.add_arc('s3', 's5', (letter), ('5'))
f1.add_arc('s4', 's5', (letter), ('5'))
f1.add_arc('sv', 's5', (letter), ('5'))
f1.add_arc('s6', 's5', (letter), ('5'))
elif letter in "dtDT":
f1.add_arc('start', 's33', (letter), (letter))
f1.add_arc('s33', 's3', (letter), ())
f1.add_arc('s11', 's3', (letter), ('3'))
f1.add_arc('s44', 's3', (letter), ('3'))
f1.add_arc('s55', 's3', (letter), ('3'))
f1.add_arc('s22', 's3', (letter), ('3'))
f1.add_arc('s66', 's3', (letter), ('3'))
f1.add_arc('s3', 's3', (letter), ())
f1.add_arc('s2', 's3', (letter), ('3'))
f1.add_arc('s1', 's3', (letter), ('3'))
f1.add_arc('s5', 's3', (letter), ('3'))
f1.add_arc('s4', 's3', (letter), ('3'))
f1.add_arc('sv', 's3', (letter), ('3'))
f1.add_arc('s6', 's3', (letter), ('3'))
""" else:
f1.add_arc('s1','s5',(letter),('1'))
f1.add_arc('s4','s5',(letter),('1'))
f1.add_arc('s6','s5',(letter),('1'))
f1.add_arc('s44','s5',(letter),('1'))
f1.add_arc('s66','s5',(letter),('1'))
f1.add_arc('s3','s5',(letter),('1'))
#f1.add_arc('s5','s5',(letter),()) """
f1.add_arc('s11', 'next', (), ())
f1.add_arc('s22', 'next', (), ())
f1.add_arc('s33', 'next', (), ())
f1.add_arc('s44', 'next', (), ())
f1.add_arc('s55', 'next', (), ())
f1.add_arc('s66', 'next', (), ())
f1.add_arc('s1', 'next', (), ())
f1.add_arc('s2', 'next', (), ())
f1.add_arc('s3', 'next', (), ())
f1.add_arc('sv', 'next', (), ())
f1.add_arc('s4', 'next', (), ())
f1.add_arc('s5', 'next', (), ())
f1.add_arc('s6', 'next', (), ())
return f1
def letters_to_numbers():
"""
Returns an FST that converts letters to numbers as specified by
the soundex algorithm
"""
# Let's define our first FST
f1 = FST('soundex-generate')
# Indicate that '1' is the initial state
f1.add_state('start')
f1.add_state('next')
f1.add_state('one')
f1.add_state('two')
f1.add_state('three')
f1.add_state('four')
f1.add_state('five')
f1.add_state('six')
f1.initial_state = 'start'
# Set all the final states
f1.set_final('next')
f1.set_final('one')
f1.set_final('two')
f1.set_final('three')
f1.set_final('four')
f1.set_final('five')
f1.set_final('six')
list_one = ['b', 'f', 'p', 'v']
list_two = ['c', 'g', 'j', 'k', 'q', 's', 'x', 'z']
list_three = ['d', 't']
list_four = ['l']
list_five = ['m', 'n']
list_six = ['r']
vowels = ['a', 'e', 'h', 'i', 'o', 'u', 'w', 'y']
# Add the rest of the arcs
# changed string.ascii_lowercase to string.letters
for letter in string.letters:
f1.add_arc('start', 'next', (letter), (letter))
for letter in string.letters:
if letter in list_one:
f1.add_arc('next', 'one', (letter), '1')
elif letter in list_two:
f1.add_arc('next', 'two', (letter), '2')
elif letter in list_three:
f1.add_arc('next', 'three', (letter), '3')
elif letter in list_four:
f1.add_arc('next', 'four', (letter), '4')
elif letter in list_five:
f1.add_arc('next', 'five', (letter), '5')
elif letter in list_six:
f1.add_arc('next', 'six', (letter), '6')
else:
f1.add_arc('next', 'next', (letter), ())
