This workshop introduces participants to constructing and analyzing term co-occurrence networks using the open-source Python package knowknow. Co-occurrence networks are particularly effective for analyzing small, linguistically diverse datasets, enabling researchers to identify features and trends that may appear in only a few documents. Participants will gain hands-on experience analyzing a curated dataset of journal articles from anthropology, economics, political science, psychology, and sociology (1970–2020), with the option to bring and work with their own datasets.

The session covers techniques for building co-occurrence datasets from academic texts and preparing them for analysis; methods for identifying meaningful terms, their relationships, and the structural properties of co-occurrence networks, such as clusters, hubs, bridges, and temporal patterns; and practical guidance on sharing workflows and datasets using Harvard Dataverse and GitHub to ensure replicability and collaboration.

Morning Session: Building and Visualizing Semantic Co-occurrence Networks. 1) Overview of co-occurrence networks and their application to social science research. 2) Preparing a dataset and building initial co-occurrence networks using knowknow. 3) Visualizing networks using built-in tools. 4) Exporting datasets, documenting workflows, and publishing on open platforms.

Afternoon Session: Interpreting Semantic Co-occurrence Networks. 1) Temporal trends and structural features of semantic networks. 2) Formulating and answering research questions about the dynamics of the social sciences. 3) Hands-on projects.

Prerequisites: Beginner-level familiarity with Python (basic scripting, running code in Jupyter Notebooks). No prior experience with knowknow is required.

Participants should bring a laptop with Python pre-installed or access to an online Python environment (e.g., Google Colab).