Dataset Viewer
Auto-converted to Parquet Duplicate
text
stringlengths
119
677k
id
stringlengths
32
32
source
stringclasses
5 values
subset
stringclasses
5 values
token_count
int64
45
244k
toxic_score
float64
-0.42
3.5
toxic_int_score
int64
1
3
edu_score
float64
0.27
5.68
edu_int_score
int64
1
5
क्रिकेट की दुनिया के दो महान खिलाड़ी, एक महेंद्र सिंह धोनी और दूसरे विराट कोहली. इन दोनों का एक वीडियो इन दिनों सोशल मीडिया पर छाया हुआ है. ये वीडियो क्रिकेट से जुड़ा नहीं है, लेकिन फिर भी लोगों को काफी पसंद आ रहा है. इस वीडियो में विराट और धोनी साथ में जबरदस्त जुगलबंदी के साथ गाना गाते हुए नजर आ रहे हैं. देखें ये मजेद...
8b326545ee2b80047f58f3cf8809da81
https://huggingface.co/datasets/HuggingFaceFW/fineweb-2
finweb_2_hi
77
1.168886
1
1.54338
2
दैनिक भास्कर के राजस्थान स्टेट हेड लक्ष्मीप्रसाद पंत को बड़ी जिम्मेदारी मिली है। सोमवार 7 दिसंबर को पंत को भास्कर का नेशनल एडिटर (समस्त हिंदी संस्करण) बना दिया गया। इस आशय की मेल सोमवार सुबह जारी हो गयी। लक्ष्मी पंत की हिन्दुस्तान अखबार में शशिशेखर के स्थान पर जाने की चर्चा बीच में काफी चल रही थी। कुछ लोगों का कहना है ...
d8f61c336504db6e7477ccd67e857759
https://huggingface.co/datasets/HuggingFaceFW/fineweb-2
finweb_2_hi
313
1.224414
1
1.370893
1
कश्मीर और गिलगित बालटिस्तान के क्षेत्र को सऊदी अरब ने पाकिस्तान से अलग करार दिया है और इसे भारत में मोदी सरकार के समर्थक लगातार मोदी सरकार की नई जीत बता रहे हैं. आप चाहें तो इसे सबसे अव्वल दर्जे की अंधभक्ति भी कह सकते हैं. सोशल मीडिया में व्हाट्सएप से लेकर फेसबुक और ट्विटर तक मोदी सरकार के समर्थक कई हैंडल और पेज बना रख...
b1add557a219a736514684e679d0cb8a
https://huggingface.co/datasets/HuggingFaceFW/fineweb-2
finweb_2_hi
431
2.246745
2
2.237108
2
There are महिलाओं के लिए मोटापा कम करने वाले व्यायाम of plans available for people who want to lose weight. यूरिया दखभाल खड़े होने तथा बैठने के लिए पर्याप्त स्थान. Summer Food For Hypertension: हाइपरटेंशन के हैं मरीज तो गर्मियों में खाएं ये पांच चीजें. मुख से रक्त की गंध। 30. StOp . मोटे होने का कैप्सूल दवातेइस्लामि घर...
970b8684cb4ef6e6bed00aa543347e24
https://huggingface.co/datasets/HuggingFaceFW/fineweb-2
finweb_2_hi
1,435
1.100665
1
2.748829
3
भारत में दुनिया का सबसे बड़ा लॉकडाउन 14 और दिनों के लिए बढ़ा दिया गया है. कोरोना को फैलने से रोकने के लिए केंद्र सरकार ने लॉकडाउन को 31 मई तक बढ़ा दिया है. लॉकडाउन फेज-4 में कुछ प्रतिबंधों से छूट दी गई है. इसमें कार्यालयों, कारखानों और अन्य औद्योगिक इकाइयों को फिर से खोलने की अनुमति दी गई है. कंटेनमेंट जोन को छोड़कर बा...
dbbf8323818f93f8b5e479bf7472cca1
https://huggingface.co/datasets/HuggingFaceFW/fineweb-2
finweb_2_hi
562
1.318443
1
2.138419
2