for letter in string.letters:
if letter in list_two:
f1.add_arc('one', 'two', (letter), '2')
elif letter in list_three:
f1.add_arc('one', 'three', (letter), '3')
elif letter in list_four:
f1.add_arc('one', 'four', (letter), '4')
elif letter in list_five:
f1.add_arc('one', 'five', (letter), '5')
elif letter in list_six:
f1.add_arc('one', 'six', (letter), '6')
else:
f1.add_arc('one', 'one', (letter), ())
for letter in string.letters:
if letter in list_one:
f1.add_arc('two', 'one', (letter), '1')
elif letter in list_three:
f1.add_arc('two', 'three', (letter), '3')
elif letter in list_four:
f1.add_arc('two', 'four', (letter), '4')
elif letter in list_five:
f1.add_arc('two', 'five', (letter), '5')
elif letter in list_six:
f1.add_arc('two', 'six', (letter), '6')
else:
f1.add_arc('two', 'two', (letter), ())
for letter in string.letters:
if letter in list_one:
f1.add_arc('three', 'one', (letter), '1')
elif letter in list_two:
f1.add_arc('three', 'two', (letter), '2')
elif letter in list_four:
f1.add_arc('three', 'four', (letter), '4')
elif letter in list_five:
f1.add_arc('three', 'five', (letter), '5')
elif letter in list_six:
f1.add_arc('three', 'six', (letter), '6')
else:
f1.add_arc('three', 'three', (letter), ())
for letter in string.letters:
if letter in list_one:
f1.add_arc('four', 'one', (letter), '1')
elif letter in list_two:
f1.add_arc('four', 'two', (letter), '2')
elif letter in list_three:
f1.add_arc('four', 'three', (letter), '3')
elif letter in list_five:
f1.add_arc('four', 'five', (letter), '5')
elif letter in list_six:
f1.add_arc('four', 'six', (letter), '6')
else:
f1.add_arc('four', 'four', (letter), ())
for letter in string.letters:
if letter in list_one:
f1.add_arc('five', 'one', (letter), '1')
elif letter in list_two:
f1.add_arc('five', 'two', (letter), '2')
elif letter in list_three:
f1.add_arc('five', 'three', (letter), '3')
elif letter in list_four:
f1.add_arc('five', 'four', (letter), '4')
elif letter in list_six:
f1.add_arc('five', 'six', (letter), '6')
else:
f1.add_arc('five', 'five', (letter), ())
for letter in string.letters:
if letter in list_one:
f1.add_arc('six', 'one', (letter), '1')
elif letter in list_two:
f1.add_arc('six', 'two', (letter), '2')
elif letter in list_three:
f1.add_arc('six', 'three', (letter), '3')
elif letter in list_four:
f1.add_arc('six', 'four', (letter), '4')
elif letter in list_five:
f1.add_arc('six', 'five', (letter), '5')
else:
f1.add_arc('six', 'six', (letter), ())
return f1
def add_zero_padding():
# Now, the third fst - the zero-padding fst
f3 = FST('soundex-padzero')
f3.add_state('st')
f3.add_state('L1')
f3.add_state('N1')
f3.add_state('N2')
f3.add_state('N3')
f3.add_state('P1')
f3.add_state('P2')
f3.add_state('P3')
f3.initial_state = 'st'
f3.set_final('N3')
f3.set_final('P3')
for letter in string.letters:
f3.add_arc('st', 'L1', (letter), (letter))
for number in xrange(10):
