Disclaimer: this is going to be a stupidly obvious question for anyone who speaks either Hindi or Bengali, but with zero knowledge of either language I can't really eke out an answer. Apologies if this comes across as a silly question.

Rabindranath Tagore's poem "This Dog" on AllPoetry.com is given in two languages, English and (supposedly) Hindi. But I thought that Tagore wrote in Bengali and not Hindi. Is this Hindi text also a translation, or is it not really Hindi at all?

Putting the "Hindi" version into Google Translate, it auto-recognises the text as Bengali. Trying Translated Labs' Language Identifier tool, it auto-recognises the text as Hindi. So I'm still confused.

What language was this poem originally published in? What language is the text on the AllPoetry website?

(Exact publication details would also be interesting, to see when and where this was originally published as well as in what language, but not strictly necessary to get the accepted answer.)


1 Answer 1



It's not Hindi at all. It's Bengali. Those languages are as different as Spanish and Italian. They don't even look the same, because they use very distinct, albeit related, scripts.

Bengali, which along with Assamese and Meitei uses the Bengali-Assamese script (duh), looks like this:

প্রত্যহ প্রভাতকালে ভক্ত এ কুকুর
স্তব্ধ হয়ে বসে থাকে আসনের কাছে
যতক্ষণে সঙ্গ তার না করি স্বীকার
করস্পর্শ দিয়ে ।

Hindi on the other hand uses the Devanagari (aka Nagari) script, as do Marathi, Nepali, Sanskrit, and some flavors of Konkani. Devanagari looks like this:

प्रत्यह प्रभातकाले भक्त ए कुकुर
स्तब्ध हये बसे थाके आसनेर काछे
यतक्षणे सञ्ग तार ना करि स्वीकार
करस्पर्श दिये ।

The Bengali lines are the first four lines of Tagore's poem, of course. The Devanagari lines are simply a letter-for-letter transliteration of the Bengali. The Bengali and Devanagari scripts are pretty much exactly equivalent; each letter in the one has a counterpart in the other, so transliterating between the two is a cinch.

Note that this is a transliteration, not a translation. The lines would not be meaningful in any of the languages that use the Devanagari script. Some of the words would be, either because of the Sanskrit-derived vocabulary common to many Indic languages (e.g., ভক্ত / भक्त / bhakta, which means a devotee) or by happenstance (e.g., তার / तार / taar, which means "his" in Bengali but also means "metal string" or "wire" in pretty much any Indic language; a Hindi speaker looking at the above transliteration would misrecognize तार as "wire").

In iTrans, the lines would be rendered as follows:

pratyaha prabhaatkaale bhakta e kukur
stabdha haye vase thane aasaner kaachhe
yatakShaNe sa~nga taar naa kari sviikaar
karasparSha diye

A literal translation would be:

Every dawn this devoted dog
silently stays seated by [my] seat
for as long as I don't accept his company
by patting him.

The poem was published in Tagore's 1941 Bengali collection আরোগ্য (आरोग्य / aarogya, meaning Convalescence). Tagore wrote original works in both Bengali and English. He also wrote a cycle of 22 poems, ভানুসিংহ ঠাকুরের পদাবলী / bhaanusi~Nh Thakurer padaavalii, in a pastiche of Brajabuli, imitating the Vaishava Padavali tradition that flourished in the 15th through the 17th C. He frequently translated his own Bengali poems, stories, etc. into English as well. He did not, however, write in Hindi. Many of his works have been translated into Hindi by other hands, but cursory Googling didn't turn up a Hindi translation of this poem.

I don't know why Translated Labs and AllPoetry think this is Hindi. Or at least I do: because ethnocentrism. But that's an explanation, not an excuse. Considering that Hindi and Bengali are each spoken by literally hundreds of millions of people, and are both among the top ten languages spoken worldwide, it's a pretty shocking mistake. The world is simply jam-packed with speakers of those languages, and those websites can't find a single speaker of either to verify which one this is?

It's mind-boggling that the web should be propagating such an indefensible error, and it's good to be able to clear it up. Thanks for this question and asking me to answer it.

Obligatory disclaimer: My Bengali is dreadful, and the translation of the four lines given above probably has deathless howlers in it, so it's strictly AYOR and corrections are welcomed.

Tangentially relevant funny story: my late husband had quite the gift for languages. He began learning Hindi in his forties. He had just mastered the Devanagari script when he and I went to watch Satyajit Ray's মহানগর (Mahanagar, The Big City), at an arthouse cinema. The titles came on and I noticed he was peering at the screen with an increasingly puzzled expression. I leaned over and whispered "It's Bengali." A look of relief came over his face as he remembered that Bengali uses a completely different script from Hindi.

  • 2
    Well there goes Translated Labs's credibility. I won't be using that tool again.
    – Rand al'Thor
    Commented Nov 9, 2020 at 6:09
  • That's a bit drastic, yo. I'm sure it's great at correctly identifying languages white people speak.
    – verbose
    Commented Nov 9, 2020 at 6:35
  • Actually, it fails on some European languges, too. It identifies the Basque national anthem as Fijian. Google translate at least identifies it as Basque, although it doesn't do a very good job of translating it.
    – Peter Shor
    Commented Nov 9, 2020 at 13:26
  • 1
    There seem to be a ton of problems with their algorithm. First, I am not sure it even has Bengali, period. I put in a few Bengali texts, and it consistently identified them as Hindi. Curiously, it identified some Bengali words as Nepali, so it apparently has that.
    – Obie 2.0
    Commented Dec 1, 2020 at 2:16
  • 1
    Second, it identifies individual Bengali graphemes as Persian (ল), Arabic (e.g. অ), Hebrew (ঔ), Russian (ঋ), Ukrainian (র) and so forth. This suggests they are using an approach which, while interesting from a programming perspective, is disastrous from a user interface perspective: recognition of characters based on their forms. When the system cannot find a perfect match in its library of characters, it just guesses and returns the best guess. This probably explains why Bengali texts, but not characters, are identified as Hindi: the joining line over the characters.
    – Obie 2.0
    Commented Dec 1, 2020 at 2:23

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.