Machine studying and different synthetic intelligence (AI) strategies have discovered immense success with scientific and technological duties corresponding to predicting how protein molecules fold and recognizing faces in a crowd. Nevertheless, the applying of those strategies to the humanities stays to be absolutely explored.

For instance, what can AI inform us about philosophy and faith? As a place to begin for such exploration, we used deep studying AI strategies to research English translations of the Bhagavad Gita, an historical Hindu textual content initially written in Sanskrit.

Utilizing a deep learning-based language mannequin referred to as BERT, we studied emotion (emotions) and semantics (that means) in translations. Regardless of the large variation in vocabulary and sentence construction, we discovered that the patterns of emotion and that means have been broadly comparable throughout all three.

This analysis opens the way in which for using AI-based applied sciences to match translations and evaluation feelings throughout a variety of classes.

an historical e book of knowledge

The Bhagavad Gita is without doubt one of the central Hindu sacred and philosophical texts. Written greater than 2,000 years in the past, it has been translated into greater than 100 languages ​​and has been of curiosity to Western philosophers because the 18th century.

The 700-verse poem is part of the bigger Mahabharata epic, which recounts the occasions of an historical struggle at Kurukshetra close to fashionable Delhi in India.



Learn extra: Indian philosophy helps us to see clearly, act correctly in an interconnected world


The textual content of the Bhagavad Gita offers with a dialog between the Hindu deity Lord Krishna and a prince named Arjuna. They talk about whether or not a soldier ought to go to struggle for morality and responsibility (or “faith”) if their shut pals or household are on the opposing facet.

The textual content has been instrumental in laying the inspiration of Hinduism. Amongst many different issues, that is the place the philosophy of karma (the non secular concept of trigger and impact) originates.

Students have additionally regarded the Bhagavad Gita as a e book of psychology, administration, management and battle decision.

numerous translations

The primary of a number of English translations of the Bhagavad Gita was revealed in 1785.
Wikimedia Commons, CC BY

There have been numerous English translations of the Bhagavad Gita, however there may be not a lot work to substantiate their high quality. Translations of songs and poems not solely break the rhythm and rhyming patterns, however may also end in lack of semantic data.

In our analysis, we have now used deep studying language fashions to research three chosen translations of the Bhagavad Gita (from Sanskrit to English), with that means and sentiment analyzes that assist in evaluating the standard of translations.

We used a pre-trained language mannequin referred to as BERT, developed by Google. We additional prolonged the mannequin utilizing a human-labeled coaching dataset primarily based on Twitter posts, which seize 10 completely different feelings.

These feelings (optimistic, grateful, empathetic, pessimistic, anxious, unhappy, offended, denial, shocked and mocking) have been tailored from our earlier analysis into social media sentiment throughout the onset of the COVID-19 pandemic.

A schematic diagram exhibiting how AI evaluation works.
Chandra, creator supplied

patterns of emotion

The three translations we studied used very completely different vocabulary and syntax, however the language mannequin acknowledged comparable sentiments within the completely different chapters of the respective translations. In response to our mannequin, optimistic, offended and shocked are probably the most expressed feelings.

Feelings captured by our language mannequin for numerous translations of the Bhagavad Gita.
Chandra and Kulkarni, creator supplied

Moreover, the mannequin confirmed how the general emotion modifications polarity (from unfavorable to optimistic) throughout the interplay between Arjuna and Lord Krishna.

Arjuna is pessimistic in the direction of the start and turns into optimistic as Lord Krishna imparts to him the data of Hindu philosophy. The feelings expressed by Krishna present that with a philosophical data of dharma and recommendation, a troubled thoughts can acquire readability to make the suitable selections in occasions of battle.

One limitation of our mannequin is that it was educated on information from Twitter, so it acknowledges “joke” as a typical emotion. It inappropriately applies this label to components of the Bhagavad Gita. Humor is advanced and strongly culturally constrained, and understanding it’s an excessive amount of to ask of our mannequin at this stage.

As a result of nature of the Sanskrit language, the Bhagavad Gita being a track with rhythm and rhyme, and completely different dates of translations, completely different translators use completely different terminology to explain comparable ideas.

The desk beneath reveals among the most semantically comparable verses from the three translations.

use of sentiment evaluation

Our analysis factors to using AI-based methods to match translations and evaluation feelings throughout a variety of texts.

This method may also be prolonged to evaluation emotions expressed in leisure media. One other potential utility is analyzing films and songs to offer perception to oldsters and authorities in regards to the suitability of content material for youngsters.


The authors wish to acknowledge the invaluable contribution of Venkatesh Kulkarni to this analysis.



Supply hyperlink