-
Notifications
You must be signed in to change notification settings - Fork 1
/
loco.py
571 lines (438 loc) · 19.7 KB
/
loco.py
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
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
import os
import re
import json
from typing import List, Dict
import openai
from google.cloud import translate
# API Keys
os.environ["GOOGLE_APPLICATION_CREDENTIALS"] = "assets/b7f9.json"
openai.api_key = "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"
google_trans_id = "delta-coast-382412"
def place_en_to_ko(location: str) -> str:
place_dict ={'강남/역삼/선릉': 'Gangnam / Yeoksam / Seolleung',
'강남구청': 'Gangnam-gu Office',
'개포/일원/수서': 'Gaepo/Irwon/Suseo',
'건대/군자/구의': 'Kondae/Gunja/Guui',
'금호/옥수': 'Kumho/Oksu',
'명동/을지로/충무로': 'Myeongdong/Euljiro/Chungmuro',
'방이': 'Bangi-dong',
'북촌/삼청': 'Bukchon/Samcheong',
'삼성/대치': 'Samsung/Daechi',
'상수/합정/망원': 'Sangsu/Hapjeong/Mangwon',
'서울역/회현': 'Seoul Station/Hoehyeon',
'서초/방배': 'Seocho/Bangbae',
'서촌': 'Seochon',
'선릉/삼성': 'Seolleung/Samsung',
'성수/뚝섬': 'Seongsu/Ttukseom',
'신사/논현': 'Sinsa/Nonhyeon',
'신촌/홍대/서교': 'Sinchon / Hongdae / Seogyo',
'압구정/청담': 'Apgujeong/Cheongdam',
'양재/도곡': 'Yangjae/Dogok',
'연남': 'Yeonnam',
'영동포/여의도': 'Yeongdongpo/Yeouido',
'용산/삼각지': 'Yongsan/Samgakji',
'이태원/한남': 'Itaewon/Hannam',
'잠실/송파': 'Jamsil/Songpa',
'종로/광화문': 'Jongno/Gwanghwamun',
'분당': 'Bundang-gu',
'수원/광교': 'Suwon/Gwanggyo',
'판교': 'Pangyo'}
gyeonggido_list = ['Bundang-gu', 'Suwon/Gwanggyo', 'Pangyo']
# gyeonggido-rization
if place_dict[location] in gyeonggido_list:
return place_dict[location] + ", GyeongGi-Do"
else:
return place_dict[location] + ", Seoul"
def translator(text: str, task: str, project_id=google_trans_id) -> str:
client = translate.TranslationServiceClient()
location = "us-central1"
# 아까 지정한 용어집 이름과 같이 맞춰줘야 함
glossary = client.glossary_path(project_id, location, 'loco_translator_glossary')
glossary_config = translate.TranslateTextGlossaryConfig(glossary=glossary)
parent = f"projects/{project_id}/locations/{location}"
# Translate text from English to Korean
if task == 'ko-en':
response = client.translate_text(
request={
"parent": parent,
"contents": [text],
"mime_type": "text/plain", # mime types: text/plain, text/html
"source_language_code": "ko",
"target_language_code": "en-US",
"glossary_config": glossary_config,
}
)
return response.glossary_translations[0].translated_text
elif task == 'en-ko':
response = client.translate_text(
request={
"parent": parent,
"contents": [text],
"mime_type": "text/plain", # mime types: text/plain, text/html
"source_language_code": "en-US",
"target_language_code": "ko",
"glossary_config": glossary_config,
}
)
return response.glossary_translations[0].translated_text
else:
pass
def translator_chatgpt(text: str, task: str) -> str:
TRANSLATION_TASK_PROMPT = ""
if task == 'ko-en':
TRANSLATION_TASK_PROMPT = f"Translate {text} into English."
elif task == 'en-ko':
TRANSLATION_TASK_PROMPT = f"Translate {text} into Korean."
else:
pass
LOCO_TRANSLATION_PROMPT = f"""{TRANSLATION_TASK_PROMPT}
You should not translate the following words: ACTIVITY, "activity_name", "start_time", "end_time", "description", "budget".