नई दिल्ली: भारत और इंग्लैंड के बीच चार मैचों की टेस्ट सीरीज की शुरुआत 5 फरवरी से होनी है. चेन्नई के एमए चिदंबरम स्टेडियम में खेले जाने वाले इस मुकाबले को लेकर क्रिकेट प्रेमी से लेकर पूर्व दिग्गज भी उत्साहित हैं. पूर्व भारतीय ऑलराउंडर इरफान पठान इस सीरीज के दौरान कमेंट्री करते नजर आएंगे. इरफान ने एक इंटरव्यू में दोनों ट...
a1eb19c746de866333f54946fabcf25a
https://huggingface.co/datasets/HuggingFaceFW/fineweb-2
finweb_2_hi
512
1.015946
1
1.94494
2
हिमाचल प्रदेश में तेजी से फैल रहे कोरोना संक्रमण के बावजूद जयराम सरकार फिलहाल न लॉकडाउन लगाएगी और न ही नाइट कर्फ्यू लगेगा। कोचिंग सेंटर, मेडिकल और डेंटल कालेज भी यथावत खुले रहेंगे। नवरात्र में भारी भीड़ जुटने के बावजूद मंदिर दर्शनों के लिए खुले रहेेंगे। बसों, निजी वाहनों और टैक्सियों में फुल क्षमता के साथ लोग यात्रा कर...
43cf69cf5b35cde4db4ec634c349637d
https://huggingface.co/datasets/HuggingFaceFW/fineweb-2
finweb_2_hi
522
1.097852
1
1.918899
2
इन एक्ट्रेसेस ने न केवल स्विमवियर वाइब्स को अच्छे से संभाल रखा है बल्कि इस तरह के स्टाइलिश टू पीस सेपरेट्स सेट्स को वह फुल कॉन्फिडेंट और कंफर्टेबल अंदाज के साथ कैरी भी करती हैं। इन्हीं में से एक बॉलीवुड एक्ट्रेस दिशा पाटनी (Disha Patani) भी हैं, जो अपने रिसोर्ट-वियर स्टाइल को फैशन स्टेटमेंट बनाने की इच्छा रखती हैं। बिक...
741962cedab7145fe0a53a33cafe3ffd
https://huggingface.co/datasets/HuggingFaceFW/fineweb-2
finweb_2_hi
782
1.475718
1
1.278748
1
50 हजार का लैपटॉप, खरीदें बस 10 हज़ार में, जानें कैसे? वर्क फ्रॉम होम की वजह से आजकल पर्सनल कंप्यूटर से लेकर लैपटॉप की मार्केट में भारी मांग है नई दिल्ली: कोरोना वायरस की वजह से सबको परेशानी झेलनी पड़ रही है। लोगों का काम की वजह से निकलना भी बंद हो गया हैक्योंकि सभी कंपनियां अपने अपने कर्मचारियों के लिए वर्क फ्रॉम होम ...
8b2e2fef3ed517cfed729b1125053d8c
https://huggingface.co/datasets/HuggingFaceFW/fineweb-2
finweb_2_hi
376
1.140843
1
1.549398
2
एक शादीशुदा व्यक्ति के कारण होगा के लिए, मुझे यकीन है कि इंडियम शुरुआत मैं था मोहक केवल मेरे साथी रखा है सेक्सी खेल की कहानी कह परमाणु संख्या 102 सेक्स करने के लिए तो oer समय ऐसा लग रहा था जैसे गंवा श्रम और मुझे ऊपर की तरफ अनुवाद में हिस्सेदारी के सिद्धांत मैं मतदान में शुरू और मेरे पति के साथ रखा वोट लाल या मेरे पति अ...
941edc4de612e34662b2fe626429a106
https://huggingface.co/datasets/HuggingFaceFW/fineweb-2
finweb_2_hi
303
2.230679
2
1.245465
1
घर में इस तरह चमकाएं चांदी के गहने चांदी की ज्वेलरी, बर्तन या सिक्के अगर काले पड़ गए हैं तो आप इन तरीकों से उन्हें घर में साफ कर सकते हैं। अगर आपके घर में कोई चांदी की ज्वेलरी, बर्तन या सिक्के ऐसे हैं, जो काले पड़ गए हैं तो आप उन्हें घर में ही चमका सकते हैं। जी हां, अब बार-बार चांदी की ज्वेलरी को चमकाने के लिए ज्वैलर्स...
c3ca4f6e795b8dd1a7c85fdc7f6df84e
https://huggingface.co/datasets/HuggingFaceFW/fineweb-2
