def __init__(self, data, count):
self.bigram_measures = nltk.collocations.BigramAssocMeasures()
with open(data, 'r') as dataDelim:
self.lst = [line.rstrip().split('-') for line in dataDelim]
with open(count, 'r') as noLines:
self.totallst = noLines.read().split('-')
self.oChain = pykov.Matrix()
def __init__(self,
messages: List[Tuple[Any, str]],
finite=False,
max_walk_length: int = 100,
enhance: bool = True):
"""Create a Chat object based on historical messages.
Args:
messages: List of tuples of sender and message. Sender may be of any type and may not be None. Messages should be in chronological order; otherwise results may be worse.
max_walk_length: Limit on number of words in a generated message
enhance: If set to true, repeated message generation for the same user may be faster at the cost of space. When dealing with a large amount of users on a system without much space this should be set to False
Examples:
>>> messages = [('Alice', 'Hey there!'),
('Bob', 'Hey, what's up?'),
('Charlie', 'Shuf is fantastic. I tried it on a 78 billion line text file.')]
>>> cc = chainchat.Chat(messages, max_walk_length = 1000, enhance = True)
"""
self.MAX_WALK_LENGTH = max_walk_length
self.messages = messages
self.Musers = pykov.Matrix()
self.enhance_speed = enhance
for mes, nxt in zip(messages[:-1], messages[1:]):
assert (mes is not None)
assert (nxt is not None)
if (mes[0], nxt[0]) not in self.Musers:
self.Musers[(mes[0], nxt[0])] = 0
self.Musers[(mes[0], nxt[0])] = self.Musers[(mes[0], nxt[0])] + 1
if finite: self.Musers[(messages[-1][0], None)] = 1
elif self.Musers.succ(key=messages[-1][0]).sum() == 0:
self.Musers[(messages[-1][0], messages[0][0])] = 1
self.Musers.stochastic()
def generate(self,
head: Any = None) -> Generator[Tuple[Any, str], None, None]:
"""Return a generator that yields a new message on every next call.
Args:
head: User to start the interaction with. If no user supplied, the sender of the earliest message is used. Note, this user will not send a message but will be the user the first generated message responds to.
Examples:
# Generate 2 messages
>>> gen = cc.generate('Alice')
>>> for _ in range(2):
tpl = next(gen)
print(": ".join(tpl))
Bob: Hey, what's up?
Charlie: Shuf is fantastic. I tried it on a 78 billion line text file.
# Generates an infinite amount of messages
>>> for tpl in cc.generate('Alice')():
print(": ".join(tpl))
Bob: Hey, what's up?
Charlie: Shuf is fantastic. I tried it on a 78 billion line text file.
...
"""
user = pykov.Vector({self.messages[0][0] if head is None else head:
1}) # starting point for markov chain
while True:
choice = (
user *
self.Musers).choose() # choose next user to write a message
if choice is None: return
user = pykov.Vector(
{choice: 1}) # create pykov vector for future user calculation
if choice in self.word_matrices:
Mword = self.word_matrices[choice]
else: # create word matrix
Mword = pykov.Matrix({(1, 1):
1}) # create matrix for word chain
for mes in [m for m in self.messages
if m[0] == choice]: # add every word to matrix
splt = mes[1].split(' ')
if (0, splt[0]) not in Mword: Mword[(0, splt[0])] = 0
Mword[(0, splt[0])] += 1
for word, nxt in zip(splt[:-1], splt[1:]):
if (word, nxt) not in Mword: Mword[(word, nxt)] = 0
Mword[(word, nxt)] += 1
if (splt[-1], 1) not in Mword: Mword[(splt[-1], 1)] = 0
Mword[(splt[-1], 1)] += 1
Mword.stochastic()
if not self.enhance_speed: self.word_matrices[choice] = Mword
C = pykov.Chain(Mword)
yield (choice, " ".join(C.walk(self.MAX_WALK_LENGTH, 0, 1)[1:-1]))
def get_symmetric_pykov_matrix(self):
states = self.get_state_names()
edge_dict = collections.OrderedDict()
edges = self.language_construct.get_edgelist()
vertex_degrees = self.language_construct.vs.degree()
transition_probabilities = [
1 / degree for degree in vertex_degrees if degree > 0
