Natural Language Processing: Crash Course Computer Science #36
HTML-код
- Опубликовано: 10 сен 2024
- Today we’re going to talk about how computers understand speech and speak themselves. As computers play an increasing role in our daily lives there has been an growing demand for voice user interfaces, but speech is also terribly complicated. Vocabularies are diverse, sentence structures can often dictate the meaning of certain words, and computers also have to deal with accents, mispronunciations, and many common linguistic faux pas. The field of Natural Language Processing, or NLP, attempts to solve these problems, with a number of techniques we’ll discuss today. And even though our virtual assistants like Siri, Alexa, Google Home, Bixby, and Cortana have come a long way from the first speech processing and synthesis models, there is still much room for improvement.
Produced in collaboration with PBS Digital Studios: / pbsdigitalstudios
Want to know more about Carrie Anne?
about.me/carri...
The Latest from PBS Digital Studios: • All PBS Digital Studio...
Want to find Crash Course elsewhere on the internet?
Facebook - / youtubecrash. .
Twitter - / thecrashcourse
Tumblr - / thecrashcourse
Support Crash Course on Patreon: / crashcourse
CC Kids: / crashcoursekids
Hi, I'm a professor of computational linguistics in Germany. I just wanted to say that your video is one of the best explanations of natural language processing for non-experts that I've ever seen. I'm really impressed - great job!
I'm a computer science student and i really need some help as i'm gonna present on this topic and needed someone to guide me through. If you can help please reply.
Please what is the difference between computational y and NLP
I agree, This needs to be the introductory course for CS majors starting their undergrad.
I had to take an introductory course at my university and the instructor told us to write an essay on computer science personalities; Grace Hopper and the like. While these personalities are definitely important people and deserve the respect. Some assignments were about choosing a preferred path in CS without giving any background on these paths. I think that the course neglected to give any introduction to computer science.
@@379dhairyaashara4 Hi I am in the same boat. Did u get any information that was helpful?
This is how computer science should be taught. The motivation and big picture is important.
I think this should be the introductory course for CS majors.
"It wasn't evil, it was efficient" might be the most terrifiying sentence I've ever heard in my life.
exactly. killing off humans will also be 'efficient'. not sure I'd enjoy that though.
It was music to my ears. If it works, it's not [insert negative adjective here].
That's because you do not understand what happened.
Framed around explainability, I'd agree. The bots diverged to using a language that humans couldn't understand, making it opaque and therefore unethically unexplainable.
INTJ
Keep calm and Carrie Ann.
Noah Browne haha
This series is so fantastic. I get super excited to hear real-world examples and helps me envision the possibilities in the future.
Will the foreshadowing of Crash Course Linguistics ever cease?!
I can't believe I've watched all 36 of these. Keep up the good computer science.
One minor issue is that with speech interfaces becoming more common and more complex, some people will over estimate how human they are and get annoyed when it doesn't understand what the user thinks is a perfectly simple request.
angeldude101 exactly me lmao
"SIRI, I TOLD YOU I WANT SOME BUFFALO WINGS, WHY ARE YOU SENDING ME PLANE TICKETS TO NEW YORK?!"
Aw yiss, this is my jam! Thanks for this video - I’m in grad school working on an MS in Computational Linguistics, so NLP is what I do every day. I actually just finished writing a simple part of speech tagger the other day!
that seems like a lot fun. I also enjoy both cs and linguistics, so that seems like a fun career path.
I'm a computer science student and in need of some help for my last year's project which includes this topic and needed someone to guide me through. If you can help please reply.
I'm 55. It has been "pretty soon" that voice recognition would work properly for most of my life. It is now in common use - and it still does not work. Perhaps if I live to a 110, but I will take no bets on that.
I SPEAK HUMAN LANGUAGE TOO, FELLOW HUMAN.
I also pose no danger to my fellow humans, (I think we are getting away with it!)
WE POSE NO DANGER TO THE HUMAN POSE NO DANGER TO THE HUMAN POSE NO DANGER TO THE HUMAN POSE NO DANGER TO THE HUMAN POSE NO DANGER TO THE HUMAN POSE NO DANGER TO THE HUMAN POSE NO DANGER TO THEsegmentation fault
Reckless Roges WHY ARE YOU YELLING
ARE YOU PERHAPS A ROBOT
Oh my gosh, this is the best channel I've subscribed to in a long time (on par with SmarterEveryDay)! Kudos!