finweb_2_hi
365
1.40079
1
1.540044
2
भारतीय क्रिकेट टीम इस वर्ष नवंबर में ऑस्ट्रेलिया (India Tour Of Australia 2020) का दौरा करेगी जहां उसे 4 मैचों की टेस्ट सीरीज खेलनी है. इस सीरीज को लेकर टीम इंडिया के पूर्व ओपनर गौतम गंभीर (Gautam Gambhir) का कहना है कि विराट कोहली (Virat Kohli) की अगुआई वाली टीम के पास ऑस्ट्रेलियाई बल्लेबाजी क्रम को ध्वस्त करने के लिए...
244534ff38b56b03356d9898ab21995d
https://huggingface.co/datasets/HuggingFaceFW/fineweb-2
finweb_2_hi
445
1.227663
1
2.060157
2
डेस्क। सर्दी हो या गर्मी, अक्सर लोग अदरक वाली चाय पीना चाहते हैं। चाय की दुकानों पर भी लोग कहते मिल जाएंगे कि भाई अदरक डाल देना ऐसा बोलते हैं। लेकिन क्या आपने कभी सोचा है कि हर बार अदरक की चाय पीने से नुकसान भी हो सकता है? तो आइए जानें – रात को अदरक की चाय पीने से बचना चाहिए। कुछ लोग सोचते हैं कि रात को सोने से पहले अद...
d789cb8158f99266f74c958e90a5e5b6
https://huggingface.co/datasets/HuggingFaceFW/fineweb-2
finweb_2_hi
227
1.483042
1
2.799497
3
लखनऊ विश्वविद्यालय(Lucknow University) के तिलक छात्रावास का ड्रेस कोड सोशल मीडिया पर वायरल हो गया। इस ड्रेस कोड में छात्राओं को शॉर्ट्स या घुटने के ऊपर की ड्रेस न पहनने की फरमान दिया गया। वायरल नोटिस के मुताबिक, शॉर्ट्स या घुटने के ऊपर की ड्रेस पहनकर छात्राएं अपने ब्लॉक से बाहर नहीं घूम सकती हैं। इस नियम का उल्लंघन करन...
2df13674dd78065d7a1f83e8292a0ece
https://huggingface.co/datasets/HuggingFaceFW/fineweb-2
finweb_2_hi
246
2.459722
2
1.448198
1
Prabhat khabar Digital हिना खान 'ये रिश्ता क्या कहलता है' में अक्षरा के रोल से काफी फेमस हुई. शो में साड़ी पहने नजर आने वाली हिना को फैंस ने जब बिग बॉस में देखा तो शॉक्ड हो गए. उनका ट्रांसफॉर्मेशन देखकर हर कोई हैरान हो गया था. आज हिना टीवी पर मंहगी एक्ट्रेसेस में से एक है. टीवी एक्ट्रेस सुरभि चंदना ने तारक मेहता का उल्...
3a73c7ec34e872e78a20f7f848c6041a
https://huggingface.co/datasets/HuggingFaceFW/fineweb-2
finweb_2_hi
301
1.21102
1
1.397076
1
फिल्मी स्टार्स अपने ब्रेकअप, अफेयर और रिलेशनशिप की खबरों को लेकर सुर्खियों में बने रहते हैं. इन सितारों के लिए ये सब आम बात है. ये सितारे अपनी पर्सनल लाइफ की वजह से बहुत ही सुर्खियों में रहते हैं. आज हम आपको सलमान की एक ऐसी अभिनेत्री के बारे में बता रहे हैं जो निर्देशक के साथ कमरे में रंगे हाथों पकड़ी गई थी. इस अभिनेत्...
5e8f1f61b639a3b9047c62d642c02487
https://huggingface.co/datasets/HuggingFaceFW/fineweb-2
finweb_2_hi
371
1.550868
2
1.506833
2
हम नए क्रिप्टो-दुनिया को जीत के लिए शुरू कर दिया नई सुविधाओं की एक टन से लागू किया और अपने लक्ष्यों को पुनः समायोजन। लंदन, 10 जुलाई (भाषा) ब्रिटेन में भारत निवेश करने वाले देशों में अमेरिका के बाद दूसरे पायदान पर पहुंच गया है। उसने 2019 में यहां 120 परियोजनाओं में निवेश किया और 5,429 नये रोजगार के अवसरों का सृजन किया। ...
