Monolingual Transduction Graham Neubig Site - - PowerPoint PPT Presentation

monolingual transduction
SMART_READER_LITE
LIVE PREVIEW

Monolingual Transduction Graham Neubig Site - - PowerPoint PPT Presentation

CS11-731 Machine Translation and Sequence-to-Sequence Models Monolingual Transduction Graham Neubig Site https://phontron.com/class/mtandseq2seq2019/ (Slides by Graham Neubig and Antonis Anastasopoulos) Examples Summarization


slide-1
SLIDE 1

CS11-731 
 Machine Translation and 
 Sequence-to-Sequence Models

Monolingual Transduction

Graham Neubig

Site https://phontron.com/class/mtandseq2seq2019/

(Slides by Graham Neubig and Antonis Anastasopoulos)

slide-2
SLIDE 2

Examples

  • Summarization
  • Paraphrasing
  • Formality
slide-3
SLIDE 3

Summarization

  • Granularity:
  • Sentence compression
  • Single document summarization
  • Multi-document summarization
  • Type:
  • Abstractive
  • Extractive
slide-4
SLIDE 4

Sentence Compression

  • Reduce sentence length
  • Common setting: first sentence in new article to title

https://www.aclweb.org/anthology/N16-1012.pdf

slide-5
SLIDE 5

https://arxiv.org/pdf/1609.09552.pdf

slide-6
SLIDE 6

https://www.aclweb.org/anthology/N18-1155.pdf

slide-7
SLIDE 7

Single Document Summarization

  • Compress a document to a sentence

https://arxiv.org/abs/1704.04368

  • Adds concept of coverage
slide-8
SLIDE 8
slide-9
SLIDE 9

Multi-document Summarization

  • Multiple documents into a single summary
  • Extractive summary with saliency
slide-10
SLIDE 10

Paraphrasing

  • Microsoft research paraphrase corpus
  • https://www.aclweb.org/anthology/I05-5002.pdf
  • https://github.com/wasiahmad/

paraphrase_identification/tree/master/dataset/msr- paraphrase-corpus

  • https://github.com/wasiahmad/

paraphrase_identification/blob/master/dataset/msr- paraphrase-corpus/msr_paraphrase_train.txt

slide-11
SLIDE 11

Paraphrasing Bilingually (Neurally)

https://www.aclweb.org/anthology/E17-1083.pdf

slide-12
SLIDE 12

Style and Attribute Transfer

  • What is style? a subset of "linguistic variation"
slide-13
SLIDE 13

Supervised Transfer

  • Just train an MT model!

https://www.aclweb.org/anthology/C12-1177.pdf

slide-14
SLIDE 14

Adversarial Transfer

https://papers.nips.cc/paper/7259-style-transfer-from-non- parallel-text-by-cross-alignment.pdf

slide-15
SLIDE 15

Retrieval-based Transfer