Research map / Label-efficient and robust learning / Recognition and long-tail
Can Embeddings Adequately Represent Medical Terminology? New Large-Scale Medical Term Similarity Datasets Have the Answer!
Claudia Schulz, Damir Juric
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use Fuzzy Jaccard instead of other tested word-embedding similarity metrics (ablation) | Medical word embeddings · New large-scale medical term similarity datasets | accuracy | Improved |
| Use pairwise word-similarity metrics rather than term-vector aggregation (ablation) | Medical word embeddings · New large-scale medical term similarity datasets | accuracy | Hurt |
| Include MeSH thesaurus information in the embeddings | Medical word embeddings · Hard Levenshtein-negative datasets | accuracy: 58.5; 55.6; 61.4; 61.2; 64.2 | Improved |
| Evaluate on hard rather than easy term pairs, where positive and negative pairs can differ lexically in misleading ways | Medical term embeddings · Levenshtein-negative datasets | accuracy: 78.3 -> 58.5 | Hurt |
| Use contextual embeddings for hard term pairs | Contextual embeddings · Hard Levenshtein-negative datasets | AUC | Hurt |
| Use contextual embeddings for lexically similar easy pairs | ELMoPubMed · Easy Levenshtein-negative datasets | accuracy | Improved |
| Use random rather than Levenshtein-based negative sampling | Medical word embeddings · New large-scale medical term similarity datasets | accuracy | Improved |
| Use small existing datasets rather than the new large-scale datasets for embedding comparisons | Medical term similarity evaluation · Existing medical similarity datasets | Spearman correlation | No clear effect |
Closest papers by meaning
- Learning Conceptual-Contextual Embeddings for Medical Text (AAAI 2020 · similarity 0.86)
- Exploiting Sentence Embedding for Medical Question Answering (AAAI 2019 · similarity 0.86)
- GraphPrompt: Graph-Based Prompt Templates for Biomedical Synonym Prediction (AAAI 2023 · similarity 0.85)
- Understanding the Semantic Content of Sparse Word Embeddings Using a Commonsense Knowledge Base (AAAI 2020 · similarity 0.84)
- BoxLM: Unifying Structures and Semantics of Medical Concepts for Diagnosis Prediction in Healthcare (ICML 2025 · similarity 0.84)
- Multimodal Medical Code Tokenizer (ICML 2025 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
