Joint prediction in MST style discourse parsing for argumentation mining Andreas Peldszus Manfred Stede Applied Computational Linguistics, University of Potsdam Dagstuhl 16161 - 18.04.2016 Peldszus, Stede (Uni Potsdam) Joint prediction for argumentation mining Dagstuhl 16161 1 / 29
Outline 1 Argumentation Mining 2 Dataset & Scheme 3 Joint prediction in argumentation mining 4 Multi-layer discourse annotation Peldszus, Stede (Uni Potsdam) Joint prediction for argumentation mining Dagstuhl 16161 2 / 29
Outline 1 Argumentation Mining 2 Dataset & Scheme 3 Joint prediction in argumentation mining 4 Multi-layer discourse annotation Peldszus, Stede (Uni Potsdam) Joint prediction for argumentation mining Dagstuhl 16161 3 / 29
What is argumentation mining? Health insurance companies should naturally cover alternative medical treatments. Not all practices and approaches that are lumped together under this term may have been proven in clinical trials, yet it's precisely their positive effect when accompanying conventional 'western' medical therapies that's been demonstrated as beneficial. Besides many general practitioners offer such counselling and treatments in parallel anyway - and who would want to question their broad expertise? Peldszus, Stede (Uni Potsdam) Joint prediction for argumentation mining Dagstuhl 16161 4 / 29
What is argumentation mining? [e1] Health insurance companies should naturally cover alternative medical treatments. [e2] Not all practices and approaches that are lumped together under this term may 2 have been proven in clinical trials, 1 c2 [e3] yet it's precisely their 3 c3 positive effect when accompanying conventional 'western' medical therapies that's been demonstrated as beneficial. [e4] Besides many general practitioners offer such 4 c4 counselling and treatments in parallel anyway - [e5] and who would want to 5 question their broad expertise? Peldszus, Stede (Uni Potsdam) Joint prediction for argumentation mining Dagstuhl 16161 4 / 29
What is argumentation mining? [e1] Health insurance companies should naturally cover alternative medical treatments. Tasks: [e2] Not all practices and approaches that are lumped together under this term may • EDU segmentation 2 have been proven in clinical trials, • ADU segmentation 1 c2 [e3] yet it's precisely their resp. argumentative relevance 3 c3 positive effect when accompanying conventional • ADU type classification 'western' medical therapies that's been demonstrated as beneficial. • Relation identification [e4] Besides many general • Relation type classification practitioners offer such 4 c4 counselling and treatments in parallel anyway - [e5] and who would want to 5 question their broad expertise? Peldszus, Stede (Uni Potsdam) Joint prediction for argumentation mining Dagstuhl 16161 4 / 29
Outline 1 Argumentation Mining 2 Dataset & Scheme 3 Joint prediction in argumentation mining 4 Multi-layer discourse annotation Peldszus, Stede (Uni Potsdam) Joint prediction for argumentation mining Dagstuhl 16161 5 / 29
Text genres: pro & contra commentaries Source: • pro & contra newspaper commentaries • in Potsdam Commentary Corpus [Stede, 2004] [Stede and Neumann, 2014] Properties: • lots of arguments • rather explicitly marked argumentation • special background knowledge required • main claim may be implicit • full range of persuasive ’tricks’ professional writers can offer Peldszus, Stede (Uni Potsdam) Joint prediction for argumentation mining Dagstuhl 16161 6 / 29
Dataset: argumentative microtexts Properties: • about 5 segments long • each segment is arg. relevant • explicit main claim • at least one possible objection considered Texts: • 23 texts: hand-crafted, covering different arg. configurations • 92 texts: collected in a controlled text generation experiment • with professional parallel translation to English • see [Peldszus and Stede, 2015b] • freely available, CC-by-nc-sa license • https://github.com/peldszus/arg-microtexts Peldszus, Stede (Uni Potsdam) Joint prediction for argumentation mining Dagstuhl 16161 7 / 29
Dataset: argumentative microtexts Properties: • about 5 segments long • each segment is arg. relevant • explicit main claim • at least one possible objection considered Texts: • 23 texts: hand-crafted, covering different arg. configurations • 92 texts: collected in a controlled text generation experiment • with professional parallel translation to English • see [Peldszus and Stede, 2015b] • freely available, CC-by-nc-sa license • https://github.com/peldszus/arg-microtexts Peldszus, Stede (Uni Potsdam) Joint prediction for argumentation mining Dagstuhl 16161 7 / 29
