Please, refer to the following work:
Citraro S., De Deyne S., Stella M., Rossetti G. (2023) Towards hypergraph cognitive networks as feature-rich models of knowledge. [ArXiv to appear]
This repository contains the basic analytical pipeline of the data we used in the work cited above.
Data includes:
- Features from the Glasgow Norms;
- English Free Associations from the Small World of Words project.
An example notebook contains the basic pipeline of the work:
- Data Preprocessing;
- Graph and Hypergraph-based representations of Free Associations;
- Features' Aggregation Strategies based on the above representations;
- Predicting a Target Feature (e.g., ground-truth concreteness) based on the other aggregated features;
Other details:
- Graph-based representations include the following strategies:
- G123 Ego-Network.
- Community Detection based representations: Louvain, EVA, Lemon;
- Prediction:
- Random Forest Regressor;
- Evaluation with RMSE, R2.