f3.add_arc('st', 'N1', (str(number)), (str(number)))
f3.add_arc('L1', 'N1', (str(number)), (str(number)))
f3.add_arc('N1', 'N2', (str(number)), (str(number)))
f3.add_arc('N2', 'N3', (str(number)), (str(number)))
f3.add_arc('L1', 'P1', (), ('0'))
f3.add_arc('N1', 'P2', (), ('0'))
f3.add_arc('N2', 'P3', (), ('0'))
f3.add_arc('P1', 'P2', (), ('0'))
f3.add_arc('P2', 'P3', (), ('0'))
return f3
def french_count():
f = FST('french')
f.add_state('start')
f.initial_state = 'start'
f.add_state('1stzero')
f.add_state('tens')
f.add_state('seventeen')
f.add_state('final_seventeen')
f.add_state('eighteen')
f.add_state('final_eighteen')
f.add_state('nineteen')
f.add_state('final_nineteen')
f.add_state('zero')
f.add_state('ones')
f.add_state('20-69')
f.add_state('70-ten')
f.add_state('80s')
f.add_state('90s')
f.add_state('100s')
f.add_state('et')
f.add_state('10-et')
f.add_state('et-un')
f.add_state('et-onze')
f.set_final('zero')
f.set_final('ones')
f.set_final('tens')
f.set_final('final_seventeen')
f.set_final('final_eighteen')
f.set_final('final_nineteen')
f.set_final('20-69')
f.set_final('70-ten')
f.set_final('80s')
f.set_final('90s')
f.set_final('et-un')
f.set_final('et-onze')
# 100 - 999
f.add_arc('start', '1stzero', '1', [kFRENCH_TRANS[100]])
for i in range(2, 10):
f.add_arc('start', '100s', str(i), [kFRENCH_TRANS[i]])
f.add_arc('100s', '1stzero', (), [kFRENCH_TRANS[100]])
# 0 - 9
f.add_arc('start', '1stzero', '0', [])
f.add_arc('1stzero', 'ones', '0', [])
for ii in range(1, 10):
f.add_arc('ones', 'ones', str(ii), [kFRENCH_TRANS[ii]])
f.add_arc('ones', 'ones', '0', [])
# for i in range(10):
# f.add_arc('ten-6', 'ten-6', str(i), kFRENCH_TRANS[(i+10])
# 10 - 16
f.add_arc('1stzero', 'tens', '1', [])
f.add_arc('tens', 'tens', '0', [kFRENCH_TRANS[10]])
f.add_arc('tens', 'tens', '1', [kFRENCH_TRANS[11]])
f.add_arc('tens', 'tens', '2', [kFRENCH_TRANS[12]])
f.add_arc('tens', 'tens', '3', [kFRENCH_TRANS[13]])
f.add_arc('tens', 'tens', '4', [kFRENCH_TRANS[14]])
f.add_arc('tens', 'tens', '5', [kFRENCH_TRANS[15]])
f.add_arc('tens', 'tens', '6', [kFRENCH_TRANS[16]])
f.add_arc('tens', 'seventeen', '7', [kFRENCH_TRANS[10]])
f.add_arc('seventeen', 'final_seventeen', (), [kFRENCH_TRANS[7]])
f.add_arc('tens', 'eighteen', '8', [kFRENCH_TRANS[10]])
f.add_arc('eighteen', 'final_eighteen', (), [kFRENCH_TRANS[8]])
f.add_arc('tens', 'nineteen', '9', [kFRENCH_TRANS[10]])
f.add_arc('nineteen', 'final_nineteen', (), [kFRENCH_TRANS[9]])
# 20 - 69
f.add_arc('1stzero', '20-69', '2', [kFRENCH_TRANS[20]])
f.add_arc('1stzero', '20-69', '3', [kFRENCH_TRANS[30]])
f.add_arc('1stzero', '20-69', '4', [kFRENCH_TRANS[40]])
f.add_arc('1stzero', '20-69', '5', [kFRENCH_TRANS[50]])
f.add_arc('1stzero', '20-69', '6', [kFRENCH_TRANS[60]])
# special cases:
for i in range(2, 10):
f.add_arc('20-69', '20-69', str(i), [kFRENCH_TRANS[i]])
# handles 20, 30 ... 60
for i in range(20, 60, 10):
