File size: 5,080 Bytes
a6326c7 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 |
from utils import (
write_jsonl_file,
parse,
)
import os
topics = {
1: "Ordinary Life",
2: "School Life",
3: "Culture & Education",
4: "Attitude & Emotion",
5: "Relationship",
6: "Tourism",
7: "Health",
8: "Work",
9: "Politics",
10: "Finance",
}
emotions = {
0: "neutral",
1: "anger",
2: "disgust",
3: "fear",
4: "happiness",
5: "sadness",
6: "surprise",
}
acts = {1: "inform", 2: "question", 3: "directive", 4: "commissive"}
def load_topics(args):
text_file = os.path.join(args.input_dir, "dialogues_text.txt")
topic_file = os.path.join(args.input_dir, "dialogues_topic.txt")
text2topic = dict()
with open(text_file, "r", encoding="utf-8") as text_reader, open(
topic_file, "r", encoding="utf-8"
) as topic_reader:
for line in text_reader:
text = line.strip()
topic = topics[int(topic_reader.readline().strip())]
# if text in text2topic and text not in [
# "Can I help you ? __eou__ I hope so . I'm looking for some material for a paper I'm writing , and I'm not quite sure where to look . __eou__ I'll certainly try to help you . What topic is your paper on ? __eou__ My paper is on the influence of television on children . __eou__ There are several possible sources you might use for that topic . I suggest you use the computer and the computer will give you a list of every scientific journal that talks about children and television . __eou__ Thank you for you help . __eou__"
# "Hey , Ann . You don't have a pen , do you ? __eou__ Sure , here you go . __eou__ Thanks . I don't suppose you have some paper , too . __eou__ Of course . There you are . __eou__ Thanks so much . I owe you one ."
# ]:
# print(text, topic, text2topic[text])
# assert text2topic[text] == topic
text2topic[text] = topic
return text2topic
def preprocess(args, split, text2topic):
input_dir = os.path.join(args.input_dir, split)
text_file = os.path.join(input_dir, f"dialogues_{split}.txt")
act_file = os.path.join(input_dir, f"dialogues_act_{split}.txt")
emotion_file = os.path.join(input_dir, f"dialogues_emotion_{split}.txt")
if split == "validation":
split = "dev"
outfile = os.path.join(args.output_dir, f"{split}.jsonl")
processed_data = []
with open(text_file, "r", encoding="utf-8") as text_reader, open(
act_file, "r", encoding="utf-8"
) as act_reader, open(emotion_file, "r", encoding="utf-8") as emotion_reader:
for line in text_reader:
text = line.strip()
if text in text2topic:
topic = text2topic[text]
else:
_text = "Sam , can we stop at this bicycle shop ? __eou__ Do you want to buy a new bicycle ? __eou__ Yes , and they have a sale on now . __eou__ What happened to your old one ? __eou__ I left it at my parent's house , but I need one here as well . __eou__ I've been using Jim's old bike but he needs it back . __eou__ Let's go then . __eou__ Look at this mountain bike . It is only £ 330 . Do you like it ? __eou__ I prefer something like this one - a touring bike , but it is more expensive . __eou__ How much is it ? __eou__ The price on the tag says £ 565 but maybe you can get a discount . __eou__ OK , let's go and ask . __eou__"
topic = text2topic[_text]
utterances = text.split("__eou__")
assert not utterances[-1]
utterances = utterances[:-1]
_acts = list(
map(lambda x: acts[int(x)], act_reader.readline().strip().split())
)
_emotions = list(
map(
lambda x: emotions[int(x)],
emotion_reader.readline().strip().split(),
)
)
dialogue = {
"turn": "multi",
"locale": "en",
"domain": [topic],
"dialog": [],
"knowledge": {"type": "list", "value": sorted(emotions.values())},
}
assert len(utterances) == len(_acts) and len(utterances) == len(
_emotions
), f"{utterances}\n{_acts}\n{_emotions}"
roles = ["ROLE1", "ROLE2"]
for idx, utterance in enumerate(utterances):
assert utterance
dialogue["dialog"].append(
{
"roles": [roles[idx % 2]],
"utterance": utterance,
"active_intents": [_acts[idx]],
"emotions": [{"emotion": _emotions[idx]}],
}
)
processed_data.append(dialogue)
write_jsonl_file(processed_data, outfile)
if __name__ == "__main__":
args = parse()
text2topic = load_topics(args)
preprocess(args, "train", text2topic)
preprocess(args, "validation", text2topic)
preprocess(args, "test", text2topic)
|