File size: 5,473 Bytes
78163ee
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1532744
 
 
 
 
78163ee
1532744
78163ee
 
 
 
 
 
1532744
 
 
 
 
78163ee
1532744
78163ee
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1532744
78163ee
 
 
 
1532744
78163ee
 
 
 
 
1532744
78163ee
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1532744
78163ee
 
 
 
 
 
 
 
 
 
1532744
78163ee
 
 
 
 
 
 
 
 
 
 
 
1532744
78163ee
 
 
1532744
78163ee
 
 
 
 
 
 
 
 
 
 
 
1532744
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
78163ee
 
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
import gradio as gr
import re
from difflib import Differ
from src.translation_agent.utils import *

LANGUAGES = {
    'English': 'English',
    'Español': 'Spanish',
    'Français': 'French',
    'Deutsch': 'German',
    'Italiano': 'Italian',
    'Português': 'Portuguese',
    'Русский': 'Russian',
    '中文': 'Chinese',
    '日本語': 'Japanese',
    '한국어': 'Korean',
    'العربية': 'Arabic',
    'हिन्दी': 'Hindi',
}


def diff_texts(text1, text2, lang):
    d = Differ()
    ic(lang)
    if lang == '中文':
        return [
            (token[2:], 
             "+" if token[0] == "+" else
             "-" if token[0] == "-" else
             "~" if token[0] == "?" else
             None)
            for token in d.compare(text1, text2)
            if token[0] in ["+", "-", "?", " "]
        ]
    else:
        words1 = re.findall(r'\S+|\s+', text1)
        words2 = re.findall(r'\S+|\s+', text2)

        return [
            (token[2:], 
             "+" if token[0] == "+" else
             "-" if token[0] == "-" else
             "~" if token[0] == "?" else
             None)
            for token in d.compare(words1, words2)
            if token[0] in ["+", "-", "?", " "]
        ]
    

def translate_text(source_lang, target_lang, source_text, country, max_tokens=MAX_TOKENS_PER_CHUNK):
    num_tokens_in_text = num_tokens_in_string(source_text)

    ic(num_tokens_in_text)

    if num_tokens_in_text < max_tokens:
        ic("Translating text as single chunk")

        #Note: use yield from B() if put yield in function B()
        translation_1 = one_chunk_initial_translation(
            source_lang, target_lang, source_text
        )
        yield translation_1, None, None, None

        reflection = one_chunk_reflect_on_translation(
            source_lang, target_lang, source_text, translation_1, country
        )
        yield translation_1, reflection, None, None

        translation_2 = one_chunk_improve_translation(
            source_lang, target_lang, source_text, translation_1, reflection
        )
        translation_diff = diff_texts(translation_1, translation_2, target_lang)
        yield translation_1, reflection, translation_diff, translation_2

    else:
        ic("Translating text as multiple chunks")

        token_size = calculate_chunk_size(
            token_count=num_tokens_in_text, token_limit=max_tokens
        )

        ic(token_size)

        text_splitter = RecursiveCharacterTextSplitter.from_tiktoken_encoder(
            model_name = "gpt-4",
            chunk_size=token_size,
            chunk_overlap=0,
        )

        source_text_chunks = text_splitter.split_text(source_text)
        
        translation_1_chunks = multichunk_initial_translation(
            source_lang, target_lang, source_text_chunks
        )
        ic(translation_1_chunks)
        translation_1 = "".join(translation_1_chunks)
        yield translation_1, None, None, None

        reflection_chunks = multichunk_reflect_on_translation(
            source_lang,
            target_lang,
            source_text_chunks,
            translation_1_chunks,
            country,
        )
        ic(reflection_chunks)
        reflection = "".join(reflection_chunks)
        yield translation_1, reflection, None, None

        translation_2_chunks = multichunk_improve_translation(
            source_lang,
            target_lang,
            source_text_chunks,
            translation_1_chunks,
            reflection_chunks,
        )
        ic(translation_2_chunks)
        translation_2 = "".join(translation_2_chunks)
        translation_diff = diff_texts(translation_1, translation_2, target_lang)
        
        yield translation_1, reflection, translation_diff, translation_2
        

def update_ui(translation_1, reflection, translation_diff):
    return gr.update(value=translation_1), gr.update(value=reflection), gr.update(value=translation_diff), gr.update(value=translation_2)

with gr.Blocks() as demo:
    gr.Markdown("# Andrew Ng's Translation Agent ")
    with gr.Row():
        source_lang = gr.Dropdown(choices=list(LANGUAGES.keys()), value='English', label="Source Language")
        target_lang = gr.Dropdown(choices=list(LANGUAGES.keys()), value='中文', label="Target Language")
        country = gr.Textbox(label="Country (for target language)")
    source_text = gr.Textbox(label="Source Text", lines=5, show_copy_button=True)

    btn = gr.Button("Translate")

    with gr.Row():
        translation_1 = gr.Textbox(label="Initial Translation", lines=3, show_copy_button=True)
        reflection = gr.Textbox(label="Reflection", lines=3, show_copy_button=True)

    translation_diff = gr.HighlightedText(
        label="Comparison",
        combine_adjacent=True,
        show_legend=True,
        color_map={
            "+": "green",   # 新增的文本显示为绿色
            "-": "red",     # 删除的文本显示为红色
            "~": "yellow"   # 修改的文本显示为黄色
        }
    )
    translation_2 = gr.Textbox(label="Final Translation", lines=5, show_copy_button=True)

    btn.click(translate_text, inputs=[source_lang, target_lang, source_text, country], outputs=[translation_1, reflection, translation_diff, translation_2], queue=True)
    btn.click(update_ui, inputs=[translation_1, reflection, translation_diff], outputs=[translation_1, reflection, translation_diff, translation_2], queue=True)

demo.launch()