Einfhrung in die Pragmatik und Diskurs: Computational Discourse - - PowerPoint PPT Presentation

▶
einf hrung in die pragmatik und diskurs computational
SMART_READER_LITE
LIVE PREVIEW

Einfhrung in die Pragmatik und Diskurs: Computational Discourse - - PowerPoint PPT Presentation

Dec 18, 2022 254 likes •590 views

Einfhrung in die Pragmatik und Diskurs: Computational Discourse Processing A. Palmer/A. Horbach Universitt des Saarlandes 16 June 2014 apalmer@coli.uni-sb.de ( Universitt des Saarlandes ) Alexis Palmer Pragmatik & Diskurs:

slide-1
SLIDE 1

Einführung in die Pragmatik und Diskurs: Computational Discourse Processing

  • A. Palmer/A. Horbach

Universität des Saarlandes

16 June 2014

Alexis Palmer apalmer@coli.uni-sb.de (Universität des Saarlandes) Pragmatik & Diskurs: Computational Discourse Processing 16/06/201416 June 2014 1 / 33

slide-2
SLIDE 2

Outline

1

Overview

2

Research Themes

3

Modeling Entity-Based Coherence: Entity Grid

4

Modeling Discourse Structure: PDTB

slide-3
SLIDE 3

Overview

Main readings

  • Bonnie Webber, Marcus Egg, and Valia Kordoni, Discourse

structure and language technology, NLE vol. 18, no. 4, 2012

  • Eleni Miltsakaki, Rashmi Prasad, Aravind Joshi, and Bonnie

Webber, The Penn Discourse Treebank, LREC 2004 Optional readings:

  • Bonnie Webber and Aravind Joshi, Discourse Structure and

Computation: Past, Present, and Future, ACL 2010

  • Penn Discourse Treebank Annotation Manual
  • Regina Barzilay and Mirella Lapata, Modeling Local Coherence: An

Entity-Based Approach, Computational Linguistics, May 2007

Alexis Palmer apalmer@coli.uni-sb.de (Universität des Saarlandes) Pragmatik & Diskurs: Computational Discourse Processing 16/06/201416 June 2014 3 / 33

slide-4
SLIDE 4

Overview

The plan for today

  • Overview of computational discourse processing
  • Research themes in computational discourse processing
  • Focus 1: Modeling entity-based coherence
  • Focus 2: Modeling discourse structure

Alexis Palmer apalmer@coli.uni-sb.de (Universität des Saarlandes) Pragmatik & Diskurs: Computational Discourse Processing 16/06/201416 June 2014 4 / 33

slide-5
SLIDE 5

Overview

Defining discourse

A multi-part definition of discourse. Following Webber et al., discourse can be thought of as

1 A sequence of sentences 2 which conveys more than its individual sentences through their

relationships with one another, and

3 which exploits special features of language that enable discourse to be

more easily understood.

Alexis Palmer apalmer@coli.uni-sb.de (Universität des Saarlandes) Pragmatik & Diskurs: Computational Discourse Processing 16/06/201416 June 2014 5 / 33

slide-6
SLIDE 6

Overview

A sequence of sentences

Example If they’re drunk and meant to be on parade and you go to their room and they’re lying in a pool of piss, then you lock them up for a day. Implementation question: unit of analysis? Research problem: automatic segmentation

Alexis Palmer apalmer@coli.uni-sb.de (Universität des Saarlandes) Pragmatik & Diskurs: Computational Discourse Processing 16/06/201416 June 2014 6 / 33

slide-7
SLIDE 7

Overview

Meaning beyond the individual sentences

Example Don’t worry about the world coming to an end today. It is already tomorrow in Australia. Research questions: how to model meaning beyond the sentence? to what extent does it connect to meaning of the sentence? how to model sentence meaning? Research problem: automatic identification/classification of meaning relations (given particular inventory)

Alexis Palmer apalmer@coli.uni-sb.de (Universität des Saarlandes) Pragmatik & Diskurs: Computational Discourse Processing 16/06/201416 June 2014 7 / 33

slide-8
SLIDE 8

Overview

Special features of language

Discourse exploits features of language that let us:

  • Talk about topics previously discussed in text
  • Indicate relations between states, events, beliefs, etc.
  • Change to new topics or resume previous topics

Alexis Palmer apalmer@coli.uni-sb.de (Universität des Saarlandes) Pragmatik & Diskurs: Computational Discourse Processing 16/06/201416 June 2014 8 / 33

slide-9
SLIDE 9

Overview

Special features of language 2

Example The police are not here to create disorder. They are here to preserve it. Example Pope John XXIII was asked ‘How many people work in the Vatican?’ He is said to have replied, ‘About half.’ Example Men have a tragic genetic flaw. As a result, they cannot see dirt until there is enough of it to support agriculture.

