Data and Code for Sensitivity Analysis of Semantic Clustering Algorithms on Static Word Embeddings
Description
This dataset supports the article “A Comparative Empirical Evaluation of Semantic Clustering Algorithms on Static Word Embeddings” submitted to the International Journal of Information Management Data Insights. The dataset includes curated word lists, sensitivity analysis results, and executable code used to assess the robustness of Phase 1 semantic clustering experiments. Pre-trained static embeddings (GloVe and fastText) are not redistributed due to licensing constraints and were loaded via the Gensim API. The materials enable replication of the reported sensitivity analyses and facilitate further comparative research on semantic clustering robustness.
Files
Steps to reproduce
1. Open the file “Sensitivity_Analysis_Notebook.ipynb” in Google Colab or a local Jupyter environment. 2. Install the required Python libraries (gensim, scikit-learn, pandas). 3. Load the pre-trained static word embeddings via the Gensim API: - glove-wiki-gigaword-300 - fasttext-wiki-news-subwords-300 4. Load the curated word lists from the file “word_lists.json”. 5. Run all notebook cells to reproduce the sensitivity analysis. 6. The resulting evaluation metrics (Silhouette Score and Davies–Bouldin Index) will be generated and can be compared to those reported in “Table_S1_Sensitivity_Analysis.csv”. Note: Pre-trained embeddings are not included in this dataset due to licensing restrictions and are downloaded automatically via the Gensim API.
Institutions
- Budapesti Corvinus Egyetem