Do not generate both before and after translation results. Begin!:
"""
response = openai.ChatCompletion.create(
model="gpt-3.5-turbo",
messages=[{"role": "user",
"content": f"{LOCO_TRANSLATION_PROMPT}."}],
temperature=1.1,
max_tokens=2048,
top_p=0.9,
frequency_penalty=0.0,
presence_penalty=0.0
)
result = response.choices[0].message.content
return result
def init():
LOCO_PREFIX_RPOMPT = """We are making LOCO, a dating application.
LOCO aims to recommend the best dating route by receiving the Meeting and Parting time, Cost, and Approximate location as input.
You must tell the user in detail where and what to do depending on the time of day.
You should not provide virtual location information, but must present location information that exists in reality.
You must not recommend physically impossible routes. For example, Human can not travel from Anam Station to Seoul Nat'l Univ. Station in 1 minutes.
You must not make recommendations beyond a given budget. For example, you shouldn't be recommending $10 spaghetti to your user if user's budget is only $5.
You must create a date path strictly according to our instructions.
Do you understand this task is?:"""
while True:
init_loco = openai.ChatCompletion.create(
model="gpt-3.5-turbo",
messages=[{"role": "user",
"content": f"{LOCO_PREFIX_RPOMPT}"}],
temperature=1.1,
max_tokens=2048,
top_p=0.9,
frequency_penalty=0.0,
presence_penalty=0.0
)
print("---------------------------------------------------")
print("\"Human\": Do you understand what this task is?")
print("\"AI\":", init_loco.choices[0].message.content)
if "yes" in init_loco.choices[0].message.content.lower():
print("\"Human\": Great, Let's begin!")
print("---------------------------------------------------")
break
else:
continue
def postprocessing(result: str) -> str:
"""
[Exception 1] OUTPUT FORMAT:
ACTIVITY
1. ~
[Exception 1-1] "activity_name" not in result
ACTIVITY
1. "Lotte World Tower Observation Deck"
[Exception 1-2] "ACTIVITY" not in result
1. Breakfast at Bongchu Jjimdak (봉추찜닭) in Jamsil
- "activity_name": "Bongchu Jjimdak (봉추찜닭)",
[Exception 1-3] Variation 1-3
ACTIVITY
1. "activity_name": "Seoul Sky Observatory",
[Exception 1-4] Variation 1-4
ACTIVITY
1. "Breakfast at Paul Bassett"
[Exception 1-5] There is one ACTIVITY
ACTIVITY
"activity_name": "Seonjeongneung Park",
"start_time": "2023-04-09T11:00:00",
"end_time": "2023-04-09T13:00:00",
"description": "A historic park featuring the tombs of two Joseon Dynasty kings and their queens.",
"budget": "0"
"activity_name": "Starfield COEX Mall",
[Exception 2] UNINTENDED OUTPUT:
"budget": "39000"
Total budget: 101000 Won <- THIS
[Exception 3] UNINTENDED OUTPUT between activities
"budget": "0"
To end the day, we can grab some coffee and dessert at Starbucks Yangjae Station Branch. It's a cozy place to chat and relax after a long day. <- THIS
"activity_name": "Starbucks Yangjae Station Branch",
"""