I remember watching this 5 years ago and slowly realising how important it would be for LLM AI and how the use of language is just another level of abstraction for the reality we live in.
Elisa was the starting point of doctoral interest in natural language for education. We've come a long way from Xerox Sigma 7s and PDP-11s working through teletypes and VDTs. I think we are about ready to start doing the natural language teaching that I dreamed about all those decades ago.
I love how you gave an example with Siri! Thank you for all your work! You are amazing.
Wait for it...
The Mongols!
Dennis E They're the exception!
One thing to keep in mind with regard to natural language is that natural language is often highly imprecise, and very dependant upon the level of eloquence of the transmitter and the knowledge and understanding of the receiver, and the importance of shared points of reference for both of them.
This is why for instance when it comes to the sciences there is such a heavy emphasis on exact word use. When writing a scientific article or instruction manual, you want to be as precise as possible with your use of words, terms and grammer and to minimize the chance of missunderanding to be as small as possible.
That is why, even if computers could understand natural language, the degrees of complexity of the tasks which they could pull off would inevitably be affected by the accuracy of the instructions(macros and preprogrammed sequences and programs are of course exempted).
thank you for the episode! i knew a lot of it, but this video managed to structure everything so neatly that i feel like i understand the topic so much better now :>
Try turning on automatic captions. For clearly spoken stuff like what is on this channel, they are super accurate now.
Awesome breakdown of the current state of NLP.
It would be great to have a similar series on implications of technical progress with philosophers and social studies experts commenting on the relationship between the tech and our society.
Hey I've beenn taking an NLP class this term.
Its hard.
*edit* This video was not really about what my class was about if you're wondering how its hard. We covered types & tokens, Part of Speech tagging, word senses, Viterbi algorithm, Naive Bayes, Lesk Algorithm, Co-occurence matrices, and briefly near the end Skip-gram and CBOW embeddings systems. Among a number of other things that I've completely forgotten about.
This class was difficult and I also screwed up, missing 3 classes being sick and not focusing enough time to do assignments. The entire course has been a nightmare and I know nothinng from it. Last class is Decemmber 7th, I will fail.
How about a crash course electrical engineering?
thanks
The best part of this video is when she spoke in Shakespearean to Siri. Who here can also fluently speak this language? I know I can.
She speaks faster than my internet connection
haha
lol
I thought my speed settings were not normal so had to check to reconfirm .. on the positive side!
I really like the pace its a crash course
Came in to watch the video... Staying to watch the series 🎉
Now the ChatGPT is dominating this field.
Your enthusiasm is awesome.
When she said this 3:34 to Siri, she didn't know what to say.
5 years later, I wrote what she said to ChatGPT and this is what I got:
Prompt: "Hey ChatGPT, methinks the mongols doth roam too much, what think ye on this most gentle mid-summer's day?"
Answer:
As an artificial intelligence, I don't have personal opinions or feelings. However, I can provide information about the Mongols and their history of conquest and expansion.
The Mongols were a nomadic people who lived in the steppes of Central Asia, and they were known for their military prowess and ability to conquer and control large territories. Under the leadership of Genghis Khan, they established a vast empire that stretched from Europe to Asia and included parts of modern-day Russia, China, and the Middle East. The Mongols were skilled horsemen and fighters, and they used a variety of tactics, including surprise attacks and psychological warfare, to defeat their enemies.
After the death of Genghis Khan, the Mongol Empire was divided among his descendants, and the different Mongol states continued to expand and conquer new territories. The Mongols were known for their mobility and their ability to adapt to different environments, which allowed them to roam and conquer many different areas.
While the Mongols were often successful in their conquests, their rule was not always welcomed by the conquered peoples. They were known for their strict and often brutal methods of governance, and many of the areas they conquered experienced significant loss of life and property.
Absolutely fantastic video. A great overview of the topic making it easy for me to learn more about the parts that interest me. I will be watching more of these this weekend
Yay, an episode on the topic of my thesis!
Koala: eats shoots and leaves. Eats, shoots, and leaves.
...and, finally, speaking of accuracy, linguists tell us that linguistics, the scientific study of language-in-general, is not an exact science, and yet, linguistics is what we use to convey the understanding of exact science-(is that a paradox)-and furthermore will correct the usage of natural language only when we stop monkeying-around for example, NOT-ELSE.
