Timestamps: 00:00:00 Advanced Preprocessing 00:00:18 Part-of-Speech (PoS) Tagging 00:01:06 Uses of PoS tags 00:02:24 Named Entity Recognition (NER) 00:03:17 Uses of NER tags 00:04:08 The challenges of NER 00:04:57 PoS- and NER-tagging as sequence labelling tasks 00:07:30 Constituency parsing 00:10:24 Dependency parsing 00:12:22 Uses of parsing 00:13:46 Which parsing approach to use 00:14:14 DEMO: advanced preprocessing with spaCy 00:21:11 Preprocessing recap
Great video (and great course). Thank you, thank you, thank you. Nitpick at 4:25 Hamilton was never a president. He was "the ten-dollar founding father without a father", but was never a president LOL
Amazing video. Thank you so much for these videos. I hope you will make more videos on other concepts in deep learning. By the way I do think you teaching this has made your speech on point as a by product. I too had doubts like others that your voice might have been generated but I picked up some words you say, that sounded Canadian, am I right hahah. Thank you and look forward to more videos!
Hello Sir! Your course is great! Please suggest us a course on Generative AI to easily learn the concepts like we are learning in this course! Very grateful Sir! Thank you!
@@futuremojo it’s so clean, no disturbance nothing and fount to be bit artificial. That’s why I got this doubt. btw your explanatory skills are amazing. Hope you make more technical videos on LLM , finetuning or on transfer learning.
Why is there no 'end of entity' tag? I'm sure some might say it's redundant and unnecessary because when you come to the 'o' you are *obviously* at the end of the entity. But it is just as possible that the 'o' is a mistake, especially in a long multi word name. An end tag would be both more explicit and eliminate any ambiguity. But maybe that's just me....
It depends on what you want to do. The transformer is a model architecture. It's not something that automatically takes care of NLP tasks for you end to end. When you load a transformer-based model under the hood using a library from Hugging Face, the library itself is taking care of things like tokenization (and you can customize what tokenizer it uses). When you want to use an LLM to embed a document, you might need to do certain types of preprocessing depending on the LLM's context length or what you're trying to accomplish (e.g. you might need to enrich your data).
@@futuremojo Thank you so much. when I learned the hugging face, it skips a lot of concepts you mentioned. After you explained, I understand now because HF provide the functions to take care of these in their tokenization process.
Thanks for the comment. I think it's just the search patterns. The three most-viewed videos in this series are: - The first introduction video. This is probably people looking for introductions to NLP. - Neural networks from scratch. This is probably because of deep learning hype and people wanting to learn fundamentals. - Transformers. Self-explanatory. None of these are surprising. I think the people most interested in NLP are the ones who go over the rest. If you look at the most popular NLP-related videos now, they're about prompt engineering, Langchain, etc. Whatever's current and applied rather than foundational.
Timestamps:
00:00:00 Advanced Preprocessing
00:00:18 Part-of-Speech (PoS) Tagging
00:01:06 Uses of PoS tags
00:02:24 Named Entity Recognition (NER)
00:03:17 Uses of NER tags
00:04:08 The challenges of NER
00:04:57 PoS- and NER-tagging as sequence labelling tasks
00:07:30 Constituency parsing
00:10:24 Dependency parsing
00:12:22 Uses of parsing
00:13:46 Which parsing approach to use
00:14:14 DEMO: advanced preprocessing with spaCy
00:21:11 Preprocessing recap
Watched lots of content on youtube for NLP. But this series is the BEST. Concise and to -the- point!
Your series is a life-saver. You an EXCELLENT teacher, and your voice is radio-like. Thanks so much!
Thank you, Davy!
This series is incredible! I can't believe we get to access such content for free online... what an era
This series is excellent! I'm glad I came across this series while searching for TF-IDF & cosine similarity related materials.
the best NLP course in youtube at least to me!! Thanks so much.
With your Great NLP lectures, understand a lot of NLP concepts
Great lectures.
Great video (and great course). Thank you, thank you, thank you.
Nitpick at 4:25
Hamilton was never a president. He was "the ten-dollar founding father without a father", but was never a president
LOL
OMFG of all the things I didn't google LOL
Amazing video. Thank you so much for these videos. I hope you will make more videos on other concepts in deep learning. By the way I do think you teaching this has made your speech on point as a by product. I too had doubts like others that your voice might have been generated but I picked up some words you say, that sounded Canadian, am I right hahah. Thank you and look forward to more videos!
LMAO which words gave me away?
Words like about haha. And words containing "O". Maybe I've been watching Robin from HIMYM too much lol.
Hello Sir!
Your course is great!
Please suggest us a course on Generative AI to easily learn the concepts like we are learning in this course!
Very grateful Sir!
Thank you!
have u find any good course> bro
Which software u used to generate that voice?
What makes you think it's software?
@@futuremojo it’s so clean, no disturbance nothing and fount to be bit artificial. That’s why I got this doubt. btw your explanatory skills are amazing. Hope you make more technical videos on LLM , finetuning or on transfer learning.
Why is there no 'end of entity' tag? I'm sure some might say it's redundant and unnecessary because when you come to the 'o' you are *obviously* at the end of the entity. But it is just as possible that the 'o' is a mistake, especially in a long multi word name. An end tag would be both more explicit and eliminate any ambiguity. But maybe that's just me....
After transformer is used, are these preprocessing steps still very useful?
It depends on what you want to do. The transformer is a model architecture. It's not something that automatically takes care of NLP tasks for you end to end.
When you load a transformer-based model under the hood using a library from Hugging Face, the library itself is taking care of things like tokenization (and you can customize what tokenizer it uses).
When you want to use an LLM to embed a document, you might need to do certain types of preprocessing depending on the LLM's context length or what you're trying to accomplish (e.g. you might need to enrich your data).
@@futuremojo Thank you so much. when I learned the hugging face, it skips a lot of concepts you mentioned. After you explained, I understand now because HF provide the functions to take care of these in their tokenization process.
Machine learning is a niche topic, sure, but how does this have so few views?
Thanks for the comment. I think it's just the search patterns. The three most-viewed videos in this series are:
- The first introduction video. This is probably people looking for introductions to NLP.
- Neural networks from scratch. This is probably because of deep learning hype and people wanting to learn fundamentals.
- Transformers. Self-explanatory.
None of these are surprising. I think the people most interested in NLP are the ones who go over the rest.
If you look at the most popular NLP-related videos now, they're about prompt engineering, Langchain, etc. Whatever's current and applied rather than foundational.
I have no idea! It beats me as well! The views on these courses should be running into millions!