f.add_arc('20-69', '20-69', '0', [])
# handles 21, 31, ... 61
f.add_arc('20-69', 'et', '1', [kFRENCH_AND])
f.add_arc('et', 'et-un', (),[kFRENCH_TRANS[1]])
# 70 - 79
f.add_arc('1stzero', '70-ten', '7', [kFRENCH_TRANS[60]])
f.add_arc('70-ten', '70-ten', '0', [kFRENCH_TRANS[10]])
# handle 71 here
f.add_arc('70-ten', '10-et', '1', [kFRENCH_AND])
f.add_arc('10-et', 'et-onze', (),[kFRENCH_TRANS[11]])
f.add_arc('70-ten', '70-ten', '2', [kFRENCH_TRANS[12]])
f.add_arc('70-ten', '70-ten', '3', [kFRENCH_TRANS[13]])
f.add_arc('70-ten', '70-ten', '4', [kFRENCH_TRANS[14]])
f.add_arc('70-ten', '70-ten', '5', [kFRENCH_TRANS[15]])
f.add_arc('70-ten', '70-ten', '6', [kFRENCH_TRANS[16]])
f.add_arc('70-ten', 'seventeen', '7', [kFRENCH_TRANS[10]])
f.add_arc('seventeen', 'final_seventeen', (), [kFRENCH_TRANS[7]])
f.add_arc('70-ten', 'eighteen', '8', [kFRENCH_TRANS[10]])
f.add_arc('eighteen', 'final_eighteen', (), [kFRENCH_TRANS[8]])
f.add_arc('70-ten', 'nineteen', '9', [kFRENCH_TRANS[10]])
f.add_arc('nineteen', 'final_nineteen', (), [kFRENCH_TRANS[9]])
# 80 - 89
f.add_arc('1stzero', '80s', '8', [kFRENCH_TRANS[4]])
f.add_arc('80s', 'ones', (), [kFRENCH_TRANS[20]])
f.add_arc('80s', '80s', '0', [kFRENCH_TRANS[20]])
# 90 - 99
f.add_arc('1stzero', '90s', '9', [kFRENCH_TRANS[4]])
f.add_arc('90s', 'tens', (), [kFRENCH_TRANS[20]])
return f
def intersection_fst(self, list_fst):
new_alphabet, new_states, new_start_state, new_accept_states, new_transitions =\
self._merge_fst(list_fst, op=INTERSECT)
return FST(new_alphabet, new_states, new_start_state, new_accept_states, new_transitions)
def letters_to_numbers():
"""
Returns an FST that converts letters to numbers as specified by
the soundex algorithm
"""
# Let's define our first FST
f1 = FST('soundex-generate')
# Indicate that '1' is the initial state
f1.add_state('start')
f1.add_state('next')
f1.initial_state = 'start'
# Set all the final states
f1.set_final('next')
# Add the rest of the arcs
for letter in string.ascii_lowercase:
f1.add_arc('start', 'next', (letter), (letter))
f1.add_arc('next', 'next', (letter), '0')
return f1
def letters_to_numbers():
"""
Returns an FST that converts letters to numbers as specified by
the soundex algorithm
"""
# Let's define our first FST
f1 = FST('soundex-generate')
aeoy = ['a','e','h','i','o','u','w','y']
one = ['b','f','p','v']
two = ['c','g','j','k','q','s','x','z']
three = ['d','t']
four = ['l']
five = ['m','n']
six = ['r']
# Indicate that '1' is the initial state
f1.add_state('initial')
f1.add_state('0')
f1.add_state('1')
f1.add_state('2')
f1.add_state('3')
f1.add_state('4')
f1.add_state('5')
f1.add_state('6')
f1.initial_state = 'initial'
# Set all the final states
f1.set_final('0')
f1.set_final('1')
f1.set_final('2')
f1.set_final('3')
f1.set_final('4')
f1.set_final('5')
f1.set_final('6')
# Add the rest of the arcs
for letter in string.ascii_letters:
f1.add_arc('initial','0',(letter),(letter))
