CIVICA ESR
COURSE CATALOGUE

24 June 2024

Sciences Po

Statistical Text Analysis

Texts — organized and stylized collections of words exchanged between people in social contexts — are one of the main forms of evidence in the social ...

Texts — organized and stylized collections of words exchanged between people in social contexts — are one of the main forms of evidence in the social sciences. Historically, text analysis involved lengthy and painstaking manual coding procedures over modestly sized collections of text. However, over the last three decades, novel algorithms for processing and representing large amounts of text have become very powerful, and the standard workflow for text analysis has become dramatically more technical in a short period of time. This course is designed for doctoral students and advanced master’s students who have some practical experience applying modern statistical text analysis methods to research questions in the social sciences and wish to deepen their knowledge of the logic, design, and mechanics underlying these tools. In particular, the main goal of the course is to help you develop an intuitive grasp of the general word association model (i.e. “word embeddings”). We will (1) collectively discuss the conceptual and theoretical problems inherent to studying meaning quantitatively; (2) survey the mathematical knowledge needed to make the leap from classical quantitative methods in the social sciences (linear regression/GLMs) to language modeling; and (3) gain hands-on experience inspecting the internal workings of language models. Couse on site, open to exchange or visiting CIVICA early stage researchers. Dates; from 24 to 28 June, from 9:30am to 12:30pm and 2:00 to 4:00pm.
Teachers:
  • Alex Kindel (Sciences Po)
Entry requirements: Prior introductory coursework in quantitative methodology (probability theory, correlation, ordinary least squares, generalized linear models) is expected. Some prior experience working with text data would be helpful, but is not required.
Assessment: Project-based (individual or team).
read more


Registration for this course is no longer possible
On site
24/06/24 - 28/06/24
Reg. deadline: 17/06/24
Credits: 0
N° of Sessions: 10