Alexis Palmer apalmer@coli.uni-sb.de (Universität des Saarlandes) Pragmatik & Diskurs: Computational Discourse Processing 16/06/201416 June 2014 9 / 33

slide-10
SLIDE 10

Overview

Types of approaches to discourse structure

Linear segmentation Discourse chunking Discourse parsing

Alexis Palmer apalmer@coli.uni-sb.de (Universität des Saarlandes) Pragmatik & Diskurs: Computational Discourse Processing 16/06/2014 16 June 2014 10 / 33

slide-11
SLIDE 11

Overview

Some applications

  • Summarization
  • Information extraction
  • Essay analysis and scoring
  • Sentiment analysis and opinion mining
  • Assessing text quality
  • Machine translation
  • ...

Alexis Palmer apalmer@coli.uni-sb.de (Universität des Saarlandes) Pragmatik & Diskurs: Computational Discourse Processing 16/06/2014 16 June 2014 11 / 33

slide-12
SLIDE 12

Outline

1

Overview

2

Research Themes

3

Modeling Entity-Based Coherence: Entity Grid

4

Modeling Discourse Structure: PDTB

slide-13
SLIDE 13

Research Themes

What does discourse structure?

Discourse structures are patterns in text. Different ways of thinking about discourse structure care about different types of elements.

  • Entities
  • Topics
  • Functions
  • Eventualities
  • Coherence/Discourse/Rhetorical relations

Alexis Palmer apalmer@coli.uni-sb.de (Universität des Saarlandes) Pragmatik & Diskurs: Computational Discourse Processing 16/06/2014 16 June 2014 13 / 33

slide-14
SLIDE 14

Research Themes

Coreference resolution

  • Entity-level analysis
  • Linking references to common entities
  • Cues: anaphoric expressions
  • pronouns
  • demonstratives (e.g. this movie)
  • alternate forms of reference (President Obama, Barack Obama,

Obama, President of the US)

  • Supervised learning models work reasonably

well ... for English ... in certain types of texts ...

Alexis Palmer apalmer@coli.uni-sb.de (Universität des Saarlandes) Pragmatik & Diskurs: Computational Discourse Processing 16/06/2014 16 June 2014 14 / 33

slide-15
SLIDE 15

Research Themes

Local coherence: Centering theory

  • Local analysis (words/phrases in pairs of

clauses/sentences)

  • Relationships between entities in adjacent

utterances

  • Coreference is an essential component
  • Some small CT-annotated corpora exist
  • CT has been used in CL for evaluating

coherence

Alexis Palmer apalmer@coli.uni-sb.de (Universität des Saarlandes) Pragmatik & Diskurs: Computational Discourse Processing 16/06/2014 16 June 2014 15 / 33

slide-16
SLIDE 16

Research Themes

Entities and topical structure

Example Gliders are aircraft which do not have a motor. They are sometimes called “sailplanes”. Gliders are controlled by their pilots by using control-sticks. Some gliders

  • nly carry one person, but some gliders can carry two persons...

Gliders cannot get into the air by themselves. They are pulled into the air by an aircraft with a motor or they are pulled up by motor on the ground.