Scheme Freeman’s theory, revised & slightly generalized: [Freeman, 1991, 2011] [Peldszus and Stede, 2013] [e1] Health insurance companies should naturally • node types = argumentative role cover alternative medical treatments. proponent (presents and defends claims) [e2] Not all practices and approaches that are lumped opponent (critically questions) together under this term may 2 have been proven in clinical trials, 1 c2 • link types = argumentative function [e3] yet it's precisely their 3 c3 positive effect when support own claims (normally, by example) accompanying conventional 'western' medical therapies that's been demonstrated as attack other’s claims (rebut, undercut) beneficial. [e4] Besides many general practitioners offer such 4 c4 counselling and treatments in parallel anyway - IAA: 3 expert annotators κ = 0 . 83 [e5] and who would want to 5 question their broad [Peldszus, 2014] expertise? Peldszus, Stede (Uni Potsdam) Joint prediction for argumentation mining Dagstuhl 16161 8 / 29
Scheme Freeman’s theory, revised & slightly generalized: [Freeman, 1991, 2011] [Peldszus and Stede, 2013] [e1] Health insurance companies should naturally • node types = argumentative role cover alternative medical treatments. proponent (presents and defends claims) [e2] Not all practices and approaches that are lumped opponent (critically questions) together under this term may 2 have been proven in clinical trials, 1 c2 • link types = argumentative function [e3] yet it's precisely their 3 c3 positive effect when support own claims (normally, by example) accompanying conventional 'western' medical therapies that's been demonstrated as attack other’s claims (rebut, undercut) beneficial. [e4] Besides many general practitioners offer such 4 c4 counselling and treatments in parallel anyway - IAA: 3 expert annotators κ = 0 . 83 [e5] and who would want to 5 question their broad [Peldszus, 2014] expertise? Peldszus, Stede (Uni Potsdam) Joint prediction for argumentation mining Dagstuhl 16161 8 / 29
Outline 1 Argumentation Mining 2 Dataset & Scheme 3 Joint prediction in argumentation mining 4 Multi-layer discourse annotation Peldszus, Stede (Uni Potsdam) Joint prediction for argumentation mining Dagstuhl 16161 9 / 29
Tasks tackled: [e1] Health insurance companies should naturally cover alternative medical Trained 4 base classifiers: treatments. • attachment (at) [e2] Not all practices and approaches that are lumped together under this term may 464 pairs yes, 2000 pairs no have been proven in clinical trials, • central claim (cc) 2 1 [e3] yet it's precisely their positive effect when 112 yes, 451 no 3 accompanying conventional 'western' medical therapies that's been demonstrated as • role (ro) beneficial. 451 proponent, 125 opponent [e4] Besides many general practitioners offer such 4 counselling and treatments in • function (fu) parallel anyway - 290 support, 174 attacks 5 [e5] and who would want to question their broad expertise? Peldszus, Stede (Uni Potsdam) Joint prediction for argumentation mining Dagstuhl 16161 10 / 29
Joint prediction in argumentation mining Key features: • MST decoding: Valid global structures from (possibly incompatible) local predictions • Joint prediction: Combine predictions of different base classifiers in the graph [Peldszus and Stede, 2015a] Peldszus, Stede (Uni Potsdam) Joint prediction for argumentation mining Dagstuhl 16161 11 / 29
Joint prediction in argumentation mining Key features: • MST decoding: Valid global structures from (possibly incompatible) local predictions • Joint prediction: Combine predictions of different base classifiers in the graph [Peldszus and Stede, 2015a] Peldszus, Stede (Uni Potsdam) Joint prediction for argumentation mining Dagstuhl 16161 11 / 29
Key feature 1: MST decoding Procedure: • predict edge score • apply classification threshold Peldszus, Stede (Uni Potsdam) Joint prediction for argumentation mining Dagstuhl 16161 12 / 29
Key feature 1: MST decoding Procedure: • predict edge score • apply classification threshold Peldszus, Stede (Uni Potsdam) Joint prediction for argumentation mining Dagstuhl 16161 12 / 29
Key feature 1: MST decoding Procedure: • predict edge score • apply classification threshold Not a tree for 85% of the texts! Peldszus, Stede (Uni Potsdam) Joint prediction for argumentation mining Dagstuhl 16161 12 / 29
Key feature 1: MST decoding Procedure: • predict edge score • apply minimum spanning tree algorithm Peldszus, Stede (Uni Potsdam) Joint prediction for argumentation mining Dagstuhl 16161 13 / 29
Recommend
More recommend