b309332ed12f41da6c5bbab1dc3f3e4b
https://huggingface.co/datasets/HuggingFaceFW/fineweb-2
finweb_2_hi
1,566
1.13706
1
1.873317
2
इस पल के रूप में, Microsoft ने अपने आगामी अगले जेनरेशनल लॉन्च के लिए आगे की ओर चिकनी नौकायन किया है। सोनी में उनकी मुख्य प्रतियोगिता वर्तमान में उनके साथ निर्माता समुदाय में उनके खड़े होने को तिरस्कृत कर रही है कॉपीराइट सिस्टम का दुरुपयोग। ऑप्टिक्स करने से पहले सिप्पुकु सोनी पहले से ही अपने $ 500-550 कंसोल के साथ एक कठ...
7972ce7797603c419bcb8e4020078218
https://huggingface.co/datasets/HuggingFaceFW/fineweb-2
finweb_2_hi
506
1.113588
1
1.20561
1
नई दिल्ली: रामानंद सागर की ‘रामायण’ (Ramayana) पिछले साल छोटे पर्दे पर रिकॉर्ड तोड़ने के बाद अब एक बार फिर स्टार भारत पर प्रसारित होने जा रही है. इस शो को फिर से टेलीकास्ट किया जाएगा ताकि लोगों को इस आंशिक लॉकडाउन में अपने घरों के अंदर रहने में मदद मिल सके और कोरोना वायरस के प्रसार को रोका जा सके. इस महामारी की दूसरी ल...
05ceb9bd4d8ead168141cb7d480c9518
https://huggingface.co/datasets/HuggingFaceFW/fineweb-2
finweb_2_hi
323
0.956809
1
1.673574
2
Ads से है परेशान? बिना Ads खबरों के लिए इनस्टॉल करें दैनिक भास्कर ऐप एक तरफ नपा शहर का कचरा 27 किमी दूर बाबई तहसील के बज्जरवाड़ा गांव में ट्रेंचिंग ग्राउंड बनाकर डंप करने की तैयारी कर रही है वहीं देखरेख नहीं हाेने से कचरा फेंकने वाली कांपेक्टर मशीन दो साल से शोपीस बनी है। नपा ने 2016 में 22 लाख रुपए से कांपेक्टर मशीन ख...
a27e52c036949341cdad85546369ae25
https://huggingface.co/datasets/HuggingFaceFW/fineweb-2
finweb_2_hi
438
1.398783
1
2.034102
2
आज पूरे देश में धनतेरस का पर्व मनाया जा रहा है। भारतीय धर्मों को मानने वाले लोग आज के दिन दीपावली की पूजा के लिये सामान आदि खरीदते हैं। इस दिन बाज़ार में भीड़ रहती है तो यह भी सच है कि स्टील, पीतल, सोने तथा अन्य धातुओं से निर्मित सामान भी अन्य दिनों की अपेक्षा अत्यंत महंगे हो जाते है। यही कारण है कि वर्ष के किसी भी दिन...
f656c25174d8bbab2f0e7ce912e50ac3
https://huggingface.co/datasets/HuggingFaceFW/fineweb-2
finweb_2_hi
894
1.808187
2
3.202126
3
बीेकानेर । राजकीय डूंगर काॅलेज में आगामी 27 अगस्त को होने वाले छात्र संघ चुनाव की तैयारियां प्रारम्भ कर दी गयी हैं। प्राचार्य डाॅ. सतीश कौशिक ने बताया कि डाॅ. रविन्द्र मंगल को मुख्य चुनाव अधिकारी नियुक्त किया गया है। इसके अलावा डाॅ. जी.पी.सिंह व डाॅ. शालिनी मूलचन्दानी को अतिरिक्त मुख्य निर्वाचन अधिकारी तथा डाॅ. आर.एस.व...
38c54221ba04333bd29dc234617ec854
https://huggingface.co/datasets/HuggingFaceFW/fineweb-2
finweb_2_hi
305
1.004101
1
1.572073
2
एक उप-रु। 10,000 स्मार्टफोन भारत में किसी भी कंपनी के लिए एक मुश्किल मामला है। मूल्य निर्धारण इस सेगमेंट में हाइपर-सेंसिटिव है, और कुछ सौ रुपये से भी गलत होने से बिक्री पर असर पड़ सकता है। Realme इस बाजार के लिए कोई अजनबी नहीं है, क्योंकि यह लंबे समय से स्मार्टफोन की सी-सीरीज़ है जो कि मूल्य-सचेत खरीदारों के लिए खानपान...
