Can You Do NLP Without Code?
A few months ago, I wanted to try the Orange3 platform to analyze how a “no-code” tool works for exploring data through Natural Language Processing (NLP). My goal was to resolve basic questions about data ingestion, preprocessing, and modeling without writing a single line of code.
What Does No-Code NLP Imply?
Extracting knowledge from a corpus of documents using visual programming. This means:
- No reliance on complex Python or R installations.
- Focusing 100% on the problem logic rather than syntax.
- Drastically reducing the learning curve for non-technical analytical profiles.
Reference Tools
In this field, my two go-to tools are:
- KNIME Analytics Platform: Powerful, modular, and with a huge community.
- Orange Data Mining: Visually excellent and very intuitive for rapid prototyping.
Practical Example with Orange3: Horizon 2020 Analysis
To test Orange3, I analyzed the European Horizon 2020 program calls related to the “Connected Car”.
The Workflow
Using the Text Mining add-on, I designed a flow to identify topics and group similar calls.
Results
- Hierarchical Clustering: After preprocessing the documents (tokenization, stopword filtering, and normalization), I used cosine distances to group the topics into 10 coherent clusters.
- Topic Modeling (LDA): The LDA algorithm allowed for identifying dominant keywords. Thanks to Orange’s interactive visualization, selecting a word automatically displays all associated documents, facilitating result validation.
Conclusion
The no-code movement is democratizing access to advanced technologies. What previously took weeks of Python development can now be prototyped in hours through visual flows, allowing the domain expert (business) to be much closer to the data analysis.