preslav nakov - the web as a training set part 3

41
1 The Future

Upload: datasciencesociety

Post on 12-Feb-2017

171 views

Category:

Data & Analytics


0 download

TRANSCRIPT

Page 1: Preslav Nakov - The Web as a Training Set Part 3

The Future

Page 2: Preslav Nakov - The Web as a Training Set Part 3

2

The Big Dream (once again)

Dave Bowman: “Open the pod bay doors, HAL”

HAL 9000: “I’m sorry Dave. I’m afraid I can’t do that.”

Page 3: Preslav Nakov - The Web as a Training Set Part 3

3

Two Important Directions We Touched

•Semantics

•Machine Translation

Critical for the overalladvancement of the field

Practical, within the reachof current technology

Page 4: Preslav Nakov - The Web as a Training Set Part 3

4

Two Important Directions We Touched

•Semantics

•Machine Translation

Page 5: Preslav Nakov - The Web as a Training Set Part 3

5

Semantics: Revolution is Needed?•If we want the dream come true, we should

- not rely on superficial statistics alone- need to get to the meaning of text

•A revolution in semantics is needed- looking at words is not enough- we need better models for

o multi-word expressions (~70% of terminology)o semantic relations (meaning is in the links!)

•The revolution will be supported by- Web-scale corpora- linguistic knowledge

“Moving Lexical Semanticsfrom Alchemy to Science”

Discussion on [Corpora-List]

• This is what Chomsky has done with syntax.• Should we expect the same for lexical semantics?

Page 6: Preslav Nakov - The Web as a Training Set Part 3

6

NAACL’2015: Accept/Reject by Area

SemanticsNLP for Web and Social Media and Social Sciences

Machine TranslationInformation Extraction and Question Answering

Tagging and Chunking and Syntax and ParsingMachine Learning for NLP

Generation and SummarizationLanguage Resources and Evaluation

Text Categorization and Topic ModelsSentiment Analysis and Opinion Mining

Phonology and Morphology and Word SegmentationSpoken Language Processing

Discourse and PragmaticsNLP-enabled Technology

Linguistic and Psycholinguistic Aspects of CLDialogue and Interactive Systems

Information RetrievalLanguage and Vision

0 10 20 30 40 50 60 70 80 90 100

Semantics has emerged from a marginal to a dominant position.

Page 7: Preslav Nakov - The Web as a Training Set Part 3

7

Two Important Directions We Have Touched

•Semantics

•Machine Translation

Page 8: Preslav Nakov - The Web as a Training Set Part 3

8

Machine Translation: Revolution?•Revolution?

- Two great revolutions so faro1993: statistical word-based translationo2003: statistical phrase-based translation

Page 9: Preslav Nakov - The Web as a Training Set Part 3

9

Machine Translation: Revolution?•Revolution?

- Two great revolutions so faro1993: statistical word-based translationo2003: statistical phrase-based translation

- Overdue for the next revolution?o2013: ???

• Syntactic translation?• Semantic translation?

SOURCE TARGET

words words

syntax syntax

semantics semantics

interlingua

phrases phrases

Page 10: Preslav Nakov - The Web as a Training Set Part 3

10

Machine Translation: Revolution?•Revolution?

- Two great revolutions so faro1993: statistical word-based translationo2003: statistical phrase-based translation

- Overdue for the next revolution?o2014: revolution in progress?

Deep neural networks – the new revolution:• Speech recognition• Machine translation• Semantics

BUT can they:- scale to the Web?- model linguistic structure- handle MWEs- use linguistic knowledge

Page 11: Preslav Nakov - The Web as a Training Set Part 3

11

The Future?

Three words: Web, semantics, linguistics

and deep neural networks?

Page 12: Preslav Nakov - The Web as a Training Set Part 3

12

The Futurev. 2.0

Page 13: Preslav Nakov - The Web as a Training Set Part 3

13

Human or Computer?

Page 14: Preslav Nakov - The Web as a Training Set Part 3

14

Human or Computer?

Page 15: Preslav Nakov - The Web as a Training Set Part 3

15

Human or Computer?

Page 16: Preslav Nakov - The Web as a Training Set Part 3

16

Human or Computer?

Page 17: Preslav Nakov - The Web as a Training Set Part 3

17

Human or Computer?

Page 18: Preslav Nakov - The Web as a Training Set Part 3

18

Human or Computer?

Page 19: Preslav Nakov - The Web as a Training Set Part 3

19

Human or Computer?

Page 20: Preslav Nakov - The Web as a Training Set Part 3

20

Human or Computer?

Page 21: Preslav Nakov - The Web as a Training Set Part 3

21

Human or Computer?

Page 22: Preslav Nakov - The Web as a Training Set Part 3

22

Human or Computer?

Page 23: Preslav Nakov - The Web as a Training Set Part 3

23

Human or Computer?

Page 24: Preslav Nakov - The Web as a Training Set Part 3

24

Human or Computer?

Page 25: Preslav Nakov - The Web as a Training Set Part 3

25

Human or Computer?

Page 26: Preslav Nakov - The Web as a Training Set Part 3

26

Human or Computer?

Page 27: Preslav Nakov - The Web as a Training Set Part 3

27

Human or Computer?

Page 28: Preslav Nakov - The Web as a Training Set Part 3

28

Human or Computer?

Page 29: Preslav Nakov - The Web as a Training Set Part 3

29

The Future is Now?•Books

- Algorithm by Philip Parker, Inseado 1,000,000+ books generatedo 100,000+ being sold at Amazon

•Robo journalism- tons of articles generated today- by 2025, can cover 90%

•What is next- Fake reviews?- Computers mining text written by other computers?- From Computational to Computer Linguistics?- …

Page 30: Preslav Nakov - The Web as a Training Set Part 3

30

The Futurev. 3.0

Page 31: Preslav Nakov - The Web as a Training Set Part 3

31

Hiroshima & Nagasaki•26/07/1945: Potsdam declaration was an ultimatum to Japan: capitulate or face “prompt and utter destruction”

•Japan’s prime minister Kantaro Suzuki at a press-conference: “No comment. We keep discussing.”

•He used the word mokusatsu, which can mean (a) no comment, or (b) we reject.

•10 days later…

Page 32: Preslav Nakov - The Web as a Training Set Part 3

32

The Future?Will the next nuclear war start because of a computer translation?

Page 33: Preslav Nakov - The Web as a Training Set Part 3

33

Moore’s Law: in 10 years…

Can NNs scale?

Page 34: Preslav Nakov - The Web as a Training Set Part 3

34

In 17 years: More Robots than Humans!

Page 35: Preslav Nakov - The Web as a Training Set Part 3

35

Maybe you do not believe it...

Page 36: Preslav Nakov - The Web as a Training Set Part 3

36

Page 37: Preslav Nakov - The Web as a Training Set Part 3

37

Artificial Intelligence as a Threat

“The development of full artificial intelligence could spell the end of the human race.”

Stephen Hawking

“I am in the camp that is concerned about super intelligence. ...and don't understand why some people are not concerned..”

Bill Gates

“I think we should be very careful about artificial intelligence. If I had to guess at what our biggest existential threat is, it’s probably that.”

Elon Musk (donated $10 million to Future of Life Institute)

Page 38: Preslav Nakov - The Web as a Training Set Part 3

38

The Future: SkyNet?

Page 39: Preslav Nakov - The Web as a Training Set Part 3

39

The Future?

Page 40: Preslav Nakov - The Web as a Training Set Part 3

40

The Future?

Page 41: Preslav Nakov - The Web as a Training Set Part 3

41