if letter in aeoy:
f1.add_arc('0','0', (letter), ())
f1.add_arc('1','0', (letter), ())
f1.add_arc('2','0', (letter), ())
f1.add_arc('3','0', (letter), ())
f1.add_arc('4','0', (letter), ())
f1.add_arc('5','0', (letter), ())
f1.add_arc('6','0', (letter), ())
else:
if letter in one:
f1.add_arc('0','1', (letter), '1')
f1.add_arc('2','1', (letter), '1')
f1.add_arc('3','1', (letter), '1')
f1.add_arc('4','1', (letter), '1')
f1.add_arc('5','1', (letter), '1')
f1.add_arc('6','1', (letter), '1')
f1.add_arc('1','0', (letter), ())
if letter in two:
f1.add_arc('0','2', (letter), '2')
f1.add_arc('1','2', (letter), '2')
f1.add_arc('3','2', (letter), '2')
f1.add_arc('4','2', (letter), '2')
f1.add_arc('5','2', (letter), '2')
f1.add_arc('6','2', (letter), '2')
f1.add_arc('2','0', (letter), ())
if letter in three:
f1.add_arc('0','3', (letter), '3')
f1.add_arc('1','3', (letter), '3')
f1.add_arc('2','3', (letter), '3')
f1.add_arc('4','3', (letter), '3')
f1.add_arc('5','3', (letter), '3')
f1.add_arc('6','3', (letter), '3')
f1.add_arc('3','0', (letter), ())
if letter in four:
f1.add_arc('0','4', (letter), '4')
f1.add_arc('1','4', (letter), '4')
f1.add_arc('2','4', (letter), '4')
f1.add_arc('3','4', (letter), '4')
f1.add_arc('5','4', (letter), '4')
f1.add_arc('6','4', (letter), '4')
f1.add_arc('4','0', (letter), ())
if letter in five:
f1.add_arc('0','5', (letter), '5')
f1.add_arc('1','5', (letter), '5')
f1.add_arc('2','5', (letter), '5')
f1.add_arc('3','5', (letter), '5')
f1.add_arc('4','5', (letter), '5')
f1.add_arc('6','5', (letter), '5')
f1.add_arc('5','0', (letter), ())
if letter in six:
f1.add_arc('0','6', (letter), '6')
f1.add_arc('1','6', (letter), '6')
f1.add_arc('2','6', (letter), '6')
f1.add_arc('3','6', (letter), '6')
f1.add_arc('4','6', (letter), '6')
f1.add_arc('5','6', (letter), '6')
f1.add_arc('6','0', (letter), ())
return f1
def truncate_to_three_digits():
"""
Create an FST that will truncate a soundex string to three digits
"""
# Ok so now let's do the second FST, the one that will truncate
# the number of digits to 3
f2 = FST('soundex-truncate')
# Indicate initial and final states
states = ['1', 'd1', 'd2', 'd3']
for state in states:
f2.add_state(state)
f2.initial_state = '1'
for state in ['d1', 'd2', 'd3']:
f2.set_final(state)
# Add the arcs
for letter in string.letters:
f2.add_arc('1', '1', (letter), (letter))
for index, state in enumerate(states):
if index > 0:
for n in range(10):
f2.add_arc(states[index-1], states[index], str(n), str(n))
for n in range(10):
f2.add_arc('d3', 'd3', str(n), ())
return f2
def french_count():
f = FST('french')
f.add_state('0')
f.add_state('1')
f.add_state('2')
f.add_state('3')
f.add_state('4')
f.add_state('5')
f.add_state('6')
f.add_state('7')
f.add_state('8')
f.add_state('9')
f.add_state('10')
f.add_state('11')
f.add_state('12')
f.add_state('13')
f.add_state('14')
f.add_state('15')
f.add_state('16')