# remove multiple spaces
result = re.sub(' +', '', result)
result = result.replace("- ", "")
pattern = """ACTIVITY
\"activity_name\""""
result = result.replace(pattern, "ACTIVITY\n\"activity_name\"")
# exception: 1-5
if result.count('ACTIVITY') == 1:
if "1. " in result:
pattern = re.compile(r'\d\. .*')
findall_num = pattern.findall(result)
for f in findall_num:
result = re.sub(f, "", result)
result = result.replace("\n\"activity_name\"", "ACTIVITY\n\"activity_name\"")
else:
result = result.replace("\n\"activity_name\"", "ACTIVITY\n\"activity_name\"")
# exception: 3
exception_pattern_3 = re.compile(r'\n\n.+\n')
findall_between = exception_pattern_3.findall(result)
if findall_between:
for f in findall_between:
if "\"activity_name\"" in f or 'ACTIVITY' in f:
continue
result = result.replace(f, "")
# exception: 1-4
activity_findall = re.compile(r'\d\. .*').findall(result)
for a in activity_findall:
if "activity_name" not in a:
result = re.sub(a, '\nACTIVITY', result)
# exception: 1-3
if "ACTIVITY\n1. \"activity_name\"" in result:
result = re.sub(r'\d\. ', 'ACTIVITY\n', result)
exception_pattern_1 = re.compile(r'\d\. ' + '\"*[a-zA-Z| *|:|é|\"]*\"*')
# exception: 1-1
if "activity_name" not in result:
activity_findall = exception_pattern_1.findall(result)
for a in activity_findall:
result = re.sub(exception_pattern_1, 'ACTIVITY\n\"activity_name\": ' + a, result)
result = re.sub(r'\d\. ', "", result)
# exception: 1-2
if "ACTIVITY" not in result:
activity_findall = exception_pattern_1.findall(result)
for _ in activity_findall:
result = re.sub(exception_pattern_1, 'ACTIVITY\n', result)
result = re.sub(r'\d\. ', "", result)
result = re.sub(r'\([^)]*\) *[a-zA-Z| *|가-힣]*', "", result)
# exception: 1
if exception_pattern_1.match(result):
result = re.sub('\d\. ' + '[a-zA-Z | *]*', 'ACTIVITY', result)
if "ACTIVITY\n\nACTIVITY" in result:
result = result.replace("ACTIVITY\n\nACTIVITY", "ACTIVITY")
if "\n\n\nACTIVITY" in result:
result = result.replace("\n\n\nACTIVITY", "\n\nACTIVITY")
if "ACTIVITY\n\nACTIVITY" in result:
result = result.replace("ACTIVITY\n\nACTIVITY", "ACTIVITY\n")
# exception: 2
exception_pattern_2 = re.compile(r"\"budget\": \"[a-zA-Z0-9]*\"")
findall_budget = exception_pattern_2.findall(result)
if findall_budget: # not empty
chop_idx_from = result.rfind(findall_budget[-1])
chop_idx_to = len(findall_budget[-1])
result = result[:chop_idx_from + chop_idx_to]
# e.g., ACTIVITY 1 -> ACTIVITY
result = re.sub(r'ACTIVITY ' + '[0-9]+', 'ACTIVITY', result)
# e.g., ACTIVITY 1: OR ACTIVITY1: OR ACTIVITY: OR ACTIVITY : -> ACTIVITY
result = re.sub(r'ACTIVITY' + '[0-9| ]*:', 'ACTIVITY', result)
# e.g., 1. ACTIVITY -> ACTIVITY
result = re.sub(r'\d\. ACTIVITY', 'ACTIVITY', result)
# remove (SOMETHING)
result = re.sub(re.compile(r'\([^)]*\) *'), "", result)
# remove #
# e.g., "budget": "0" # free admission -> "budget": "0"
result = re.sub(re.compile(r'# *.*'), "", result)
if "ACTIVITYACTIVITY" in result:
