200140 SE Vertiefungsseminar: Geist und Gehirn (2023W)

In this course students will learn how to use natural language processing tools to automatically analyze the content of large numbers of texts. After this course students will be able to perform sentiment analyses of modern textual sources from the internet but also of historical sources such as fiction work (movies and books) from the past. Students will be able to quantify the temporal change of values and preferences expressed in these textual sources and its relationship to historical events and socio-economic dynamics. BASIC KNOWLEDGE OF PYTHON AND R are highly recommended.

Contents

Week 1 - Introduction to digital humanities. Brief introduction to computational social sciences; description of the general set of tools and datasets.

Week 2 - Bag of words approach. Refresher of Python syntax and objects; How to create a bag of words using dictionary tools like WordNet.

Week 3 - Preprocessing in NLP and Frequency analysis. Why preprocessing, general steps; How to extract the most frequent words after preprocessing; How to extract the most frequent words in a text;

Week 4 - Introduction to datasets with Pandas. Iterate over cells and texts; Generate clouds of words using WordCloud and average frequency plots.

Week 5 - Sentiment analysis. Overview of techniques used to implement sentiment analysis; Implementation examples.

Week 6 - Introduction to Web scrapping; How to read and extract information from a website; Using BeautifulSoup to read HTML and XML files.

Week 7 - Iteratively collect information from a website using loops; Implementation of the sentiment analysis pipeline to online information.

Week 8 - Webscrapping using API. What are APIs. How to use APIs to extract information from the internet (e.g. IMDB, Twitter); Implementation of the sentiment analysis pipeline to API information.

Week 9 - Culture as fossilized psychology. The challenge of analyzing historical texts. How to use psychometric tools to build valid bags-of-words; plot historical sentiment time series using Seaborn.

Week 10 - Handling Socioeconomical data. Surveying the major datasets for contemporary and historical socioeconomic estimates; Understanding the meaning and differences between socioeconomic variables and how they are affected by historical events; Extracting socio-economic data from datasets and adding them to a dataset of textual word frequencies.

Week 11 - Testing hypotheses on the relationship between historic events, socioeconomics, and word-frequencies. Quasi-experimental methods, time series trends, linear mixed models; Linear mixed model trend comparison, model selection with stepAIC; Implementation with R.

Week 12 - Project presentations and Q&A