f.add_state('17')
f.add_state('18')
f.add_state('19')
f.add_state('20')
f.add_state('21')
f.add_state('22')
f.add_state('23')
f.add_state('24')
f.add_state('25')
f.initial_state = '0'
f.set_final('1')
f.set_final('3')
f.set_final('6')
f.set_final('7')
f.set_final('8')
f.set_final('9')
f.set_final('11')
f.set_final('13')
f.set_final('14')
f.set_final('18')
f.set_final('20')
zero = [0]
one = [1]
two_to_six = [2,3,4,5,6]
one_to_six = [1,2,3,4,5,6]
seven = [7]
seven_eight_nine = [7,8,9]
eight = [8]
nine = [9]
singles_all = [1,2,3,4,5,6,7,8,9]
singles = [2,3,4,5,6,7,8,9]
tens = [20,30,40,50]
# Edge from initial to final, if preceding zero in input
for i in zero:
# f.add_arc('0','9', str(i), [kFRENCH_TRANS[i]])
f.add_arc('0','0', str(i), ())
f.add_arc('4','6', str(i), ())
f.add_arc('5','8', str(i), ())
f.add_arc('0','9', str(i), [kFRENCH_TRANS[i]])
f.add_arc('10','11', str(i), [kFRENCH_TRANS[i+10]])
f.add_arc('12','13', str(i), [kFRENCH_TRANS[20]])
f.add_arc('16','18', str(i), [kFRENCH_TRANS[20],kFRENCH_TRANS[10]])
f.add_arc('17','19', str(i), ())
f.add_arc('19','9', str(i), ())
for i in one:
f.add_arc('0','2', str(i), ())
f.add_arc('17','2', str(i), ())
f.add_arc('0','17', str(i), [kFRENCH_TRANS[100]])
f.add_arc('0','5', str(i), [kFRENCH_TRANS[i*10]])
f.add_arc('17','5', str(i), [kFRENCH_TRANS[i*10]])
f.add_arc('4','7', str(i), [kFRENCH_AND, kFRENCH_TRANS[i]])
f.add_arc('10','11', str(i), [kFRENCH_AND, kFRENCH_TRANS[i+10]])
f.add_arc('12','14', str(i), [kFRENCH_TRANS[20], kFRENCH_AND, kFRENCH_TRANS[i]])
f.add_arc('16','20', str(i), [kFRENCH_TRANS[20], kFRENCH_AND, kFRENCH_TRANS[i+10]])
for i in one_to_six:
f.add_arc('2','3', str(i), [kFRENCH_TRANS[i+10]])
for i in two_to_six:
f.add_arc('0','4', str(i), [kFRENCH_TRANS[i*10]])
f.add_arc('17','4', str(i), [kFRENCH_TRANS[i*10]])
f.add_arc('10','11', str(i), [kFRENCH_TRANS[i+10]])
f.add_arc('16','20', str(i), [kFRENCH_TRANS[20],kFRENCH_TRANS[i+10]])
for i in singles:
f.add_arc('4','7', str(i), [kFRENCH_TRANS[i]])
f.add_arc('0','17', str(i), [kFRENCH_TRANS[i],kFRENCH_TRANS[100]])
f.add_arc('12','14', str(i), [kFRENCH_TRANS[20], kFRENCH_TRANS[i]])
for i in singles_all:
f.add_arc('0','1', str(i), [kFRENCH_TRANS[i]])
f.add_arc('19','1', str(i), [kFRENCH_TRANS[i]])
for i in seven_eight_nine:
f.add_arc('5','8', str(i), [kFRENCH_TRANS[i]])
f.add_arc('10','11', str(i), [kFRENCH_TRANS[10], kFRENCH_TRANS[i]])
f.add_arc('16','20', str(i), [kFRENCH_TRANS[20], kFRENCH_TRANS[10], kFRENCH_TRANS[i]])
for i in seven:
f.add_arc('0','10',str(i), [kFRENCH_TRANS[60]])
f.add_arc('17','10',str(i), [kFRENCH_TRANS[60]])
for i in eight:
f.add_arc('0','12',str(i), [kFRENCH_TRANS[4]])
f.add_arc('17','12',str(i), [kFRENCH_TRANS[4]])
for i in nine:
f.add_arc('0','16',str(i), [kFRENCH_TRANS[4]])
f.add_arc('17','16',str(i), [kFRENCH_TRANS[4]])
return f
def letters_to_numbers():
"""
Returns an FST that converts letters to numbers as specified by
