Text to ideology or text to party status?. Hirst, G., Riabinin, Y., Graham, J., Boizot-Roche, M., & Morris, C. Kaal, B., Maks, I., & van Elfrinkhof, A., editors. From Text to Political Positions: Text analysis across disciplines, pages 61–79. John Benjamins Publishing Company, Amsterdam, 2014. Send e-mail to the first author to request a copyPaper doi abstract bibtex Several recent papers have used support-vector machines with word features to classify political texts — in particular, legislative speech — by ideology. Our own work on this topic led us to hypothesize that such classifiers are sensitive not to expressions of ideology but rather to expressions of attack and defence, opposition and government. We tested this hypothesis by training on one parliament and testing on another in which party roles have been interchanged, and we find that the performance of the classifier completely disintegrates. But removing the factor of government-opposition status, as in the European Parliament, enables a more-ideological classification. Our results suggest that the language of attack and defence, of government and opposition, may dominate and confound any sensitivity to ideology in these kinds of classifiers.
@InBook{ hirstt2pp,
author = {Graeme Hirst and Yaroslav Riabinin and Jory Graham and
Magali Boizot-Roche and Colin Morris},
chapter = {Text to ideology or text to party status?},
editor = { Bertie Kaal and Isa Maks and Annemarie van Elfrinkhof},
title = {From Text to Political Positions: Text analysis across
disciplines},
address = {Amsterdam},
publisher = {John Benjamins Publishing Company},
year = {2014},
pages = {61--79},
doi = {doi:10.1075/dapsac.55.05hir},
url = {https://benjamins.com/#catalog/books/dapsac.55.05hir/details}
,
abstract = {Several recent papers have used support-vector machines
with word features to classify political texts — in
particular, legislative speech — by ideology. Our own
work on this topic led us to hypothesize that such
classifiers are sensitive not to expressions of ideology
but rather to expressions of attack and defence, opposition
and government. We tested this hypothesis by training on
one parliament and testing on another in which party roles
have been interchanged, and we find that the performance of
the classifier completely disintegrates. But removing the
factor of government-opposition status, as in the European
Parliament, enables a more-ideological classification. Our
results suggest that the language of attack and defence, of
government and opposition, may dominate and confound any
sensitivity to ideology in these kinds of classifiers.},
note = {Send e-mail to the first author to request a copy}
}
Downloads: 0
{"_id":{"_str":"53d57a49f414ae191e0003e3"},"__v":0,"authorIDs":[],"author_short":["Hirst, G.","Riabinin, Y.","Graham, J.","Boizot-Roche, M.","Morris, C."],"bibbaseid":"hirst-riabinin-graham-boizotroche-morris-fromtexttopoliticalpositionstextanalysisacrossdisciplines-2014","bibdata":{"bibtype":"inbook","type":"inbook","author":[{"firstnames":["Graeme"],"propositions":[],"lastnames":["Hirst"],"suffixes":[]},{"firstnames":["Yaroslav"],"propositions":[],"lastnames":["Riabinin"],"suffixes":[]},{"firstnames":["Jory"],"propositions":[],"lastnames":["Graham"],"suffixes":[]},{"firstnames":["Magali"],"propositions":[],"lastnames":["Boizot-Roche"],"suffixes":[]},{"firstnames":["Colin"],"propositions":[],"lastnames":["Morris"],"suffixes":[]}],"chapter":"Text to ideology or text to party status?","editor":[{"firstnames":["Bertie"],"propositions":[],"lastnames":["Kaal"],"suffixes":[]},{"firstnames":["Isa"],"propositions":[],"lastnames":["Maks"],"suffixes":[]},{"firstnames":["Annemarie"],"propositions":["van"],"lastnames":["Elfrinkhof"],"suffixes":[]}],"title":"From Text to Political Positions: Text analysis across disciplines","address":"Amsterdam","publisher":"John Benjamins Publishing Company","year":"2014","pages":"61–79","doi":"doi:10.1075/dapsac.55.05hir","url":"https://benjamins.com/#catalog/books/dapsac.55.05hir/details","abstract":"Several recent papers have used support-vector machines with word features to classify political texts — in particular, legislative speech — by ideology. Our own work on this topic led us to hypothesize that such classifiers are sensitive not to expressions of ideology but rather to expressions of attack and defence, opposition and government. We tested this hypothesis by training on one parliament and testing on another in which party roles have been interchanged, and we find that the performance of the classifier completely disintegrates. But removing the factor of government-opposition status, as in the European Parliament, enables a more-ideological classification. Our results suggest that the language of attack and defence, of government and opposition, may dominate and confound any sensitivity to ideology in these kinds of classifiers.","note":"Send e-mail to the first author to request a copy","bibtex":"@InBook{\t hirstt2pp,\n author\t= {Graeme Hirst and Yaroslav Riabinin and Jory Graham and\n\t\t Magali Boizot-Roche and Colin Morris},\n chapter\t= {Text to ideology or text to party status?},\n editor\t= { Bertie Kaal and Isa Maks and Annemarie van Elfrinkhof},\n title\t\t= {From Text to Political Positions: Text analysis across\n\t\t disciplines},\n address\t= {Amsterdam},\n publisher\t= {John Benjamins Publishing Company},\n year\t\t= {2014},\n pages\t\t= {61--79},\n doi\t\t= {doi:10.1075/dapsac.55.05hir},\n url\t\t= {https://benjamins.com/#catalog/books/dapsac.55.05hir/details}\n\t\t ,\n abstract\t= {Several recent papers have used support-vector machines\n\t\t with word features to classify political texts — in\n\t\t particular, legislative speech — by ideology. Our own\n\t\t work on this topic led us to hypothesize that such\n\t\t classifiers are sensitive not to expressions of ideology\n\t\t but rather to expressions of attack and defence, opposition\n\t\t and government. We tested this hypothesis by training on\n\t\t one parliament and testing on another in which party roles\n\t\t have been interchanged, and we find that the performance of\n\t\t the classifier completely disintegrates. But removing the\n\t\t factor of government-opposition status, as in the European\n\t\t Parliament, enables a more-ideological classification. Our\n\t\t results suggest that the language of attack and defence, of\n\t\t government and opposition, may dominate and confound any\n\t\t sensitivity to ideology in these kinds of classifiers.},\n note\t\t= {Send e-mail to the first author to request a copy}\n}\n\n","author_short":["Hirst, G.","Riabinin, Y.","Graham, J.","Boizot-Roche, M.","Morris, C."],"editor_short":["Kaal, B.","Maks, I.","van Elfrinkhof, A."],"key":"hirstt2pp","id":"hirstt2pp","bibbaseid":"hirst-riabinin-graham-boizotroche-morris-fromtexttopoliticalpositionstextanalysisacrossdisciplines-2014","role":"author","urls":{"Paper":"https://benjamins.com/#catalog/books/dapsac.55.05hir/details"},"metadata":{"authorlinks":{}}},"bibtype":"inbook","biburl":"www.cs.toronto.edu/~fritz/tmp/compling.bib","creationDate":"2014-07-27T22:16:41.900Z","downloads":0,"keywords":[],"search_terms":["text","political","positions","text","analysis","disciplines","hirst","riabinin","graham","boizot-roche","morris"],"title":"From Text to Political Positions: Text analysis across disciplines","year":2014,"dataSources":["n8jB5BJxaeSmH6mtR","6b6A9kbkw4CsEGnRX"]}