def Q_implicature_simulation_datapoint(specific_cost, dcost=1.0, alpha=2.0):
# Values to obtain:
is_max = False
listener_val = None
speaker_val = None
# Set-up:
lexica = Lexica(baselexicon=BASELEXICON, costs={GENERAL_MSG: 0.0, SPECIFIC_MSG: specific_cost}, join_closure=True, nullsem=True, nullcost=5.0, disjunction_cost=dcost)
ref_probs = np.repeat(1.0/len(lexica.states), len(lexica.states))
lexprior = np.repeat(1.0/len(lexica.lexica2matrices()), len(lexica.lexica2matrices()))
# Run the model:
mod = Pragmod(lexica=lexica.lexica2matrices(), messages=lexica.messages, states=lexica.states, costs=lexica.cost_vector(), lexprior=lexprior, prior=ref_probs, alpha=alpha)
langs = mod.run_expertise_model(n=3, display=False, digits=2)
# Get the values we need:
speaker = mod.speaker_lexical_marginalization(langs[-2])
listener = mod.listener_lexical_marginalization(langs[-3])
general_msg_index = lexica.messages.index(GENERAL_MSG)
general_only_state = lexica.states.index(GENERAL_ONLY_REF)
disj_state_index = lexica.states.index(DISJ_REF)
disj_msg_index = lexica.messages.index(DISJ_MSG)
speaker_val = speaker[disj_state_index, disj_msg_index]
listener_val = listener[general_msg_index, general_only_state]
# Determine whether max, with a bit of rounding to avoid spurious mismatch diagnosis:
maxspkval = np.max(speaker[disj_state_index])
is_max = np.round(speaker_val, 10) == np.round(maxspkval, 10)
# Return values:
return (listener_val, speaker_val, is_max)
def I_implicature_simulation_datapoint(common_ref_prob, dcost=1.0, alpha=2.0):
# Values to obtain:
is_max = False
listener_val = None
speaker_val = None
# Set-up:
lexica = Lexica(baselexicon=BASELEXICON, costs=LEXICAL_COSTS, join_closure=True, nullsem=True, nullcost=5.0, disjunction_cost=dcost)
ref_probs = np.array([common_ref_prob, (1.0-common_ref_prob)/2.0, (1.0-common_ref_prob)/2.0])
lexprior = np.repeat(1.0/len(lexica.lexica2matrices()), len(lexica.lexica2matrices()))
# Run the model:
mod = Pragmod(lexica=lexica.lexica2matrices(), messages=lexica.messages, states=lexica.states, costs=lexica.cost_vector(), lexprior=lexprior, prior=ref_probs, alpha=alpha)
langs = mod.run_expertise_model(n=3, display=False, digits=2)
# Get the values we need:
speaker = mod.speaker_lexical_marginalization(langs[-2])
listener = mod.listener_lexical_marginalization(langs[-3])
superkind_term_index = mod.messages.index(SUPERKIND_MSG)
common_state_index = mod.states.index(COMMON_REF)
disj_term_index = mod.messages.index(DISJ_MSG)
disj_state_index = mod.states.index(DISJ_REF)
# Fill in listener_val and speaker_val:
listener_val = listener[superkind_term_index, common_state_index]
speaker_val = speaker[disj_state_index, disj_term_index]
# Determine whether max, with a bit of rounding to avoid spurious mismatch diagnosis:
maxspkval = np.max(speaker[disj_state_index])
is_max = np.round(speaker_val, 10) == np.round(maxspkval, 10)
# Return values:
return (listener_val, speaker_val, is_max)
def disjunction(n=3):
"""Settings for Bergen et al.'s figure 6. Seems to reproduce the effects they report."""
