# Understanding the Stability of Medical Concept Embeddings

## TL;DR

This paper investigates the stability of medical concept embeddings, revealing that low-frequency concepts are surprisingly stable and identifying context word noisiness as a key factor influencing stability across domains.

## Contribution

It introduces the concept noisiness of context words as a new factor affecting embedding stability, extending understanding beyond frequency effects.

## Key findings

- Low-frequency medical concepts are as stable as high-frequency ones.
- Context word noisiness correlates linearly with embedding stability.
- The proposed factor applies to general domain word embeddings.

## Abstract

Frequency is one of the major factors for training quality word embeddings. Several work has recently discussed the stability of word embeddings in general domain and suggested factors influencing the stability. In this work, we conduct a detailed analysis on the stability of concept embeddings in medical domain, particularly the relation with concept frequency. The analysis reveals the surprising high stability of low-frequency concepts: low-frequency (<100) concepts have the same high stability as high-frequency (>1000) concepts. To develop a deeper understanding of this finding, we propose a new factor, the noisiness of context words, which influences the stability of medical concept embeddings, regardless of frequency. We evaluate the proposed factor by showing the linear correlation with the stability of medical concept embeddings. The correlations are clear and consistent with various groups of medical concepts. Based on the linear relations, we make suggestions on ways to adjust the noisiness of context words for the improvement of stability. Finally, we demonstrate that the proposed factor extends to the word embedding stability in general domain.