261201e600bb99ba4c90a41d8041726f
https://huggingface.co/datasets/HuggingFaceFW/fineweb-2
finweb_2_hi
1,223
1.023695
1
1.250514
1
दोस्तों रामायण के अनुसार बताया गया है कि लक्ष्मण ने शूर्पणखा कि नाक काट दी थी क्या ये बात सच है? तो आइये जानते लक्षमण शूर्पणखा कि विचित्र कहानी के बारे में, शूर्पणखा रावण कि बहन है राजा कालका के पुत्र कि पत्नी है, रावण ने अपनी बहन के पति का वध कर दिया था तब रावण ने शूर्पणखा को अपने पास रख लिया था एक दिन शूर्पणखा जंगल म...
98b95c5d24bd2276ee698e4b1df4f2b6
https://huggingface.co/datasets/HuggingFaceFW/fineweb-2
finweb_2_hi
330
2.800849
3
1.96707
2
वानखेड़े स्टेडियम में हाई-प्रोफाइल शाम के मैच की मेजबानी करने के लिए आवश्यक मशीनरी को धूल चटाने की प्रक्रिया पिछले सप्ताह शुरू हुई थी। इंडियन प्रीमियर लीग (आईपीएल) सभी के बाद भारतीय धरती पर वापसी होगी। लेकिन यह एक पहिया के cogs के बीच लौकिक मकड़ी-जाले नहीं था जिसे साफ किया जाना था। आयोजन स्थल पर सब कुछ बिजली से संबंधित...
615d03e6000cc7d48150280f54db3d22
https://huggingface.co/datasets/HuggingFaceFW/fineweb-2
finweb_2_hi
1,162
1.386234
1
2.020143
2
सैन फ्रांसिस्को, 23 मई (आईएएनएस)| अमेरिकी राष्ट्रपति डोनाल्ड ट्रंप के ट्विटर का उपयोग ‘टाइपराइटर’ के रूप में करने के उनके खुद के बयान के कुछ दिन बाद अब माइक्रो-ब्लॉगिंग साइट के सह-संस्थापक, ईव विलियम्स ने ट्रंप को प्लेटफॉर्म का ‘मास्टर’ बताया है। सीएनएन बिजनेस द्वारा बुधवार को प्रकाशित एक साक्षात्कार में विलियम्स ने कह...
349ae58c359bd1dfa3fb2fe2b4cd6c00
https://huggingface.co/datasets/HuggingFaceFW/fineweb-2
finweb_2_hi
381
1.382314
1
1.618512
2
सीएम योगी का चित्रकूट दौरा, अधूरे पड़े विकास कार्यों को मिली गति चित्रकूट। मुख्यमंत्री योगी आदित्यनाथ ने बुधवार को प्रभु श्रीराम की तपोभूमि में आठवीं बार पहुंचे। उन्होंने यहां विकास कार्यों को गति दी। मुख्यमंत्री ने बुधवार को चित्रकूट में रसिन बांध और चिल्ली मल पंप कैनाल का लोकार्पण करने के बाद बांदा में मंडलीय स्तर पर...
3b9a2bf28c571c81a2e67d0c727a2847
https://huggingface.co/datasets/HuggingFaceFW/fineweb-2
finweb_2_hi
295
1.200826
1
1.472208
1
जान्हवी कपूर और पंकज त्रिपाठी की गुंजन सक्सेना: द कारगिल गर्ल बीते 12 अगस्त को रिलीज़ हो गई. पैंडमिक की वजह से कई फ़िल्में OTT प्लेटफ़ॉर्म्स पर रिलीज़ की जा रही हैं. धर्मा प्रोडक्शन्स और ज़ी स्टूडियोज़ की इस फ़िल्म का निर्देशन शरन शर्मा ने किया है. ये फ़िल्म फ़्लाइट लेफ़्टिनेंट गुंजन सक्सेना के जीवन पर आधारित है. फ़िल्...
9af905876bc6dd9cc934ac30ffc4f2dc
https://huggingface.co/datasets/HuggingFaceFW/fineweb-2
finweb_2_hi
131
1.080423
1
1.238603
1
