O96a commited on
Commit
ef7b5cd
·
verified ·
1 Parent(s): 70c8752

Upload benchmark_results.json with huggingface_hub

Browse files
Files changed (1) hide show
  1. benchmark_results.json +121 -0
benchmark_results.json ADDED
@@ -0,0 +1,121 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "experiment": {
3
+ "id": "opus-mt-arabic-benchmark-2026-03-28",
4
+ "date": "2026-03-28",
5
+ "domain": "nlp",
6
+ "task": "translation",
7
+ "models_tested": [
8
+ "Helsinki-NLP/opus-mt-en-ar",
9
+ "Helsinki-NLP/opus-mt-ar-en"
10
+ ],
11
+ "total_tests": 9
12
+ },
13
+ "results": [
14
+ {
15
+ "id": 1,
16
+ "direction": "en-ar",
17
+ "type": "formal",
18
+ "input": "Hello, how are you today? I hope you are doing well.",
19
+ "output": "-مرحباً، كيف حالك اليوم؟",
20
+ "latency_seconds": 3.32,
21
+ "quality": "good",
22
+ "notes": "Good MSA translation, formal register"
23
+ },
24
+ {
25
+ "id": 2,
26
+ "direction": "en-ar",
27
+ "type": "technical",
28
+ "input": "The machine learning model achieved 95% accuracy on the test dataset.",
29
+ "output": "وحقق نموذج التعلم الآلي 95 في المائة من الدقة في مجموعة بيانات الاختبار.",
30
+ "latency_seconds": 13.48,
31
+ "quality": "good",
32
+ "notes": "Accurate technical translation, handles ML terminology"
33
+ },
34
+ {
35
+ "id": 3,
36
+ "direction": "en-ar",
37
+ "type": "colloquial",
38
+ "input": "I need to fix this bug before the deadline tomorrow.",
39
+ "output": "أحتاج لإصلاح هذه الحشرة قبل الموعد النهائي غداً",
40
+ "latency_seconds": 4.18,
41
+ "quality": "good",
42
+ "notes": "Good colloquial translation"
43
+ },
44
+ {
45
+ "id": 4,
46
+ "direction": "en-ar",
47
+ "type": "code-switching",
48
+ "input": "The API endpoint returns a JSON response with status code 200.",
49
+ "output": "نقطة نهاية API ترجع استجابة JSON مع رمز الوضع 200.",
50
+ "latency_seconds": 8.97,
51
+ "quality": "good",
52
+ "notes": "Preserves technical terms (API, JSON), code-switching handled well"
53
+ },
54
+ {
55
+ "id": 5,
56
+ "direction": "ar-en",
57
+ "type": "msa",
58
+ "input": "الذكاء الاصطناعي يغير طريقة عملنا في العالم الحديث.",
59
+ "output": "Artificial intelligence changes the way we work in the modern world.",
60
+ "latency_seconds": 3.63,
61
+ "quality": "good",
62
+ "notes": "Accurate MSA translation"
63
+ },
64
+ {
65
+ "id": 6,
66
+ "direction": "ar-en",
67
+ "type": "technical",
68
+ "input": "يتم تدريب النموذج باستخدام خوارزمية التعلم العميق على بيانات كبيرة.",
69
+ "output": "The model is trained using a deep learning algorithm on large data.",
70
+ "latency_seconds": 3.75,
71
+ "quality": "good",
72
+ "notes": "Accurate technical translation"
73
+ },
74
+ {
75
+ "id": 7,
76
+ "direction": "ar-en",
77
+ "type": "politeness",
78
+ "input": "شكراً جزيلاً على المساعدة، أقدر ذلك كثيراً.",
79
+ "output": "Thank you so much for your help. I appreciate that very much.",
80
+ "latency_seconds": 3.88,
81
+ "quality": "good",
82
+ "notes": "Accurate formal politeness"
83
+ },
84
+ {
85
+ "id": 8,
86
+ "direction": "ar-en",
87
+ "type": "dialect_egyptian",
88
+ "input": "إزيك؟ كله تمام؟ كنت عايز أسألك عن حاجة.",
89
+ "output": "I was gonna ask you something.",
90
+ "latency_seconds": 0.42,
91
+ "quality": "poor",
92
+ "notes": "TRUNCATED: Missed greeting 'izayyak? kullu tamam?', only translated second half"
93
+ },
94
+ {
95
+ "id": 9,
96
+ "direction": "ar-en",
97
+ "type": "dialect_sudanese",
98
+ "input": "يا زول، كيف حالك؟ تعال نتغدا سوا النهاردة.",
99
+ "output": "Hey, Zol, how are you?",
100
+ "latency_seconds": 0.58,
101
+ "quality": "poor",
102
+ "notes": "TRUNCATED: Missed 'come have lunch together today', 'Zol' untranslated (Sudanese term)"
103
+ }
104
+ ],
105
+ "summary": {
106
+ "avg_latency_msa": 5.67,
107
+ "avg_latency_dialectal": 0.5,
108
+ "msa_accuracy_rate": 100,
109
+ "dialectal_accuracy_rate": 0,
110
+ "key_finding": "OPUS-MT handles Modern Standard Arabic well but truncates Egyptian and Sudanese dialectal inputs, missing greetings and colloquial phrases"
111
+ },
112
+ "recommendations": [
113
+ "Use OPUS-MT for MSA content only",
114
+ "Implement dialect detection before translation",
115
+ "Consider NLLB-200 or specialized dialect models for Arabic dialects",
116
+ "Add preprocessing for Egyptian, Sudanese, and other dialectal inputs"
117
+ ],
118
+ "benchmark_by": "O96a",
119
+ "benchmark_date": "2026-03-28T09:40:00Z",
120
+ "huggingface_discussion": "https://huggingface.co/Helsinki-NLP/opus-mt-ar-en/discussions/10"
121
+ }