August 9, 2017, Montreal, Canada Digital Humanities 2017 The Course of Emotion in Three Centuries of German Text A Methodological Framework Sven Buechel 1 , Johannes Hellrich 1,2 and Udo Hahn 1 1 Jena University Language & Information Engineering 2 Graduate School (JULIE) Lab ’The Romantic Model’ http://www.julielab.de http://www.modellromantik.uni-jena.de Friedrich Schiller University Jena, Jena, Germany The Course of Emotion in Three Centuries of German Text Sven Buechel, Johannes Hellrich and Udo Hahn 1
August 9, 2017, Montreal, Canada Digital Humanities 2017 Motivation: Emotion in the Humanities https://de.wikipedia.org/wiki/Laokoon-Gruppe https://de.wikipedia.org/wiki/Der_Schrei http://www.br.de/telekolleg/faecher/deutsch/literatur/goethe-weimarer-klassik-100.html The Course of Emotion in Three Centuries of German Text Sven Buechel, Johannes Hellrich and Udo Hahn 2
August 9, 2017, Montreal, Canada Digital Humanities 2017 Resource Problem in Sentiment Analysis Contemporary Resources Historical Resources ? The Course of Emotion in Three Centuries of German Text Sven Buechel, Johannes Hellrich and Udo Hahn 3
August 9, 2017, Montreal, Canada Digital Humanities 2017 Methodological Framework adapt + expand apply for emotion analysis lexicon historical text lexicon historically modern adapted The Course of Emotion in Three Centuries of German Text Sven Buechel, Johannes Hellrich and Udo Hahn 4
August 9, 2017, Montreal, Canada Digital Humanities 2017 Methods The Course of Emotion in Three Centuries of German Text Sven Buechel, Johannes Hellrich and Udo Hahn 5
August 9, 2017, Montreal, Canada Digital Humanities 2017 Representing Emotion: VAD 1.0 Joy ● (in control—being controlled) Anger ● Disgust ● Surprise ● 0.5 Dominance Fear ● 0.0 Sadness ● 1.0 0.5 − 0.5 0.0 − 0.5 − 1.0 − 1.0 − 1.0 − 0.5 0.0 0.5 1.0 Valence (displeasure—pleasure) The Course of Emotion in Three Centuries of German Text Sven Buechel, Johannes Hellrich and Udo Hahn 6
August 9, 2017, Montreal, Canada Digital Humanities 2017 Emotion Lexicons • Lexical resources • Store emotion values of individual words • Many lexicons available from psychology Lemma Valence Arousal Dominance sunshine 8.1 5.3 5.4 terrorism 1.6 7.4 2.7 calm 6.9 1.7 7.4 (Warriner et al., 2013) (1-to-9 scales) The Course of Emotion in Three Centuries of German Text Sven Buechel, Johannes Hellrich and Udo Hahn 7
August 9, 2017, Montreal, Canada Digital Humanities 2017 Lexicon Expansion • Automatic expansion based on Seed • Turney-Littman-Algorithm: mouse filth rat computer keyboard The Course of Emotion in Three Centuries of German Text Sven Buechel, Johannes Hellrich and Udo Hahn 8
August 9, 2017, Montreal, Canada Digital Humanities 2017 Measuring Similarity: Word Embeddings • Dominant approach to similarity • Dense, low-dimensional vectors listen as word representation opera • Most common: word2vec song (SGNS) (Mikolov et al., 2013) poem • SVD PPMI supposedly better suited for humanities (Levy et al., novel 2015, Hellrich & Hahn, 2016) read Ø Johannes’ talk for details! (Friday, 9am, LP-26) The Course of Emotion in Three Centuries of German Text Sven Buechel, Johannes Hellrich and Udo Hahn 9
August 9, 2017, Montreal, Canada Digital Humanities 2017 Classical Expansion Methodology contemp. contemp. expanded contemp. expansion embeddings lexicon corpus contemp. lexicon The Course of Emotion in Three Centuries of German Text Sven Buechel, Johannes Hellrich and Udo Hahn 10
August 9, 2017, Montreal, Canada Digital Humanities 2017 Modifications to Expansion Methodology historical expanded+ historical. embeddings adapted expansion corpus Lexicon contemp. lexicon The Course of Emotion in Three Centuries of German Text Sven Buechel, Johannes Hellrich and Udo Hahn 11
August 9, 2017, Montreal, Canada Digital Humanities 2017 Intuition of Temporal Adaptation (Before) mouse filth rat computer keyboard The Course of Emotion in Three Centuries of German Text Sven Buechel, Johannes Hellrich and Udo Hahn 12
August 9, 2017, Montreal, Canada Digital Humanities 2017 Intuition of Temporal Adaptation (After) mouse filth rat computer keyboard The Course of Emotion in Three Centuries of German Text Sven Buechel, Johannes Hellrich and Udo Hahn 13
August 9, 2017, Montreal, Canada Digital Humanities 2017 Experiments I (Lexicon Induction) The Course of Emotion in Three Centuries of German Text Sven Buechel, Johannes Hellrich and Udo Hahn 14
August 9, 2017, Montreal, Canada Digital Humanities 2017 Methodological Framework adapt + expand apply for emotion analysis lexicon historical text lexicon historically modern adapted The Course of Emotion in Three Centuries of German Text Sven Buechel, Johannes Hellrich and Udo Hahn 15
August 9, 2017, Montreal, Canada Digital Humanities 2017 Resources • Seed Lexicon: – ANGST (Schmidtke et al. (2014) – A thousand German word-emotion pairs – Well established acquisition methodology • Target corpus: – Deutsches Textarchiv (DTA) [“German Text Archive”] – Balanced and manually transcribed – Rich meta-data (e.g., genre and subgenre) The Course of Emotion in Three Centuries of German Text Sven Buechel, Johannes Hellrich and Udo Hahn 16
August 9, 2017, Montreal, Canada Digital Humanities 2017 Target Corpus: DTA Im DTA verfügbare Werke Belles lettres Functional Academia nach Genre und Dekade 150 http://www.deutsches-textarchiv.de/doku/textauswahl 100 Documents Werke 50 0 1601ff. 1611ff. 1621ff. 1631ff. 1641ff. 1651ff. 1661ff. 1671ff. 1681ff. 1691ff. 1701ff. 1711ff. 1721ff. 1731ff. 1741ff. 1751ff. 1761ff. 1771ff. 1781ff. 1791ff. 1801ff. 1811ff. 1821ff. 1831ff. 1841ff. 1851ff. 1861ff. 1871ff. 1881ff. 1891ff. 1901ff. 1700s 1800s 1900s 1600s Belletristik Gebrauchsliteratur Wissenschaft • 1 st third shows different genre distribution • Individual decades comprise too little text Ø Aggregate 30-years slices Ø Select 1690-1899 (~ 1k documents, 7 slices) The Course of Emotion in Three Centuries of German Text Sven Buechel, Johannes Hellrich and Udo Hahn 17
August 9, 2017, Montreal, Canada Digital Humanities 2017 Target Corpus: DTA Im DTA verfügbare Werke Belles lettres Functional Academia nach Genre und Dekade 150 http://www.deutsches-textarchiv.de/doku/textauswahl 100 Documents Werke 50 0 1601ff. 1611ff. 1621ff. 1631ff. 1641ff. 1651ff. 1661ff. 1671ff. 1681ff. 1691ff. 1701ff. 1711ff. 1721ff. 1731ff. 1741ff. 1751ff. 1761ff. 1771ff. 1781ff. 1791ff. 1801ff. 1811ff. 1821ff. 1831ff. 1841ff. 1851ff. 1861ff. 1871ff. 1881ff. 1891ff. 1901ff. 1700s 1800s 1900s 1600s Belletristik Gebrauchsliteratur Wissenschaft • 1 st third shows different genre distribution • Individual decades comprise too little text Ø Aggregate 30-years slices Ø Select 1690-1899 (~ 1k documents, 7 slices) The Course of Emotion in Three Centuries of German Text Sven Buechel, Johannes Hellrich and Udo Hahn 18
August 9, 2017, Montreal, Canada Digital Humanities 2017 Affective Course of Sünde [Sin] ● 4 ● ● Standard Deviations ● ● ● ● 2 Valence Arousal 0 Dominance ● ● − 2 ● ● ● ● ● ● ● ● ● ● ● ● − 4 1705 1735 1765 1795 1825 1855 1885 Year • Strong increase in Valence since onset of Enlightenment • Backing up our findings with corpus linguistics… The Course of Emotion in Three Centuries of German Text Sven Buechel, Johannes Hellrich and Udo Hahn 19
August 9, 2017, Montreal, Canada Digital Humanities 2017 Top 10 Collocations for Sünde [Sin] Lemma and Translation Rank 1690s 1890s 1 todt- death (prefix use) Lamm lamb 2 Erzürnung infuriation hinwegnehmen to take away 3 läßlich minor (clerical) Verzeihung forgiveness 4 beichten to confess Ausschweifung excess 5 Nachlaß abatement Gotte god 6 Grobheit crudeness Schande shame 7 verschweigen to conceal Reue repentance 8 beweinen to beweep Ärgernis nuisance 9 pichen to pitch Laster vice 10 beichten to confess aufrichtig sincere The Course of Emotion in Three Centuries of German Text Sven Buechel, Johannes Hellrich and Udo Hahn 20
August 9, 2017, Montreal, Canada Digital Humanities 2017 Experiments II (Application to Historical Text) The Course of Emotion in Three Centuries of German Text Sven Buechel, Johannes Hellrich and Udo Hahn 21
August 9, 2017, Montreal, Canada Digital Humanities 2017 Methodological Framework adapt + expand apply for emotion analysis lexicon historical text lexicon historically modern adapted The Course of Emotion in Three Centuries of German Text Sven Buechel, Johannes Hellrich and Udo Hahn 22
August 9, 2017, Montreal, Canada Digital Humanities 2017 Sentence-Level Emotion Prediction: JE M AS (Buechel & Hahn, ECAI 2016) Available: (sunshine, <8, 3, 5>) (terrorism, <2, 7, 3>) https://github.com/JULIELab/JEmAS (calm, <7, 2, 7>) word emotion lexicon full text BOW VAD score linguistic documents representation calculation normalization The Course of Emotion in Three Centuries of German Text Sven Buechel, Johannes Hellrich and Udo Hahn 23
Recommend
More recommend