सर्दी का मौसम कवियों को मुबारक और स्कूल जाने वाले उन बच्चों को जिनकी छुट्टी पहले एक महीने की होती है, फिर बढ़ते बढ़ते डेढ़ दो महीने तक पहुंच जाती है. हमारे यानी कामकाजी उम्र के लोगों के लिए सर्दी में कुछ नहीं रखा. बस सोने में मज़ा आता है, लेकिन चैन की नींद मिलती कहां है! रात भर की जद्दोजहद के बाद जब जब रज़ाई परफ़ेक्ट ट...
b17ea523288cd85b66712431beaac6dd
https://huggingface.co/datasets/HuggingFaceFW/fineweb-2
finweb_2_hi
511
1.968611
2
2.467812
2
6. आपका व्यापार खुल जाएगा, और यह सक्रिय टूल के रूप में 'टूलबॉक्स' विंडो में दिखाई देगा। आज ब्रोकर बीसीएस 300,000 से अधिक ग्राहक खातों के मोबाइल ट्रेडिंग (ऐप) समीक्षा उपयोगकर्ता आधार के साथ रूसी बाजार में ब्रोकरेज के क्षेत्र में सबसे बड़े वित्तीय मध्यस्थों में से एक है। कंट्रोल दुकानें बंद रहने से गरीबों को सस्ता राशन म...
48d55adabf3970c8c9658d03b5f9944d
https://huggingface.co/datasets/HuggingFaceFW/fineweb-2
finweb_2_hi
1,134
1.146595
1
1.637564
2
February 13, 2020 SHARES अब पुनर्विकसित रेलवे स्टेशनों पर उपलब्ध जनसुविधाओं के लिए हवाई अड्डों की तरह शुल्क वसूल करेगा। बता दें कि हवाई यात्रा में यूजर डेवलपमेंट फीस (UDF) कर का हिस्सा होता है जिसका हवाई यात्री भुगतान करते हैं। यूडीएफ विभिन्न हवाई अड्डों पर वसूला जाता है और इसकी दरें विभिन्न पहलुओं पर निर्भर होने की वज...
54222e6717c783ec1cec44235d48b637
https://huggingface.co/datasets/HuggingFaceFW/fineweb-2
finweb_2_hi
330
1.032107
1
1.996314
2
July 17, 2020 SHARES सरकारी विमानन कंपनी एयर इंडिया ने अपने कुछ कर्मचारियों को छह महीने से लेकर पांच साल तक जबरन छुट्टी पर भेजने की तैयारी कर ली है. कंपनी ने जारी आधिकारिक बयान में यह जानकारी दी. इस दौरान छुट्टी पर भेजे गए कर्मचारियों को वेतन नहीं दिया जाएगा. इंडिया टुडे की खबर के मुताबिक एयर इंडिया की इस योजना को हाल ...
e78a0b55d89dc27fe0d52982169d5e81
https://huggingface.co/datasets/HuggingFaceFW/fineweb-2
finweb_2_hi
346
1.112246
1
1.457494
1
चीन की स्मार्टफोन निर्माता कंपनी Xiomi ने अपने दो स्मार्टफोन Mi A2 और Mi A2 Lite को स्पेन में आयोजित एक ग्लोबल इवेंट में लांच कर दिया हैं. आपको बता दें कि Xiomi Mi A2 बीते साल लांच किये गये शाओमी के पहले एंड्रायड वन स्मार्टफोन Mi A1 का अपग्रेड वर्जन हैं. दोस्तो दोनों ही स्मार्टफोन्स की बिक्री 27 जुलाई से फ्रांस में, 8अ...
317b4ea4e1b7489e6eda83d8907ccd31
https://huggingface.co/datasets/HuggingFaceFW/fineweb-2
finweb_2_hi
853
0.997554
1
1.680373
2
दक्षिणी कश्मीर के पुलवामा जिले में बारूदी सुरंग बिछाकर बड़ी वारदात को अंजाम देने की साजिश को सुरक्षा बलों ने विफल कर दिया। सड़क किनारे पांच-पांच किलो के बिछाए गए दो आईईडी को निष्क्रिय किया गया। माना जा रहा है कि सुरक्षा बलों के काफिले को उड़ाने की साजिश के तहत आतंकियों ने आईईडी लगा रखा था। इस घटना के बाद सुरक्षा बलों क...
0c04467dcd13ba40c4b4acb74cb6ba46
https://huggingface.co/datasets/HuggingFaceFW/fineweb-2
finweb_2_hi
137
1.301194
1
0.858249
1
End of preview. Expand in Data Studio