]
for edge in edges:
state1 = states[edge[0]]
state2 = states[edge[1]]
edge_dict[(state1, state2)] = transition_probabilities[edge[0]]
edge_dict[(state2, state1)] = transition_probabilities[edge[1]]
return pykov.Matrix(edge_dict)
def buildMarkovChain(phrasesFile):
phrases = construct_phrases_list(phrasesFile)
print(phrases)
wordCount = buildWordCount(phrases)
connectionsCount = buildConnectionsCount(phrases)
vector = pykov.Matrix()
for connection in connectionsCount:
initialState = connection[0]
#print("(" + str(connectionsCount[connection]) + ", " + str(wordCount[initialState]) + ")")
probability = (float(connectionsCount[connection]) /
float(wordCount[initialState]))
#print(probability)
vector[connection] = probability
return pykov.Chain(vector)
def markov_chain(states):
T = pykov.Matrix()
markov_states = []
for i in xrange(0, len(states) - 1):
markov_states.append((
states[i],
states[i + 1],
))
for i in xrange(len(markov_states)):
total = len([j for j in markov_states if j[0] == markov_states[i][0]]),
amount = markov_states.count(markov_states[i])
markov_states[i] += (amount / float(total[0]), )
for i in markov_states:
if (i[0], i[1]) not in T:
T[(i[0], i[1])] = i[2]
return pykov.Chain(T)
def generate():
import pykov
phrasesData = ''
with open('sayings.txt', "rt") as f:
phrasesData = f.read()
phrasesLines = phrasesData.split("\n")
phrases = phrasesLines
wordCount = dict()
for phrase in phrases:
words = phrase.split(" ")
for index in range(len(words)):
word = words[index]
if word not in wordCount:
wordCount[word] = 0
wordCount[word] = wordCount.get(word) + 1
connectionsCount = dict()
for phrase in phrases:
words = phrase.split(" ")
for index in range(len(words) - 1):
currState = (words[index], words[index + 1])
if currState not in connectionsCount:
connectionsCount[currState] = 0
connectionsCount[currState] = connectionsCount.get(currState) + 1
vector = pykov.Matrix()
for connection in connectionsCount:
initialState = connection[0]
probability = (float(connectionsCount[connection]) /
float(wordCount[initialState]))
vector[connection] = probability
mc = pykov.Chain(vector)
newPhrases = mc.walk(100, '+', '+\r')
newPhrases = newPhrases[1:len(newPhrases) - 1]
newPhrases = " ".join(newPhrases)
newPhrase = newPhrases.split('+')[0].strip()
return newPhrase.replace("í", "'")
import pykov
d = pykov.Matrix({
('coal', 'WW'): 0.3483,
('coal', 'COD'): 0.2034,
('coal', 'NHN'): 0.2658,
('oil', 'WW'): 0.1399,
('oil', 'COD'): 0.1820,
('oil', 'NHN'): 0.2666,
('chemistry', 'WW'): 0.1397,
('chemistry', 'COD'): 0.1826,
('chemistry', 'NHN'): 0.1916,
('smelt', 'WW'): 0.0759,
('smelt', 'COD'): 0.0515,
('smelt', 'NHN'): 0.0417,
('farm', 'WW'): 0.0458,
('farm', 'COD'): 0.1116,
('farm', 'NHN'): 0.0355,
})
T = pykov.Chain(d)
import matplotlib.pyplot as plt
import numpy as np
# create some data to use for the plot
dt = 0.001
t = np.arange(0.0, 10.0, dt)
r = np.exp(-t[:1000] / 0.05) # impulse response
x = np.random.randn(len(t))
s = np.convolve(x, r)[:len(x)] * dt # colored noise
import pykov
import math
s = [0.99, 0.01, 0.00, 0.00, 0.00, 0.00]
i = [0.00, 0.80, 0.04, 0.01, 0.00, 0.15]
h = [0.00, 0.15, 0.80, 0.05, 0.00, 0.00]
u = [0.00, 0.00, 0.05, 0.80, 0.15, 0.00]
o = [0.00, 0.00, 0.00, 0.00, 1.00, 0.00]
r = [0.00, 0.00, 0.00, 0.00, 0.00, 1.00]
a = np.array([s, i, h, u, o, r])
str_states = ['S', 'I', 'H', 'U', 'O', 'R']
states = dict(zip(list(range(1, len(a) + 1)), str_states))
dic = {(states[i + 1], states[j + 1]): a[i][j]
for i in range(len(a)) for j in range(len(a[i]))}
M = pykov.Matrix(dic)
print(M, end='\n\n')
# Todos os estados
print(M.states(), end='\n\n')
# Antecessores de cada estado
print(M.pred(), end='\n\n')
# Sucessores de cada estado
print(M.succ(), end='\n\n')
C = pykov.Chain(dic)