  • entity chains
  • lexical cohesion
  • lexical chains
  • Entity Grid (Barzilay and Lapata)

Alexis Palmer apalmer@coli.uni-sb.de (Universität des Saarlandes) Pragmatik & Diskurs: Computational Discourse Processing 16/06/2014 16 June 2014 16 / 33

slide-17
SLIDE 17

Research Themes

Topics and structure

  • Text/text-passage level analysis
  • Concerned with aboutness
  • Topics used to model structure
  • Topic models ~ underlying topics defined in terms of

which words are used

  • Topic transitions often co-occur with

document-internal boundaries

  • Unsupervised models perform well

Alexis Palmer apalmer@coli.uni-sb.de (Universität des Saarlandes) Pragmatik & Diskurs: Computational Discourse Processing 16/06/2014 16 June 2014 17 / 33

slide-18
SLIDE 18

Research Themes

Functional structure

Different types of functional structure:

  • Genre-related structure (e.g. scientific research papers)
  • Conventionalized high-level functional structure (e.g. Wikipedia, news)
  • Temporal structure
  • Narrative structure
  • Intentional structure (discourse relations)

Alexis Palmer apalmer@coli.uni-sb.de (Universität des Saarlandes) Pragmatik & Diskurs: Computational Discourse Processing 16/06/2014 16 June 2014 18 / 33

slide-19
SLIDE 19

Research Themes

Genre

  • Text-level analysis
  • Genre influences various aspects of texts
  • Structure
  • Themes and topics (but != domain)
  • Choice of vocabulary
  • Linguistic register/style
  • ....
  • Many different classification schemes

Alexis Palmer apalmer@coli.uni-sb.de (Universität des Saarlandes) Pragmatik & Diskurs: Computational Discourse Processing 16/06/2014 16 June 2014 19 / 33

slide-20
SLIDE 20

Research Themes

Discourse modes

  • Text-passage level analysis
  • Following Smith 2003 Modes of Discourse:
  • Narrative
  • Description
  • Report
  • Information
  • Argument
  • Discourse modes “do coherence” in different

ways

Alexis Palmer apalmer@coli.uni-sb.de (Universität des Saarlandes) Pragmatik & Diskurs: Computational Discourse Processing 16/06/2014 16 June 2014 20 / 33

slide-21
SLIDE 21

Research Themes

Narrative structure

  • Eventuality level analysis

Structuring by eventualities (events, states, beliefs, etc.) and their spatio-temporal relations Russian folk tale structure

  • an interdiction is addressed to the protagonist, where the hero is told not to do

something;

  • the interdiction is violated, where the hero does it anyway;
  • the hero leaves home, on a search or journey;
  • the hero is tested or attacked, which prepares teh way for receiving a magic

agent or helper.

Alexis Palmer apalmer@coli.uni-sb.de (Universität des Saarlandes) Pragmatik & Diskurs: Computational Discourse Processing 16/06/2014 16 June 2014 21 / 33

slide-22
SLIDE 22

Research Themes

Temporal structure

  • Eventuality level analysis

TempEval: three tasks In Washington today, the Federal Aviation Administration released air traffic control tapes from the night the TWA Flight eight hundred went down.

  • Extracting and normalizing time expressions (aka Timex, time stamping)
  • Extracting and classifying events
  • Identifying temporal relations/links between time expressions and events

Alexis Palmer apalmer@coli.uni-sb.de (Universität des Saarlandes) Pragmatik & Diskurs: Computational Discourse Processing 16/06/2014 16 June 2014 22 / 33

slide-23
SLIDE 23

Research Themes

Discourse relations and structure

  • Clause/sentence/EDU-level analysis
  • Relations between clauses: causality, temporal

structure, etc.

  • Higher-level structure: discourse parse for entire

texts

  • Resources: corpora
  • Penn Discourse Treebank (PDTB)
  • Rhetorical Structure Theory (RST) Bank
  • DISCOR: texts labeled with SDRT structures

Why not address intentional structure?

Alexis Palmer apalmer@coli.uni-sb.de (Universität des Saarlandes) Pragmatik & Diskurs: Computational Discourse Processing 16/06/2014 16 June 2014 23 / 33

slide-24
SLIDE 24

Research Themes

Structure of discourse relations

Relations holding between the semantic content of two units of discourse.

Example The kite was created in China, about 2800 years ago. Later it spread into

  • ther Asian countries, like India, Japan and Korea. However, the kite only

appeared in Europe by about the year 1600.

  • explicit vs. implicit relations
  • unit of analysis (arguments)
  • sense of the relation

Alexis Palmer apalmer@coli.uni-sb.de (Universität des Saarlandes) Pragmatik & Diskurs: Computational Discourse Processing 16/06/2014 16 June 2014 24 / 33

slide-25
SLIDE 25

Outline

1

Overview

2

Research Themes

3

Modeling Entity-Based Coherence: Entity Grid

4

Modeling Discourse Structure: PDTB

slide-26
SLIDE 26

Modeling Entity-Based Coherence: Entity Grid

Entity Grid

Entirely automatic approach for modeling local coherence in a computationally-feasible way.

