ProbabilisticPython/eve.py at master · NancyFulda/ProbabilisticPython · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
import numpy as np
from parser import Parser
from inferencedriver import InferenceDriver

# ----------------------------------
# Globals

EMOTIONS = { 0: "none", 1: "anger", 2: "disgust", 3: "fear",
             4: "happiness", 5: "sadness", 6: "surprise"}
emotions = ["none", "anger", "disgust", "fear",
            "happiness", "sadness", "surprise"]
TOPICS = {1: "ordinary_life", 2: "school_life", 3: "culture_and_education",
          4: "attitude_and_emotion", 5: "relationship", 6: "tourism",
          7: "health", 8: "work", 9: "politics", 10: "finance"}
ACT = {1: "inform", 2: "question", 3: "directive", 4: "commissive"}

EVE_EMOTIONAL_BIAS = np.array([.15, .05, .05, .05, .5, .05, .15])
PARSER = Parser(num_of_observations=100)

# ----------------------------------
# Define a model

def sampled_user_text(pp, loop_iter, emotion):
    return pp.choice(elements=PARSER.emotion_turns[emotion],
            name="sample_user_text", loop_iter=loop_iter)

def generated_eve_text(pp, loop_iter, emotion, text):
    emotive = ""
    answer = pp.choice(elements=PARSER.emotion_turns[emotion], name="sample_eve_text", loop_iter=loop_iter)
    offer = ""
    response =  emotive + " " + answer + " " + offer
    return response.strip()

def emotional_response(pp, name, loop_iter, text, emotional_bias):
    emotion = PARSER.text_emotions[text]
    temp_bias = emotional_bias.copy()
    temp_bias[emotion] += 1.0
    temp_bias /= np.sum(temp_bias)
    return pp.choice(elements=emotions, p=emotional_bias, name=name, loop_iter=loop_iter)


def model(pp):

    # generate user emotional bias
    user_none = pp.random(name="user_none")
    user_anger = pp.random(name="user_anger")
    user_disgust = pp.random(name="user_disgust")
    user_fear = pp.random(name="user_fear")
    user_happiness = pp.random(name="user_happiness")
    user_sadness = pp.random(name="user_sadness")
    user_surprise = pp.random(name="user_surprise")
    user_emotional_bias = [user_none, user_anger, user_disgust, user_fear,
                                user_happiness, user_sadness, user_surprise]
    user_emotional_bias /= np.sum(user_emotional_bias)

    for i in range(len(PARSER.observations)):
        user_emo_prev = pp.choice(elements=emotions, p=user_emotional_bias, name="user_emo_prev", loop_iter=i)
        print("User emotion prev:", user_emo_prev)

        user_text = sampled_user_text(pp, i, user_emo_prev) # Condition here
        # print("User text:", user_text)

        eve_emo = emotional_response(pp, "eve_emo", i, user_text, EVE_EMOTIONAL_BIAS)
        # print("Eve emotion:", eve_emo)

        eve_text = generated_eve_text(pp, i, eve_emo, user_text)
        # print("Eve text:", eve_text)

        user_emo = emotional_response(pp, "user_emo", i, eve_text, user_emotional_bias) # Condition here
        # print("User emotion:", user_emo)

        user_emo_prev = user_emo


# ----------------------------------
# Run inference

#create the inference object
driver = InferenceDriver(model)

# observations = [["happiness",
# "The lower branches on that tree are hanging very low . Would you like me to cut them off for you ?",
# ]] # only one loop

# condition
for i in range(len(PARSER.observations)):
    driver.condition(label="user_emo_prev-{}".format(i), value=PARSER.observations[i][0])
    driver.condition(label="sample_user_text-{}".format(i), value=PARSER.observations[i][1])
    driver.condition(label="eve_emo-{}".format(i), value=PARSER.observations[i][2])
    driver.condition(label="sample_eve_text-{}".format(i), value=PARSER.observations[i][3])
    driver.condition(label="user_emo-{}".format(i), value=PARSER.observations[i][4])

# init and establish priors
driver.init_model()
driver.prior(label="user_none-0", value=.15)
driver.prior(label="user_anger-0", value=.05)
driver.prior(label="user_disgust-0", value=.05)
driver.prior(label="user_fear-0", value=.05)
driver.prior(label="user_happiness-0", value=.5)
driver.prior(label="user_sadness-0", value=.05)
driver.prior(label="user_surprise-0", value=.15)

# burn in
driver.burn_in(steps=100)

# run inference
driver.run_inference(interval=5, samples=1)

# print estimated preferences
# print(driver.return_traces())
# print("Estimated:", driver.return_string_values(keys=["user_emo", "eve_emo",
        # "sample_eve_text", "sample_user_text"]))

found_user_bias = driver.return_values(keys=["user_none", "user_anger",
                "user_disgust", "user_fear", "user_happiness", "user_sadness", "user_surprise"])
found_user_bias = [found_user_bias["user_none-0"], found_user_bias["user_anger-0"],
                   found_user_bias["user_disgust-0"],found_user_bias["user_fear-0"],
                   found_user_bias["user_happiness-0"], found_user_bias["user_sadness-0"],
                   found_user_bias["user_surprise-0"]]
found_user_bias /= np.sum(found_user_bias)

print("Estimated:", found_user_bias)

# graph your likelihood
driver.graph_ll()