state = pykov.Vector(S=1)
# Distribuição de prob após N dias, começando no estado state
n_inicio = 282 # dias, entre o dia do primeiro infetado em Portugal e o dia da submissão do projeto
dia274 = C.pow(state, n_inicio)
n_fim = 365 - (31 + 28 + 1)
for st in json_data['statesArray']:
states_dict[st['id']] = st['text'] # 'text' is the name of the state
try:
states_mirdef[st['text']] = st['mir'][0]
except:
states_mirdef[st['text']] = {
"sfx.duration": "* TO 3",
"sfx.inharmonicity.mean": "0.1"
} #default value
try:
states_dur[st['text']] = st['duration'] #default value
except:
states_dur[st['text']] = 1. # default duration
sd = states_dict
T = pykov.Matrix()
for st in json_data['linkArray']:
# print( float(st['text']) )
T[sd[st['from']], sd[st['to']]] = float(st['text'])
try:
T.stochastic() #check
except Exception, e:
print(e)
exit(1)
#########################
#FIXME: Time
# duration = 1 #FIXME: hardcoded (default duration)
# time_bt_states = 1 # (delay within states...)
#########################
def main():
if len(sys.argv) < 2:
print(("\nBad amount of input arguments\n\t", Usage, "\n"))
sys.exit(1)
# JSON config file
config = ""
try:
config = json.load(open(".apicultor_config.json", 'r'))
except Exception as e:
print(e)
print("No json config file or error.")
# sys.exit(2)
api_key = config["Freesound.org"][0]["API_KEY"]
#JSON composition file
json_data = ""
try:
json_comp_file = sys.argv[1]
# with open(json_file,'r') as file:
# json_data = json.load( file )
json_data = json.load(open(json_comp_file, 'r'))
except Exception as e:
print(e)
print("JSON composition file error.")
sys.exit(2)
print("Starting MIR state machine")
states_dict = dict() # id to name conversion
states_dur = dict() #states duration
start_state = json_data['statesArray'][0][
'text'] #TODO: add as property (start: True)
for st in json_data['statesArray']:
states_dict[st['id']] = st['text']
try:
states_dur[st['text']] = st['duration']
except:
states_dur[st['text']] = 1. # default duration
sd = states_dict
T = pykov.Matrix()
for st in json_data['linkArray']:
# print( float(st['text']) )
T[sd[st['from']], sd[st['to']]] = float(st['text'])
try:
T.stochastic() #check
except Exception as e:
print(e)
exit(1)
#########################
#FIXME: Time
# duration = 1 #FIXME: hardcoded (default duration)
# time_bt_states = 1 # (delay within states...)
#########################
#########################
# Init conditions
#state = 'idle' #start state
# state = "A" #start state
state = start_state
previous_state = "H"
#Fixed amount or infinite with while(1 ) ()
# events = 10 # or loop with while(1)
# for i in range(events):
while (1):
print(("State: %s" %
state)) # TODO: call the right method for the state here
#(optional) change sound in the same state or not (add as json config file)
if state != previous_state:
#retrieve new sound
call = '/list/samples' #gets only wav files because SuperCollider
response = urllib.request.urlopen(URL_BASE + call).read()
audioFiles = list()
for file in response.split('\n'):
if len(file) > 0: #avoid null paths
audioFiles.append(file)
# print file
file_chosen = audioFiles[random.randint(0, len(audioFiles) - 1)]
print(file_chosen)
# Hardcoded sound for each MIR state
# file_chosen = snd_dict[state]
pyo_synth(file_chosen, dry_val)
# granular_synth(file_chosen)
# external_synth(file_chosen)
time_bt_states = states_dur[state]
# time_between_notes = random.uniform(0.,2.) #in seconds
#time.sleep(time_between_notes)
# MIDI
# notes = Notein(poly=10, scale=1, mul=.5)
# p = Port(notes['velocity'], .001, .5)
# # Add inputs to the mixer
# mm.addInput(voice=new_voice, input=sfplay)
#mm.addInput(voice=new_voice, input=pvs)
# Delay within states
time.sleep(time_bt_states)
#next state
previous_state = state
state = T.succ(state).choose() #new state