This videos are just so awesome!!!, great job.
big fan of Crash course. great job guys.
Buffalo buffalo Buffalo buffalo buffalo buffalo Buffalo buffalo
Police police police police police police, who in turn police police.
Crash blossoms.
My favorite is:
"Wouldn't the sentence 'I want to put a hyphen between the words Fish and And and And and Chips in my Fish-And-Chips sign' have been clearer if quotation marks had been placed before Fish, and between Fish and and, and and and And, and And and and, and and and And, and And and and, and and and Chips, as well as after Chips?"
I recall on my TRS-80 Model 1, Level 2 with EI and disk - I had a speech synthesizer hooked up and a speech recognition module. That was great fun - learned about phonemes through that.
I recently saw that clip from the Bell Labs speech generation system in the movie "Gizmo!" from 1977. HIGHLY recommended for fans of obscure technology that didn't quite make it, and random feats of badassery. And great music. Let's not forget the great music.
Thank you, I always wanted to know more about this topic 👍
What bothers me about those "speech recognizers" is. Most of the time we humans don't say things like "I am going to ...". But talk more like this "Am goingto" (yes, I exaggerated a bit). So, we combine two words, because you talk faster. Or we even skip words. But the computer is programmed to recognize each word individually. Which isn't the way humans naturally speak.
There is even Vocaloid, a speech synthesizer software by Yamaha aimed to singing. There is a whole culture aroind it because the companies that made the voice banks have created anime-style characters for them.
Quite comprehensive!
A video on
1) firmware, drivers, microcontroller, daq and PLD
2) database
Would be very useful and complete...
This makes me want a crash course linguistics so bad.
Brilliant! So clearly explained. Well done!
Very interesting. Another outstanding lesson. Thanks!
Teresa White Very true
My TA is preparing PhD on NLP things which brought me to this channel an' boy I'm glad i found it
Bezos, Brin, Page, and Zuckerburg really want to know what you have to say so they can... make our lives better ?
I remember downloading a chat bot called Daisy back in the day...very simple, but if you spoke to it in a consistent manor, and did it a LOT, it would spit out some fairly interesting things at times...I think still have some of the conversations saved somewhere.
What were some interesting things it said?
I dont quite remember anymore, this wouldve been in the early 2000s...I'll see if I can find the conversations.
R u sure that wasnt an actual person. . .?
lol yep
Still looking for the old conversations...not being able to find them makes me want to find them even more...
did you ever find those
Damn, Computer Science is so exciting.
Damn, she speaks faaaaaast !!
Little correction: Phonems are from the realm of Phonology, not Phonetics. The corresponding unit from Phonetics is a Phon. The distinction is important, because Phons are a physics based unit while Phonems are a mind based unit. One Phonem can contain many multiple Phons, which are then referred to as Allophones.
Perhaps it's an acceptable variation, but I would have typed "phoneme", "phonemes", and "phones".
Ah, good catch. You are fully correct, the mistake stems from my being a native german speaker. Thank you for your correction.
Bitte shoen. "Akkusativ", "Dativ", und "Nominativ" gefallt mir.
The extra "e" is rather pesky to me but it is a bit amusing to spell my native language since there is almost a method to the madness.
Excellent video!
thats really informative ...thank you sooo much ..keep it up ...
WHAT THINK YE ON THIS GENTLE MIDSUMMER'S DAY? >intense concentration< maybe my favorite part in the entire series so far
Crash Course - English Grammar!@!
Next series please.
"It wasn't evil; it was *efficient*."
Do you realize how ominous that sounds? I'm not talking about AI alignment, which makes that sort of sentence actually worrying, just about how it sounds when you say the sentence out loud.
I wonder if there are some academic papers that introducing these concepts? I hope wish I can find good citation for my dissertation, thank you!
How fitting. The ad for this video was for the Echo.
'Blabbing on about'. lol I like kerryan choice pf words. :)
Probably the best channel/playlist that I have encountered on youtube till date apart from 3blue1brown.
Carrie Ann, Carry on!
No mention of Watson (the Jeopardy!-playing computer that beat Ken Jennings and Brad Rutter) in a video about computers that understand human languages? What the heck?
Great introduction to a very complex area!
