Sentence Similarity Evaluation Data for Siamese BiLSTM Model

Experimental results evaluating a Siamese BiLSTM model for sentence-level semantic similarity. The dataset likely contains performance metrics and feature-level analysis comparing the model to traditional methods like TF-IDF and cosine similarity. Author Weihong Zhao published the data on figshare under a CC-BY-4.0 license in May 2026.

Use Cases

• Benchmarking sentence similarity models based on reported RMSE and MAE metrics
• Analyzing semantic feature importance based on the mention of TF-IDF, Jaccard similarity, and embedding distances
• Evaluating model transparency and attention distribution based on the application of SHAP and LIME explainable AI techniques

Strengths

• Model performance is quantified with RMSE of 0.16 and MAE of 0.107
• Evaluation includes comparisons to multiple baseline methods: TF-IDF, cosine similarity, and feature-based regressors
• Analysis incorporates explainable AI techniques (SHAP, LIME) for model transparency

Limitations

• Row count is unknown, which may limit suitability assessment
• Column-level documentation is absent; field semantics must be inferred after download
• The dataset is very small (5.5 KB), indicating limited scope

Provenance

Source

figshare

Freshness

Last updated 2026-05-21 17:40:34; freshness should be verified