the soundex algorithm
"""
# Let's define our first FST
f1 = FST('soundex-generate')
# Indicate that '1' is the initial state
states = ['q1', 'q2', 'n1', 'n2', 'n3', 'n4', 'n5', 'n6']
for state in states:
f1.add_state(state)
f1.initial_state = 'q1'
# Set all the final states
for state in ['q2', 'n1', 'n2', 'n3', 'n4', 'n5', 'n6']:
f1.set_final(state)
# Add the rest of the arcs
for letter in string.ascii_lowercase:
f1.add_arc('q1', 'q2', (letter), (letter))
if letter in set('aehiouwy'):
for state in ['q2', 'n1', 'n2', 'n3', 'n4', 'n5', 'n6']:
f1.add_arc(state, state, (letter), ())
else:
if letter in set('bfpv'):
for state in ['q2', 'n2', 'n3', 'n4', 'n5', 'n6']:
f1.add_arc(state, 'n1', (letter), ('1'))
f1.add_arc('n1', 'n1', (letter), ())
elif letter in set('cgjkqsxz'):
for state in ['q2', 'n1', 'n3', 'n4', 'n5', 'n6']:
f1.add_arc(state, 'n2', (letter), ('2'))
f1.add_arc('n2', 'n2', (letter), ())
elif letter in set('dt'):
for state in ['q2', 'n1', 'n2', 'n4', 'n5', 'n6']:
f1.add_arc(state, 'n3', (letter), ('3'))
f1.add_arc('n3', 'n3', (letter), ())
elif letter in set('l'):
for state in ['q2', 'n1', 'n2', 'n3', 'n5', 'n6']:
f1.add_arc(state, 'n4', (letter), ('4'))
f1.add_arc('n4', 'n4', (letter), ())
elif letter in set('mn'):
for state in ['q2', 'n1', 'n2', 'n3', 'n4', 'n6']:
f1.add_arc(state, 'n5', (letter), ('5'))
f1.add_arc('n5', 'n5', (letter), ())
elif letter in set('r'):
for state in ['q2', 'n1', 'n2', 'n3', 'n4', 'n5']:
f1.add_arc(state, 'n6', (letter), ('6'))
f1.add_arc('n6', 'n6', (letter), ())
return f1
def french_count():
f = FST('french')
f.add_state('start')
f.add_state('0xx')
f.add_state('nxx')
f.add_state('00x')
f.add_state('n0x')
f.add_state('n1x')
f.add_state('nnx')
f.add_state('n7x')
f.add_state('n8x')
f.add_state('n9x')
f.add_state('last')
f.initial_state = 'start'
f.set_final('last')
for ii in xrange(10):
f.add_arc('00x', 'last', [str(ii)], [kFRENCH_TRANS[ii]])
if ii == 0:
f.add_arc('start', '0xx', [str(ii)], ())
f.add_arc('0xx', '00x', [str(ii)], ())
f.add_arc('nxx', 'n0x', [str(ii)], ())
f.add_arc('n0x', 'last', [str(ii)], ())
f.add_arc('n1x', 'last', [str(ii)], [kFRENCH_TRANS[ii + 10]])
f.add_arc('nnx', 'last', [str(ii)], ())
f.add_arc('n7x', 'last', [str(ii)], [kFRENCH_TRANS[ii+10]])
f.add_arc('n8x', 'last', [str(ii)], ())
f.add_arc('n9x', 'last', [str(ii)], [kFRENCH_TRANS[ii+10]])
if ii == 1:
f.add_arc('start', 'nxx', [str(ii)], [kFRENCH_TRANS[100]])
f.add_arc('0xx', 'n1x', [str(ii)], ())
f.add_arc('nxx', 'n1x', [str(ii)], ())
f.add_arc('n0x', 'last', [str(ii)], [kFRENCH_TRANS[ii]])
f.add_arc('n1x', 'last', [str(ii)], [kFRENCH_TRANS[ii + 10]])
f.add_arc('nnx', 'last', [str(ii)], [kFRENCH_AND,kFRENCH_TRANS[ii]])
f.add_arc('n7x', 'last', [str(ii)], [kFRENCH_AND,kFRENCH_TRANS[ii+10]])
f.add_arc('n8x', 'last', [str(ii)], [kFRENCH_TRANS[ii]])
f.add_arc('n9x', 'last', [str(ii)], [kFRENCH_TRANS[ii+10]])