result = result.replace("ACTIVITYACTIVITY", "ACTIVITY")
if "ACTIVITY\nACTIVITY" in result:
result = result.replace("ACTIVITY\nACTIVITY", "ACTIVITY")
return result
def remove_verb(p: str) -> str:
p = p.replace("Artistic exploration at ", "")
p = p.replace("Breakfast at ", "")
p = p.replace("Brunch at ", "")
p = p.replace("Chimaek at ", "")
p = p.replace("Coffee at ", "")
p = p.replace("Coffee and Cake at ", "")
p = p.replace("Coffee and dessert at ", "")
p = p.replace("Coffee break at ", "")
p = p.replace("Cycling in ", "")
p = p.replace("Dinner at ", "")
p = p.replace("Explore ", "")
p = p.replace("Go to ", "")
p = p.replace("Hiking at ","")
p = p.replace("Karaoke at ", "")
p = p.replace("Lunch at ", "")
p = p.replace("Night view at ", "")
p = p.replace("Night view of ", "")
p = p.replace("Picnic at ", "")
p = p.replace("Riding the Cable Car at ", "")
p = p.replace("Relax at ", "")
p = p.replace("Shopping at ", "")
p = p.replace("Shopping in ", "")
p = p.replace("Stroll around ", "")
p = p.replace("Tea time at ", "")
p = p.replace("Visit ", "")
p = p.replace("Visit to ", "")
p = p.replace("Visiting ", "")
p = p.replace("Virtual Reality Experience at ", "")
p = p.replace("Walking tour of ", "")
p = p.replace("Walking tour in ", "")
p = p.replace("Walk along ", "")
p = p.replace("Walk in ", "")
p = p.replace("Walk around ", "")
return p
def change_route(meeting_time: str,
parting_time: str,
budget: str,
place: str,
user_request: str,
prior_places: List[str]) -> str:
m_year, m_time = meeting_time.split("T")
m_year = m_year.split('-')
meeting_time = m_time.strip() + ", " + m_year[0] + "/" + m_year[1] + "/" + m_year[2]
p_year, p_time = parting_time.split("T")
p_year = p_year.split('-')
parting_time = p_time.strip() + ", " + p_year[0] + "/" + p_year[1] + "/" + p_year[2]
print(f"time: from {meeting_time} to {parting_time}, budget: {budget}, place: {place}")
print("---------------------------------------------------")
USER_REQUEST_PROMPT = ""
# user_request: "가격이 너무 비싸요"
if user_request:
user_request = translator(user_request, task='ko-en') # 'ko-en' or 'en-ko'
USER_REQUEST_PROMPT = f"You should consider this message: {user_request}. "
# prior_places: ['서울올림픽미술관', '꼬꼬춘천치킨', '코엑스 아쿠아리움', '스타필드 코엑스몰']
if prior_places:
temp_places = ""
for idx, pp in enumerate(prior_places):
if idx == len(prior_places) - 1:
temp_places += translator(pp, task="ko-en")
else:
temp_places = temp_places + translator(pp, task="ko-en") + ", "
USER_REQUEST_PROMPT += f"You must not include the following locations: {temp_places}"
print("USER_REQUEST_PROMPT: ", USER_REQUEST_PROMPT)
LOCO_INSTRUCTION_RPOMPT = f"""Plan a date from {meeting_time} to {parting_time}, budget is {budget}, and Meeting place is {place}.
You should actively recommend names of place that exist in reality, what it actually costs.
{USER_REQUEST_PROMPT}.
You should recommend only one place and don't give me a choice.