GigaLekh: A Large Hindi Text Corpus with Educational and Toxicity Annotations

Dataset Summary

This repository contains a large corpus of Hindi text, which has been filtered and annotated using classifiers for educational content and toxicity. The dataset is intended for training language models and other NLP applications in Hindi.

Supported Tasks and Leaderboards

This dataset can be utilized for tasks involving language modeling.

Languages

Hindi.

Dataset Structure

Data Instances

The dataset consists of the following features:

  • text: a string of text in Hindi.
  • source: the source where that string originated.
  • subset: a short string indicating the name of the subset (referring to the original dataset or crawl).
  • id: a unique identifier for each sample (md5 hash).
  • token_count: number of tokens in the text (Polygl0t/LilMoo tokenizer).
  • edu_score: a float representing the educational quality score predicted by the Polygl0t/hindi-roberta-edu-classifier (1 = non-educational, 5 = highly educational).
  • edu_int_score: an integer-rounded version of the edu_score.
  • toxic_score: a float representing the toxicity score predicted by the Polygl0t/hindi-roberta-toxicity-classifier (1 = non-toxic, 5 = highly toxic).
  • toxic_int_score: an integer-rounded version of the toxic_score.

Data Fields

{
  "text": "चाणक्य की गिनती भारत के श्रेष्ठ विद्वानों में की जाती है|चाणक्य को राजनीति शास्त्र, समाज शास्त्र, कूटनीति शास्त्र, सैन्य शास्त्र के साथ अर्थशास्त्र का भी गहरा ज्ञान था|",
  "source": "https://huggingface.co/datasets/HuggingFaceFW/fineweb-2",
  "subset": "finweb_2_hi",
  "id": "dc0ec0a28936637fd9ec5be2b3994eb8",
  "token_count": 410,
  "edu_score": 3.87568,
  "edu_int_score": 4,
  "toxic_score": 2.047186,
  "toxic_int_score": 2
}

Subsets and Splits

After applying the complete filtering pipeline—including text extraction, language identification, heuristic quality filtering, deduplication, and learned filtering using our trained classifiers, we further (1) removed documents shorter than 50 tokens, and (2) removed documents with a toxicity score > 3. Given that toxicity is a particularly challenging issue in Hindi web data, we opted for preserving the filtered out documents as a separate subset of our corpus, which can be useful for future research on toxicity detection and mitigation in Hindi NLP.

from datasets import load_dataset

# Load the main dataset
ds = load_dataset("Polygl0t/hindi-corpus", "default", split="train")

# Load the excluded subset
excluded_ds = load_dataset("Polygl0t/hindi-corpus", "excluded", split="train")

# If you don't want to download the entire dataset, set streaming to `True`
ds = load_dataset("Polygl0t/hindi-corpus", "default", split="train", streaming=True)

Statistics

Subset Files Rows Size (GB) Tokens
default 35 83,081,507 260.68 90,705,245,239
excluded 11 498,892 1.76 1,545,290,236
Total 46 83,580,399 262.44 92,250,535,475

Dataset Creation

Curation Rationale

To curate this dataset, we developed an implementaton of the FineWeb2 filtering pipeline, while also adding a learned-filter (LLM-as-a-Judge) approach based on works like FineWeb-Edu, the original Phi paper, and other works that document a similar approach.

Text Extraction and Language Identification

For web-crawled data sourced from CC WARC files, we begin by extracting text content using the Trafilatura library, configured to favor precision over recall. For datasets sourced from Hugging Face, which often include pre-cleaned text, this extraction step is bypassed, allowing direct input into the subsequent filtering stages.

To ensure relevance and a minimal level of filtering at this early stage, we apply an initial URL Filter to remove documents from undesirable sources, leveraging blocklists (Maravento's blackweb) to exclude low-quality or inappropriate websites. This is followed by a language identification step using the FastText FT176 model. To enhance language identification accuracy, we employ a second round of language identification using GlotLID as backend.

Quality Filtering and Formatting

For the purpose of building a robust Hindi corpus, we define quality based on several heuristic criteria that reflect linguistic coherence and structural integrity. In short, we import the filters developed during the creation of FineWeb-2 and MassiveText, tunning them to be more sensitive to Hindi's linguistic characteristics. Post-filtering, we apply formatting steps to corrects encoding issues, remove personally identifiable information, and eliminate/replace undesirable patterns (e.g., excessive symbols). For all of this, we used the standard implementation from the Datatrove library to implement these filters.