I am a linguist but I know nothing about computer science and NLP but I am interested to get a job in this field. Do you have any recommendations on how can I start my learning? What should I study to become more attractive towards employers?
Your classes are amazing...
Hmm, I never looked into the other parts of this channel, mostly just history and astromony. I must say I'm actually glad i see this, reasons being my desire to be an astro phycsist to have nasa send me to mars, and away from my crazy ex.
Brilliant
Omg like x1000. Great coverage girl...
Very impressive young lady, Thks
I mean all I got was the surveillance. Then again, as a tool this is so cool.
Methinks thou art more wonderful than a summer's day. Long ago I created a 3D CGI puppet, using phonemes as the anchor points in speech as they correspond to the shapes our mouths form when speaking. Being able to detect phonemes in speech would have been a massive timesaver. I supposed that's what is going on with animojies?
Nope. Animojis use the phone's 3D tracking camera to map points on your face onto the 3D model. So it's just copying your 3D expression, not basing it on sound at all.
Fascinating, thank you.
I had to make a chatbot for my cs class, so this topic is good timing.
catsaur if you didn't know about neutral networks before this video I doubt you will need to implement one to pass your cs class. A set of if statements would probably suffice :)
One of the simpler ways to make a "dumb" chatbot is to give it a small set of default "conversation starters" to randomly choose from. Whenever a user says something it doesn't know about, it responds "I didn't understand that", but adds what the user said to its conversation starters. Whenever it chooses a conversation starter, it logs what the user responds as an "appropriate response" and now it knows how to respond to one more possible input :)
Of course, even for a "dumb bot", you may still want to do some input normalization to get a more generic set of matched inputs for responses, but that's up to you :)
I'm amazed by the Mac in the background.
Dear Siri, you are not alone. I am not sure I got that too :)
If I do a voice search I do it the same way I do a text search. I don't bother with sentences.
And yet I've still never seen a web-based version of Eliza or any other Turing test contender that works using HTML speech recognition and generation. Anyone know of any?
I
WE NEED CRASHCOURSE ENGLISH LANGUAGE!!!!!! ASAP!!!!!!
I actually work with mainframes and I daily program in a language called Natural! Does anyone else work with it around here?
This is awesome you rock comp sci goddess.
great stuff!!
thanks for this information I will take a 10 in my class of english
GREAT.
Talk nerdy to me
Does anyone else find it annoying that to take advantage of these awesome technologies you have to send your voice data off to some server somewhere? I wish that there were local speech recognition solutions (that actually compiled (looking at you, AUR version of Simon))
Those wouldn't be very accurate; but more importantly, if it was all local, the global systems couldn't learn from your interactions, and overall technological progress of them would slow down. Which would suck.
but can a robot Ramble (which I define as "speaking faster than it can think about what its saying) ? if so, what will happen ?
Random Guy depends on how you define "think". If you write on a mobile phone and always pick the predicted word that could be viewed as the computer "speaking without thinking much" since it doesn't "think" about the context of each predicted word. For example if I let my phone "ramble": I would have to provide a link to the stack of the galaxy 2. As you can see the words almost sound like a normal sentence, however it's a bit "off". However if a computer were to "talk faster than it could think" it would probably only repeat the last known word on the page page page page page page....
true
no human can ramble according to your definition. If you relax it a bit and define it as: speaking unfinished thoughts out loud, which might lead to a dead end then sure.
Daisy, daisy... give me your answer do
I am PG Students from India Linguistics ,but I studied General Linguistics in master degree, can anybody guide me self learning Computational Linguistics course available???? Pls.
apart from the vocabularies concept can you give short description on Entity extraction, aspect of entity extraction , like how can these interrelate to Natural language processing.
Many complain about your speech being to fast, but I have to kindly disagree.
It was all perfect, fast enough and smoothly cut.
It's a crash course after all.
[clicks like before watching]
Good and Interesting!😀😁
the mongols rose from leaves? does it mean mango grow up from tree or mongols came from grassland?confusing...
She WOZ happy
Esperanto would make this all so much easier :)
Hey, that's a copy of Mitnick's "Ghost in the Wires" in the background!
I still ask myself why people give these videos a thumbs down.
To let RUclips know that this video is not their cup-of-tea and to improve the RUclips recommendation algorithm.
Because people are dicks.
*Balls to me to me to me to me, to me to me?*
Dope!