Natural Language Processing at Hike messaging app

The need for NLP at Hike is driven by two sets of applications: - A key focus area is around developing contextual, conversational systems - the byte-sized content app - apply NLP to make sense of the content and tag it


One of the most popular features on Hike is the stickers. More than 20k stickers in over 40 languages in Hike, it’s a great way to say about anything. But when you get many stickers you have a problem with the selection, right? Finding the right sticker when you need it is quite difficult. Hike suggest stickers based on the current context of the chating.

For recommendations, Conversational text is noisy since users don’t care about spellings or grammar. There are many exaggerated forms of expressing something when it comes to chatting. users also chat using a mix of Indian languages with an English word in a transliterated format.

Hike write “At Hike, we adhere to very high standards when it comes to protecting users’ data. The data used to train these conversational models. that anonymize on the user’s device by stripping off any identifiable information. This helps us safeguard against the risk of leaking user data to malicious parties.”

Content tagging for apps such as Hike News is another application ripe for NLP. Hike working towards problems such as

  • classifying articles.
  • detect key entities that are the reference in an article.
  • finding duplicate articles in the content.

All this requires developing a unique “Named Entity Recognition”. for Indian languages, multilingual embedding, etc.


Please enter your comment!
Please enter your name here