Text Mining Systematic Reviews
About This Workshop
This workshop teaches you how to build a complete, transparent pipeline for extracting structured knowledge from published scientific papers. You will go from a search query all the way to a visual causal knowledge graph, using only deterministic models, meaning every result is traceable back to an exact sentence in an exact paper.
The workshop is built around a real tool: rasoultilburg/SocioCausaNet, a model specifically trained to detect causal claims in social science text and extract the cause-effect pairs from them.
What You Will Learn
Bulk paper retrieval
Set up API access and use the eScience Center package to download papers on a topic of your choice.
PDF to sentences
Get text out of PDFs, clean away the page furniture and citations, and split what is left into sentences.
Causal claims
Run SocioCausaNet to find the sentences that state a cause and an effect, and mark both spans.
Harmonizing constructs
Three ways to deal with the same construct being worded differently in every paper: a published taxonomy, similarity groups, or clusters.
The causal map
Assemble the pairs into a graph, follow causal chains, and trace any edge back to the sentence it came from.
See the Schedule for the timing and a fuller description of each part.
Who Should Come
This workshop is for:
- Social science PhD students who want to make their literature reviews faster and more systematic
- Researchers interested in automating qualitative reviews
- Anyone who wants to extract and visualize the causal theory embedded in a field’s published literature
- People curious about NLP tools but with no background in machine learning
You need basic Python familiarity. Nothing else is assumed.
This Workshop Is Relevant For
- Knowledge extraction from published papers
- Automating qualitative systematic reviews
- Transparent and reproducible text analysis
- Constructing theory from published causal claims
- Creating a nomological map for a research area