OUTPUT FORMAT:
```
ACTIVITY
"activity_name": "Starbucks Yangjae Station Branch",
"start_time": "2023-04-09T13:00:00",
"end_time": "2023-04-09T15:00:00",
"description": "Starbucks is the world's largest multinational coffee chain.",
"budget": "10000"
```
"activity_name" should be a place only and "description" should consist of one sentence
"""
LOCO_SUFFIX_PROMPT = """You must follow the OUTPUT FORMAT, Begin!:"""
response = openai.ChatCompletion.create(
model="gpt-3.5-turbo",
messages=[{"role": "user",
"content": f"{LOCO_INSTRUCTION_RPOMPT}. {LOCO_SUFFIX_PROMPT}"}],
temperature=1.1,
max_tokens=2048,
top_p=0.9,
frequency_penalty=0.0,
presence_penalty=0.0
)
result = response.choices[0].message.content
return result
def generate_route(meeting_time: str,
parting_time: str,
budget: str,
place: str,
user_request: str) -> str:
m_year, m_time = meeting_time.split("T")
m_year = m_year.split('-')
meeting_time = m_time.strip() + ", " + m_year[0] + "/" + m_year[1] + "/" + m_year[2]
p_year, p_time = parting_time.split("T")
p_year = p_year.split('-')
parting_time = p_time.strip() + ", " + p_year[0] + "/" + p_year[1] + "/" + p_year[2]
print(f"time: from {meeting_time} to {parting_time}, budget: {budget}, place: {place}")
print("---------------------------------------------------")
USER_REQUEST_PROMPT = ""
# user_request: "가격이 너무 비싸요"
if user_request:
user_request = translator(user_request, task='ko-en') # 'ko-en' or 'en-ko'
USER_REQUEST_PROMPT = f"You should consider this message: {user_request}. "
# date time <= 2 hours -> recommend a place
SINGLE_PLACE_PROMPT = ""
if int(p_time.split(":")[0]) - int(m_time.split(":")[0]) <= 3:
SINGLE_PLACE_PROMPT = "If date time is less than 2 hours, you recommend only one place. Do not give me a choice."
LOCO_INSTRUCTION_RPOMPT = f"""Plan a date from {meeting_time} to {parting_time}, budget is {budget}, and Meeting place is {place}.
You should actively recommend names of place that exist in reality, what it actually costs.
{USER_REQUEST_PROMPT}. {SINGLE_PLACE_PROMPT}
OUTPUT FORMAT:
```
ACTIVITY
"activity_name": "Starbucks Yangjae Station Branch",
"start_time": "2023-04-09T13:00:00",
"end_time": "2023-04-09T15:00:00",
"description": "Starbucks is the world's largest multinational coffee chain.",
"budget": "10000"
```
"activity_name" should be a place only, "budget" should be only numbers and "description" should consist of one sentence
"""
LOCO_SUFFIX_PROMPT = """You must follow the OUTPUT FORMAT, Begin!:"""
response = openai.ChatCompletion.create(
model="gpt-3.5-turbo",
messages=[{"role": "user",
"content": f"{LOCO_INSTRUCTION_RPOMPT}. {LOCO_SUFFIX_PROMPT}"}],
temperature=1.1,
max_tokens=2048,
top_p=0.9,
frequency_penalty=0.0,
presence_penalty=0.0
)
result = response.choices[0].message.content
return result
class LOCO:
def __init__(self):
print("Initializing LOCO")
init()
def inference(self,
meeting_time: str,
parting_time: str,
place: str,
budget: int,
user_request: str,
prior_places: List[str]) -> List[Dict]:
# preprocessing
place = place_en_to_ko(place)
budget = str(budget) + " Won"
if prior_places: # user-requested try (not empty)
print("Change a meeting place")
print("---------------------------------------------------")
# change a meeting place
result = change_route(meeting_time,
parting_time,
budget,
place,
user_request,
prior_places)
else: # first try
# generate schedule
print("Generate Routes")
result = generate_route(meeting_time,
parting_time,
budget,
place,
user_request)
print(result)
print("---------------------------------------------------")
# postprocessing
result = postprocessing(result)
print(result)
print("---------------------------------------------------")
# split paragraph
parsed_result = result.replace("\n\n", "").replace("\n", "").split("ACTIVITY")[1:]
# remove verb from activity, translate en to ko, and json formatting
parsed_result = [json.loads("{" + remove_verb(p) + "}") for p in parsed_result]
for r in parsed_result:
print(r)
print("---------------------------------------------------")
# translation and then budget; str to int
for idx in range(len(parsed_result)):
parsed_result[idx]["activity_name"] = translator(parsed_result[idx]["activity_name"], task='en-ko')
parsed_result[idx]["description"] = translator_chatgpt(parsed_result[idx]["description"], task='en-ko')
try:
parsed_result[idx]["budget"] = int(parsed_result[idx]["budget"])
except:
parsed_result[idx]["budget"] = 0
return parsed_result