IndicGenBench/prompt.py at main · codingstark-dev/IndicGenBench · 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
"""
This file contains prompt templates for each task in IndicGenBench.
These templates are designed to be consistent across different models.
"""

TASK_DESCRIPTIONS = {
    "crosssum_in": "Given an English article, generate a concise summary in {language}.",
    "flores_in": "Translate the text from {source_language} to {target_language} accurately.",
    "xquad_in": "Answer the question in {language} based on the provided passage in {language}.",
    "xorqa_in": "Answer the question in {language} based on the provided English passage."
}

# CrossSum: English-to-Indic summarization
CROSSSUM_PROMPT = """# Task: Cross-lingual Summarization
# Description: Summarize the following English text in {language}.
# Instructions: Provide a concise and accurate summary of the main points in {language}.

English Text:
{text}

Summary in {language}:
"""

# Flores: Translation in both directions
FLORES_EN_TO_INDIC_PROMPT = """# Task: Machine Translation
# Description: Translate from English to {language}.
# Instructions: Provide an accurate and fluent translation.

English Text:
{source}

{language} Translation:
"""

FLORES_INDIC_TO_EN_PROMPT = """# Task: Machine Translation
# Description: Translate from {language} to English.
# Instructions: Provide an accurate and fluent translation.

{language} Text:
{source}

English Translation:
"""

# XQuAD: Indic QA (question and context in same language)
XQUAD_PROMPT = """# Task: Question Answering
# Description: Answer the question in {language} based on the passage in {language}.
# Instructions: Extract the answer from the passage. Your answer should be short and precise.

Passage ({language}):
{context}

Question ({language}):
{question}

Answer ({language}):
"""

# XorQA: Cross-lingual QA (Indic question, English context)
XORQA_PROMPT = """# Task: Cross-lingual Question Answering
# Description: Answer the {language} question based on the English passage.
# Instructions: Extract the answer from the passage and provide it in {language}.

Passage (English):
{context}

Question ({language}):
{question}

Answer ({language}):
"""

def get_prompt_for_task(task, example, language_name=None):
    """Generate the appropriate prompt based on task and example"""
    if language_name is None:
        language_name = example.get('lang', '')

    if task == "crosssum_in":
        return CROSSSUM_PROMPT.format(
            language=language_name,
            text=example.get('text', '')
        )

    elif task == "flores_in":
        # Check direction (en->indic or indic->en)
        src_lang = example.get('src_lang', '')
        # tgt_lang = example.get('tgt_lang', '')

        if src_lang == 'en':
            return FLORES_EN_TO_INDIC_PROMPT.format(
                language=language_name,
                source=example.get('source', '')
            )
        else:
            return FLORES_INDIC_TO_EN_PROMPT.format(
                language=language_name,
                source=example.get('source', '')
            )

    elif task == "xquad_in":
        return XQUAD_PROMPT.format(
            language=language_name,
            context=example.get('context', ''),
            question=example.get('question', '')
        )

    elif task == "xorqa_in":
        return XORQA_PROMPT.format(
            language=language_name,
            context=example.get('context', ''),
            question=example.get('question', '')
        )

    return ""