11/2/2023 0 Comments Deep synonym medical![]() ![]() The key distinguishing factor is the context of how the terms appear in the clinical note. However, we focus solely on synonyms as this allows for a more precise evaluation. ![]() Within an ontology, there exists other relationships (eg hypernym, hyponym) beyond synonymy. For example, the 2 terms “diabetes type 2” and “diabetes” can be synonymous in that they refer to the same underlying concept, or they could refer to 2 different types of diabetes. However, type level representations cannot address the second challenge: synonym determination is often contextual. Along these lines, work by Wang et al 15 proposed word embeddings for this task, and found that Word2vec representations improved over the previous best approach. A solution to this problem is to use representations of terms that move away from using the lexical items themselves. First, while typical synonyms tend to be lexically dissimilar, medical concepts are both lexically similar (“dilated RA” and “dilated RV”) or dissimilar (“cerebrovascular accident” and “stroke”). While synonym discovery is critical within clinical NLP, the task is especially challenging for disorder mentions in the clinical domain. In this work, we consider the task of identifying whether 2 textual mentions of a disorder refer to the same underlying medical concept. For these reasons, we are interested in methods that can automatically identify new synonyms from clinical notes without supervision. ![]() An additional challenge is that a fixed synonym list may not accurately reflect meaning, as abbreviations and shortened references have meanings that are contextually dependent. As these types of terms are by definition not present in annotated data, models which can leverage large unannotated datasets to infer meaning from surrounding text are critical in representing these terms. Alternatively, new or rare terms may not be present in a standardized ontology. 17 However, while ontologies contain synonyms for a specific term, these often do not contain the variety of synonyms that can occur in clinical notes across different authors and different domains. 13–15 Medical synonyms aid in a variety of clinical tasks, such as medical concept linking, 16 automatic phenotyping, and cohort selection for comparative effectiveness research. Synonym discovery is especially important within the clinical medical domain. 7–10 This task builds on work in measuring semantic similarity between words and phrases. Synonym discovery, and the related task of paraphrase identification, 4–6 have been explored using a variety of methods. Both tasks rely on an expanded list of synonyms to ensure that relevant concepts or documents are retrieved even if they do not contain the query term. Constructing lists of synonyms can be helpful in a range of downstream applications, such as linking concepts to a knowledge base 3 or query expansion in information retrieval. Given a word or phrase, synonym discovery identifies other words or phrases that have the same or similar meaning to the original. Models relying solely on non-annotated data can be trained on a wider variety of texts without the cost of annotation, and thus may capture a broader variety of language. The ability to discover synonyms from models trained on large amounts of unannotated data removes the need to rely on annotated pairs of similar words. Therefore, the ability to discover synonyms, especially without reliance on training data, is an important component in processing training notes. Medical professionals utilize extensive variation of medical terminology, often not evidenced in structured medical resources. However, medical synonyms can be lexically similar (“dilated RA” and “dilated RV”) or dissimilar (“cerebrovascular accident” and “stroke”) contextual information can determine if 2 strings are synonymous. Synonyms can inform learned representations of patients or improve linking mentioned concepts to medical ontologies. An important component of processing medical texts is the identification of synonymous words or phrases.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |