We are happy to have Prof Ken Benoit from the London School of Economics giving his talk “Beyond a bag of words: Identifying and using multi-word expressions to improve political text analysis”.
Time: 1pm, 21st November 2018
Ken Benoit is associate editor of the American Political Science Review.
He is Professor of Quantitative Social Research Methods at the LSE. He is also the current Director of the Social and Economic Data Science (SEDS) Research Unit. Ken’s research focuses on automated, quantitative methods of processing large amounts of textual and other forms of big data – mainly political texts and social media – and the methodology of text mining. He is the creator and co-author of several popular R packages for text analysis, including quanteda, spacyr, and readtext. He has published extensively on applications of measurement and the analysis of text as data in political science, including machine learning methods and text coding through crowd-sourcing, an approach that combines statistical scaling with the qualitative power of thousands of human coders working in tandem on small coding tasks.
The rapid growth of applications treating text as data has transformed our ability to gain insight into important political phenomena. Almost universal among existing approaches is the adoption of the bag of words approach, counting each word as a feature without regard to grammar or order. This approach remains extremely useful despite being an ob- viously inaccurate model of how observed words are generated in natural language. Many politically meaningful textual features, however, occur not as unigram words but rather as pairs of words or phrases, especially in language relating to policy, political economy, and law. Here we present a hybrid model for detecting these associated words, known as collocations. Using a combination of statistical detection, human judgement,and machine learning, we extract and validate a dictionary of meaningful collocations from three large corpora totalling over 1 billion words, drawn from political manifestos and legislative floor debates. We then examine how the word scores of phrases in a text model compare to the scores of their component terms.