This article discusses the development of annotated lexicons for identifying suicidal ideation (SI) and self-harm (SS) within clinical notes. Utilizing a combination of manual chart reviews and semi-automatic enhancement through word embeddings, a team of clinical experts created a lexicon aimed at improving NLP methodologies in healthcare. Initially, manually reviewed terms had limited application, prompting further enrichment of the lexicon to ensure better identification of SI and SS. The study emphasizes the importance of combining clinical insights with computational methods to enhance data accuracy and relevance.
The iterative approach to lexicon development, combining manual reviews and semi-automatic methods, significantly enhances the identification of suicidal ideation from clinical notes.
The use of word embeddings to enrich manually created lexicons demonstrates the blend of qualitative and quantitative approaches in advancing NLP applications in healthcare.
Initial findings from the lexicon for suicidal ideation revealed limited yield, necessitating extensive reviews by clinical experts to develop a more comprehensive list.
A manual review of clinical notes by psychiatrists and psychologists plays a crucial role in ensuring the accuracy and relevance of the lexicons for NLP tasks.
Collection
[
|
...
]