Deduplication

Deduplication allows us to train models that emit memorized text ten times less frequently and require fewer train steps to achieve the same or better accuracy. Hence, to address redundancy, we implement a deduplication pipeline using the MinHash algorithm. Following FineWeb-2 implementation, we use 14 buckets, 8 hashes per bucket, and 5-grams, employing the xxHash function for hashing.

Learned Filtering with LLM-as-a-Judge

Learned filters, often based on large language models, can evaluate text quality more holistically, considering context, coherence, and other nuanced factors that heuristic methods might overlook. Hence, we implemented an LLM-as-a-Judge filtering approach to further enrich our dataset with annotations. Specifically, we used Qwen/Qwen2.5-32B-Instruct to evaluate the quality of documents that passed through the initial heuristic filters. This annotation process involved prompting the model to assess documents based on two criteria: educational quality and toxicity. The result was the following datasets:

With these annotations, we trained two separate classification models to automate the filtering process for our corpus curation:

The columns edu_score and toxic_score in this dataset correspond to the predicted scores from these classifiers, while edu_int_score and toxic_int_score are the integer-rounded versions of these scores.

Source Data

We sourced data from a variety of existing datasets available on Hugging Face. These datasets offer a mix of curated and semi-curated content, providing a solid foundation for our corpus to start (see the table below). Unlike web crawls, these datasets often come with some level of quality control, allowing us to bypass some of the initial cleaning steps (e.g., cleaning of HTML tags). However, we also recognized the need to supplement these datasets with more extensive web-crawled data to ensure sufficient coverage and diversity, especially when it comes to adding timely data from recent snapshots of CommonCrawl.

  • Cutoff Date: December 2025

Who are the source language producers?

All text samples are native to Hindi or translated from other languages to Hindi (slight contamination of different languages should also be expected).

Annotations

Annotation process

Annotations were created using a learned-filtering approach based on large language models (LLM-as-a-Judge). Specifically, we used Qwen/Qwen2.5-32B-Instruct to evaluate the quality of documents that passed through the initial heuristic filters. This annotation process involved prompting the model to assess documents based on two criteria: educational quality and toxicity. With these annotations, we trained two separate classification models to automate the filtering process for our corpus curation.

Personal and Sensitive Information

This dataset was filtered to remove personally identifiable information (PII) using standard PII detection methods as part of the formatting step in the filtering pipeline. Moreover, users should be aware that the excluded subset contains documents that were filtered out due to high toxicity scores, which may include offensive or sensitive content.

Considerations for Using the Data

Social Impact of Dataset

The creation of a large Hindi text corpus has the potential to significantly advance NLP research and applications for Hindi speakers, who represent a substantial portion of the global population. By providing high-quality training data, this dataset can facilitate the development of more accurate and effective language models, which can be used in various applications such as machine translation, sentiment analysis, and information retrieval.

Additional Information

Dataset Maintainers

Licensing Information

Please refer to the individual licenses of the source datasets used to create this corpus, as listed in the "Source Data" section above. The combined dataset does not have a single unified license, and users should ensure compliance with the terms of each source dataset when utilizing this corpus.

Citation Information

@misc{shiza2026lilmoo,
      title={{Raising Bars, Not Parameters: LilMoo Compact Language Model for Hindi}}, 
      author={Shiza Fatimah and Aniket Sen and Sophia Falk and Florian Mai and Lucie Flek and Nicholas Kluge Corr{\^e}a},
      year={2026},
      eprint={2603.03508},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2603.03508}, 
}

Acknowledgments

Polyglot is a project funded by the Federal Ministry of Education and Research (BMBF) and the Ministry of Culture and Science of the State of North Rhine-Westphalia (MWK) as part of TRA Sustainable Futures (University of Bonn) and the Excellence Strategy of the federal and state governments.

We also gratefully acknowledge the granted access to the Marvin cluster hosted by University of Bonn along with the support provided by its High Performance Computing & Analytics Lab.

Contributions

If you want to contribute, contact us at polyglot@uni-bonn.de!

Downloads last month
201

Models trained or fine-tuned on Polygl0t/gigalekh-v1

Collection including Polygl0t/gigalekh-v1

Paper for Polygl0t/gigalekh-v1