# bl = {'p': [r'$w_1$', r'$w_2$'], 'q':[r'$w_2$', r'$w_3$']}
bl = {'p': ['1', '2'], 'q':['1', '3'], 'p & q': ['1'], 'p v q': ['1', '2', '3']}
lexica = Lexica(baselexicon=bl,
atomic_states=['1', '2', '3'],
disjunction_cost=1.0,
conjunction_cost=0.0,
null_cost=5.0,
join_closure=True,
meet_closure=False,
block_trivial_messages=True,
block_ineffability=False)
lexica.display()
for key, val in lexica.lexica[8].items():
print key, len(val), val
mod = Pragmod(lexica=lexica.lexica2matrices(),
messages=lexica.messages,
meanings=lexica.states,
costs=lexica.cost_vector(),
prior=np.repeat(1.0/len(lexica.states), len(lexica.states)),
lexprior=np.repeat(1.0/len(lexica), len(lexica)),
temperature=1.0,
alpha=1.0)
#mod.plot_expertise_listener(output_filename='../paper/fig/scalardisj-expertise-listener-marginalized.pdf', n=n)
#mod.plot_expertise_speaker(output_filename='../paper/fig/scalardisj-expertise-speaker.pdf', n=n)
#mod.plot_expertise_speaker(output_filename='../paper/fig/scalardisj-expertise-speaker-lexsum.pdf', n=n, lexsum=True)
mod.plot_expertise_listener(n=n)
def build(self):
lex = Lexica(
baselexicon=self.baselexicon,
join_closure=True,
disjunction_cost=self.disjunction_cost,
nullsem=True,
null_cost=self.null_cost,
costs=copy.copy(self.lexical_costs),
unknown_word=self.unknown_word,
)
self.lexica = lex.lexica2matrices()
self.states = lex.states
self.messages = lex.messages
if self.prior == None:
self.prior = np.repeat(1.0 / len(self.states), len(self.states))
if self.lexprior == None:
self.lexprior = np.repeat(1.0 / len(self.lexica), len(self.lexica))
self.model = Pragmod(
lexica=self.lexica,
messages=self.messages,
meanings=self.states,
costs=lex.cost_vector(),
prior=self.prior,
lexprior=self.lexprior,
temperature=self.temperature,
alpha=self.alpha,
beta=self.beta,
)
self.langs = self.model.run_expertise_model(n=self.n, display=True)
def scalars(n=3):
"""Scalar example without and with disjunction; compare with Bergen et al.'s figure 5 and figure 9"""
baselexicon = {'some': [r'$w_{\exists\neg\forall}$', r'$w_{\forall}$'], 'all': [r'$w_{\forall}$']}
basic_lexica = Lexica(baselexicon=baselexicon)
lexica = Lexica(baselexicon=baselexicon, join_closure=True, disjunction_cost=0.1)
lexica.display()
mod = Pragmod(lexica=lexica.lexica2matrices(),
messages=lexica.messages,
meanings=lexica.states,
costs=lexica.cost_vector(),
prior=np.repeat(1.0/len(lexica.states), len(lexica.states)),
lexprior=np.repeat(1.0/len(lexica), len(lexica)),
temperature=1.0)
#mod.run_expertise_model(n=n, display=True, digits=4)
#mod.plot_expertise_listener(output_filename='../paper/fig/scalar-expertise-listener-marginalized.pdf', n=n)
mod.plot_expertise_speaker(output_filename='../paper/fig/scalar-expertise-speaker.pdf', n=n)
def manner(n=3):
"""Settings for Bergen et al.'s figure 6. Seems to reproduce the effects they report."""
lexica = Lexica(baselexicon={'SHORT': [r'$w_{RARE}$', r'$w_{FREQ}$'], r'long': [r'$w_{RARE}$', r'$w_{FREQ}$']},
costs={'SHORT':1.0, r'long':2.0},
null_cost=5.0,
join_closure=False,
disjunction_cost=0.1)
lexica.display()
mod = Pragmod(lexica=lexica.lexica2matrices(),
messages=lexica.messages,
meanings=lexica.states,
costs=lexica.cost_vector(),
prior=np.array([2.0/3.0, 1.0/3.0]),
lexprior=np.repeat(1.0/len(lexica), len(lexica)),
temperature=1.0,
alpha=3.0)
mod.plot_expertise_listener(output_filename='../paper/fig/manner-expertise-listener-marginalized.pdf', n=n)
mod.plot_expertise_speaker(output_filename='../paper/fig/manner-expertise-speaker.pdf', n=n)
mod.plot_expertise_speaker(output_filename='../paper/fig/manner-expertise-speaker-lexsum.pdf', n=n, lexsum=True)