  • Barzilay and Lapata 2008
  • Converts text into a set of entity transition sequences
  • Uses syntactic, referential, and distributional information/features

Alexis Palmer apalmer@coli.uni-sb.de (Universität des Saarlandes) Pragmatik & Diskurs: Computational Discourse Processing 16/06/2014 16 June 2014 26 / 33

slide-27
SLIDE 27

Modeling Entity-Based Coherence: Entity Grid

Definition

  • A local entity transition is a sequence [S,O,X,-]n that

represents entity occurrences and their syntactic roles in n adjacent occurrences.

  • S=Subject, O=Object, X=Other arguments, -=not present
  • Each transition has a probability: frequency of occurrence
  • ver total number of transitions of that length.

Alexis Palmer apalmer@coli.uni-sb.de (Universität des Saarlandes) Pragmatik & Diskurs: Computational Discourse Processing 16/06/2014 16 June 2014 27 / 33

slide-28
SLIDE 28

Modeling Entity-Based Coherence: Entity Grid

Entity Grid: Example

Entities in text marked with syntactic roles

  • 1. [The justice department]-S is conducting an [anti-trust trial]-O against

[Microsoft Corp.]-X with [evidence]-X that [the company]-S is increasingly attempting to crush [competitors]-O.

  • 2. [Microsoft]-O is accused of trying to forcefully buy into [markets]-X where

[its own products]-S are not competitive enough to unseat [established brands]-O.

  • 3. [The case]-S revolves around [evidence]-O of [Microsoft]-S aggressively

pressuring [Netscape]-O into merging [browser software]-O.

  • 4. [Microsoft]-S claims [its tactics]-S are commonplace and good

economically.

  • 5. [The government]-S may file [a civil suit]-O ruling that [conspiracy]-S to

curb [competition]-O through [collusion]-X is [a violation of the Sherman Act]-O.

  • 6. [Microsoft]-S continues to show [increased earnings]-O despite [the

trial]-X.

Alexis Palmer apalmer@coli.uni-sb.de (Universität des Saarlandes) Pragmatik & Diskurs: Computational Discourse Processing 16/06/2014 16 June 2014 28 / 33

slide-29
SLIDE 29

Modeling Entity-Based Coherence: Entity Grid

Entity Grid: Example

Grid shows which entities occur where, with which role

Alexis Palmer apalmer@coli.uni-sb.de (Universität des Saarlandes) Pragmatik & Diskurs: Computational Discourse Processing 16/06/2014 16 June 2014 29 / 33

slide-30
SLIDE 30

Modeling Entity-Based Coherence: Entity Grid

Entity Grid: Example

Probability of each transition type is computed, then used as feature Probability of [O-] = 7/75 = 0.093

Alexis Palmer apalmer@coli.uni-sb.de (Universität des Saarlandes) Pragmatik & Diskurs: Computational Discourse Processing 16/06/2014 16 June 2014 30 / 33

slide-31
SLIDE 31

Modeling Entity-Based Coherence: Entity Grid

Evaluation

Entity grid approach is evaluated in three applications:

1 Information ordering 2 Evaluation of summary coherence 3 Readability assessment

Alexis Palmer apalmer@coli.uni-sb.de (Universität des Saarlandes) Pragmatik & Diskurs: Computational Discourse Processing 16/06/2014 16 June 2014 31 / 33

slide-32
SLIDE 32

Outline

1

Overview

2

Research Themes

3

Modeling Entity-Based Coherence: Entity Grid

4

Modeling Discourse Structure: PDTB

slide-33
SLIDE 33

Modeling Discourse Structure: PDTB

Penn Discourse Treebank

Corpus of texts from the Wall Street Journal annotated with discourse

  • relations. Has enabled much research, both empirical analysis and

development of systems for automatic analysis.

Example Slides from Nikos Bampounis

Alexis Palmer apalmer@coli.uni-sb.de (Universität des Saarlandes) Pragmatik & Diskurs: Computational Discourse Processing 16/06/2014 16 June 2014 33 / 33