1 of 26

Birds of a Feather: Natural Language Processing

2 of 26

3 of 26

4 of 26

5 of 26

6 of 26

7 of 26

8 of 26

9 of 26

10 of 26

What makes language such a difficult problem to solve

  • Languages are constantly evolving�The English Language In 24 Accents

11 of 26

What makes language such a difficult problem to solve

  • Languages are constantly evolving�The English Language In 24 Accents
  • Figurative speech and sarcasm:�-Xavi really shouldn’t play the piano- he has Van Gogh’s ear for music.�-There are no sacred cows in this organization. We will audit everyone from top to bottom.�-Come on Parker, use your loaf! I know you can solve this problem!

12 of 26

What makes language such a difficult problem to solve

  • Languages are constantly evolving�The English Language In 24 Accents
  • Figurative speech and sarcasm:�-Xavi really shouldn’t play the piano- he has Van Gogh’s ear for music.�-There are no sacred cows in this organization. We will audit everyone from top to bottom.�-Come on Parker, use your loaf! I know you can solve this problem!
  • Not good at deducing the meaning of words it hasn’t seen before

13 of 26

What makes language such a difficult problem to solve

  • Synonymous words or phrases (also known as word sense disambiguation)

14 of 26

What makes language such a difficult problem to solve

  • Synonymous words or phrases (also known as word sense disambiguation)
  • Machine translation�French word: Entarter

15 of 26

What makes language such a difficult problem to solve

  • Synonymous words or phrases (also known as word sense disambiguation)
  • Machine translation�French word: Entarter�Means to throw a pie at someone’s face

16 of 26

What makes language such a difficult problem to solve

  • Synonymous words or phrases (also known as word sense disambiguation)
  • Machine translation�French word: Entarter�Means to throw a pie at someone’s face
  • Computers just don’t perceive words in the same sense that we do

17 of 26

18 of 26

19 of 26

Classical NLP

20 of 26

One hot encoded vectors

21 of 26

Problems with one hot-encoded vectors

  • Sparsity of representations
  • No meaningful representation

22 of 26

Word Vector

23 of 26

24 of 26

25 of 26

26 of 26

Cool papers!