Plug and Play Language Model : A Simple Baseline for Controlled - - PowerPoint PPT Presentation

plug and play language model a simple baseline for
SMART_READER_LITE
LIVE PREVIEW

Plug and Play Language Model : A Simple Baseline for Controlled - - PowerPoint PPT Presentation

Plug and Play Language Model : A Simple Baseline for Controlled Language Generation ICLR20 Sumanth Dathathri CMS, Caltech Eric Frank Uber AI Andrea Madotto HKUST Janice Lan Uber AI Jane Hung Uber AI Piero Molino Uber AI Jason Yosinski Uber


slide-1
SLIDE 1

Plug and Play Language Model : A Simple Baseline for Controlled Language Generation

ICLR20

Xiachong Feng Sumanth Dathathri CMS, Caltech Eric Frank Uber AI Andrea Madotto HKUST Janice Lan Uber AI Jane Hung Uber AI Piero Molino Uber AI Jason Yosinski Uber AI Rosanne Liu Uber AI

slide-2
SLIDE 2

Author

  • 1. Sumanth Dathathri CMS, Caltech
  • 2. Andrea Madotto HKUST
  • 3. Janice Lan Uber AI
  • 4. …… Uber AI
slide-3
SLIDE 3

Background : Pre-trained LM

  • GPT-2
  • BERT
  • Transformer XL
  • ……..
slide-4
SLIDE 4

Background : GPT-2

slide-5
SLIDE 5

Background : GPT-2

https://gpt2.apps.allenai.org/?text=Joel%20is%20a

slide-6
SLIDE 6

Background : Gradient

𝑧 = 𝑦$ 𝑧% = 2𝑦 𝑦 = 1 𝑧% = 2

slide-7
SLIDE 7

Task : Controlled Generation

slide-8
SLIDE 8

Overview : Plug and Play LM for controlled language generation

slide-9
SLIDE 9

P(x) : Language Modeling With Transformers

slide-10
SLIDE 10

P(x) : Language Modeling With Transformers

slide-11
SLIDE 11

P(x) : Language Modeling With Transformers

slide-12
SLIDE 12

P(x) : Language Modeling With Transformers

slide-13
SLIDE 13

P(x) : Language Modeling With Transformers

slide-14
SLIDE 14

P(a|x)

  • Bag of Words (BoW)
  • Discriminator
  • Sentiment
slide-15
SLIDE 15
  • Suppose we want x && a==positive
  • 1. Generate x à p(x)
  • 2. Classifier à p(a|x)
  • 3. If a==positive : Done
  • 4. Else : Generate x……
slide-16
SLIDE 16

Method

slide-17
SLIDE 17

Method : Gradient based

slide-18
SLIDE 18

Method : Gradient based

slide-19
SLIDE 19

Method

slide-20
SLIDE 20

Fluency

  • Kullback–Leibler (KL) Divergence
  • Post-norm Geometric Mean Fusion
slide-21
SLIDE 21

Bag of Words (BoW)

slide-22
SLIDE 22

Discriminator

slide-23
SLIDE 23

Thanks!