Concretely, What is NLP?

Module 1: Lecture 2

2 / 14

Fundamental and Scientific Goal

Deep understanding of broad language

​

Engineering Goal

Design, implement, and test systems that process natural languages for practical applications

What do we do in NLP?

Concretely, it (was) a collection of tasks:

• Translation, sentiment analysis, retrieval, chatbots, Question Answering, Reading Comprehension

High-level

• NLU, NLG, NLI
• Next token prediction. Next sentence prediction

Mostly from Prof. Pawan Goyal’s Slides

Module 1: Lecture 2

3 / 14

Ambitious goal (achieved?): Good quality translation

What do we do in NLP?

Then (2021?)

Now (2026)

Module 1: Lecture 2

4 / 14

Ambitious goal (achieved?) : Good quality translation

What do we do in NLP?

Then (2021?)

Now (2026)

Well, even humans have made blunders

Module 1: Lecture 2

Pepsi Chinese blunder

“Come alive with the Pepsi Generation”, when translated into Chinese meant, “Pepsi brings your relatives back from the dead.”

​

KFC’s Chinese blunder

KFC’s slogan, “Finger lickin’ good”, when translated into Chinese meant “We’ll eat your fingers off.”

What do we do in NLP?

5 / 14

Module 1: Lecture 2

5 / 14

Well, even humans ...

What do we do in NLP?

Module 1: Lecture 2

6 / 14

Goals can be very ambitious: Open Domain Chatbots

What do we do in NLP?

Module 1: Lecture 2

7 / 14

Goals can be very ambitious: Open Domain Chatbots

What do we do in NLP?

Today’s

Capabilities

are beyond

our wildest

dreams.

- Agents with tools

​

But so are their vulnerabilities

Module 1: Lecture 2

8 / 14

And Goals Can be Practical: Auto Completion

What do we do in NLP?

Module 1: Lecture 2

9 / 14

And Goals can be Practical: Search Engines

What do we do in NLP?

Module 1: Lecture 2

10 / 14

And Goals can be Practical: Information Extraction

What do we do in NLP?

Module 1: Lecture 2

11 / 14

And Goals can be Practical: Domain-specific Chatbots

What do we do in NLP?

And Goals can be Practical: Domain-specific Chatbots

Module 1: Lecture 2

12 / 14

1

¹http://www.news.gatech.edu/2016/05/09/artificial-intelligence-course-creates-ai-teaching-assist

What do we do in NLP?

Module 1: Lecture 2

13 / 14

And Goals can be Practical: Sentiment Analysis

What do we do in NLP?

Module 1: Lecture 2

Other Goals

Spam detection

Machine Translation services on the Web Text Summarization

. . .

What do we do in NLP?

14 / 14

Other Goals

Spam detection

Machine Translation services on the Web Text Summarization

. . .

​

​

Natural Language Technology is perfect sometimes, yet inexplicably imperfect the rest of the times (JAGGED INTELLIGENCE)

Currently so good, that people are promoting ”atoms” vs “bits”.

Any job to do with atoms are here to stay: civil, humans, etc.

Any job to do with manipulating bits and pixels will be automated.

Module 1: Lecture 2

What do we do in NLP?

14 / 14