How

[@pratyushmittal9473]

This is fabulous. Thanks a lot for recording this. Notes: - Give corpus to Byte-Pair encoding to get tokens - Convert tokens to vec using Word2Vec - Train neural network on above dataset using Bidirectional LSTM (with crf) - Use "sequence tagging" for named entity recognition

[@damercy]

I love how Arpit simplifies complex concepts. Truly a great engineer. Hope I can learn to become like him. Thank you for these.

[@mryaswanth1078]

So Insightful! Thanks buddy

[@HarshSingh-hk8fe]

Wow!, so much info in this short video really man you're great

[@as_if]

So vectorization is done to maintain the semantics, one of our greatest power as human.

[@jithinb7047]

Awesome content! Thanks Arpit.

[@architshukla8076]

Very Informative Content . Thanks a lot Arpit !!

[@shishirchaurasiya7374]

This is dayummm....awesome best explaination

[@krum.00]

While the general idea has been covered, any discussion on modern NLP is incomplete without transfer learning and transformers. For most practical purpose word2vec and lstms are dead, expect for rare cases where your scale and latency req. demands it, like in the case of gmail smart compose.

[@bytesizedbraincog]

Typesense is yet another good production ready package on the lines of elastic search and solar.

[@skanderbegvictor6487]

why not combine this with BM25?

[@grx4.067]

questions how can i handle searches in which there are some spelling mistakes

[@Paaaaaaaaaaal]

Next video : How Zomato fired top college grads just after 4 months of joining ☹️

[@kalyanben10]

How Zomato Scaled its Loss... and then scaled firing

[@socialmedia5173]

Zomato is founded to make money not to give jobs, have they promised u that they will not fire 🤷