---
date: '2022-11-28T11:37:23'
hypothesis-meta:
  created: '2022-11-28T11:37:23.032429+00:00'
  document:
    title:
    - 1809.09672.pdf
  flagged: false
  group: __world__
  hidden: false
  id: BmIgdm8REe2-umvTlBFiag
  links:
    html: https://hypothes.is/a/BmIgdm8REe2-umvTlBFiag
    incontext: https://hyp.is/BmIgdm8REe2-umvTlBFiag/arxiv.org/pdf/1809.09672.pdf
    json: https://hypothes.is/api/annotations/BmIgdm8REe2-umvTlBFiag
  permissions:
    admin:
    - acct:ravenscroftj@hypothes.is
    delete:
    - acct:ravenscroftj@hypothes.is
    read:
    - group:__world__
    update:
    - acct:ravenscroftj@hypothes.is
  tags:
  - rl
  - bandit
  - NLProc
  - summarization
  target:
  - selector:
    - end: 10812
      start: 10640
      type: TextPositionSelector
    - exact: "Extractive summarization may be regarded as acontextual bandit as follows.\
        \ Each document is acontext, and each ordered subset of a document\u2019ssentences\
        \ is a different action"
      prefix: h ev-ery episode has length one.
      suffix: . Formally, assumethat each cont
      type: TextQuoteSelector
    source: https://arxiv.org/pdf/1809.09672.pdf
  text: We can represent extractive summarization as a bandit problem by treating
    the document as the context and possible reorderings of sentences as actions an
    agent could take
  updated: '2022-11-28T11:37:23.032429+00:00'
  uri: https://arxiv.org/pdf/1809.09672.pdf
  user: acct:ravenscroftj@hypothes.is
  user_info:
    display_name: James Ravenscroft
in-reply-to: https://arxiv.org/pdf/1809.09672.pdf
tags:
- rl
- bandit
- NLProc
- summarization
- hypothesis
type: annotation
url: /annotations/2022/11/28/1669635443

---



 <blockquote>Extractive summarization may be regarded as acontextual bandit as follows. Each document is acontext, and each ordered subset of a document’ssentences is a different action</blockquote>We can represent extractive summarization as a bandit problem by treating the document as the context and possible reorderings of sentences as actions an agent could take