if ii in range(2,7):
f.add_arc('start', 'nxx', [str(ii)], [kFRENCH_TRANS[ii],kFRENCH_TRANS[100]])
f.add_arc('0xx', 'nnx', [str(ii)], [kFRENCH_TRANS[ii * 10]])
f.add_arc('nxx', 'nnx', [str(ii)], [kFRENCH_TRANS[ii * 10]])
f.add_arc('n0x', 'last', [str(ii)], [kFRENCH_TRANS[ii]])
f.add_arc('n1x', 'last', [str(ii)], [kFRENCH_TRANS[ii + 10]])
f.add_arc('nnx', 'last', [str(ii)], [kFRENCH_TRANS[ii]])
f.add_arc('n7x', 'last', [str(ii)], [kFRENCH_TRANS[ii+10]])
f.add_arc('n8x', 'last', [str(ii)], [kFRENCH_TRANS[ii]])
f.add_arc('n9x', 'last', [str(ii)], [kFRENCH_TRANS[ii + 10]])
if ii == 7:
f.add_arc('start', 'nxx', [str(ii)], [kFRENCH_TRANS[ii],kFRENCH_TRANS[100]])
f.add_arc('0xx', 'n7x', [str(ii)], [kFRENCH_TRANS[6 * 10]])
f.add_arc('nxx', 'n7x', [str(ii)], [kFRENCH_TRANS[6*10]])
f.add_arc('n0x', 'last', [str(ii)], [kFRENCH_TRANS[ii]])
f.add_arc('n1x', 'last', [str(ii)], [kFRENCH_TRANS[10], kFRENCH_TRANS[ii]])
f.add_arc('nnx', 'last', [str(ii)], [kFRENCH_TRANS[ii]])
f.add_arc('n7x', 'last', [str(ii)], [kFRENCH_TRANS[10], kFRENCH_TRANS[ii]])
f.add_arc('n8x', 'last', [str(ii)], [kFRENCH_TRANS[ii]])
f.add_arc('n9x', 'last', [str(ii)], [kFRENCH_TRANS[10], kFRENCH_TRANS[ii]])
if ii == 8:
f.add_arc('start', 'nxx', [str(ii)], [kFRENCH_TRANS[ii],kFRENCH_TRANS[100]])
f.add_arc('0xx', 'n8x', [str(ii)], [kFRENCH_TRANS[4], kFRENCH_TRANS[20]])
f.add_arc('nxx', 'n8x', [str(ii)], [kFRENCH_TRANS[4], kFRENCH_TRANS[20]])
f.add_arc('n0x', 'last', [str(ii)], [kFRENCH_TRANS[ii]])
f.add_arc('n1x', 'last', [str(ii)], [kFRENCH_TRANS[10], kFRENCH_TRANS[ii]])
f.add_arc('nnx', 'last', [str(ii)], [kFRENCH_TRANS[ii]])
f.add_arc('n7x', 'last', [str(ii)], [kFRENCH_TRANS[10], kFRENCH_TRANS[ii]])
f.add_arc('n8x', 'last', [str(ii)], [kFRENCH_TRANS[ii]])
f.add_arc('n9x', 'last', [str(ii)], [kFRENCH_TRANS[10], kFRENCH_TRANS[ii]])
if ii == 9:
f.add_arc('start', 'nxx', [str(ii)], [kFRENCH_TRANS[ii],kFRENCH_TRANS[100]])
f.add_arc('0xx', 'n9x', [str(ii)], [kFRENCH_TRANS[4], kFRENCH_TRANS[20]])
f.add_arc('nxx', 'n9x', [str(ii)], [kFRENCH_TRANS[4], kFRENCH_TRANS[20]])
f.add_arc('n0x', 'last', [str(ii)], [kFRENCH_TRANS[ii]])
f.add_arc('n1x', 'last', [str(ii)], [kFRENCH_TRANS[10], kFRENCH_TRANS[ii]])
f.add_arc('nnx', 'last', [str(ii)], [kFRENCH_TRANS[ii]])
f.add_arc('n7x', 'last', [str(ii)], [kFRENCH_TRANS[10], kFRENCH_TRANS[ii]])
f.add_arc('n8x', 'last', [str(ii)], [kFRENCH_TRANS[ii]])
f.add_arc('n9x', 'last', [str(ii)], [kFRENCH_TRANS[10], kFRENCH_TRANS[ii]])
return f
from fst import FST
import string, sys
from fsmutils import composechars, trace
f1 = FST('soundex-generate')
f1.add_state('start')
f1.add_state('next')
f1.initial_state = 'start'
f1.set_final('next')
list_one = ['b', 'f', 'p', 'v']
list_two = ['c', 'g', 'j', 'k', 'q', 's', 'x', 'z']
list_three = ['d', 't']
list_four = ['l']
list_five = ['m', 'n']
list_six = ['r']
vowels = ['a', 'e', 'h', 'i', 'o', 'u', 'w', 'y']
