Medical entity recognition of Esophageal Carcinoma based on word clustering

Abstract

Electronic Medical Records (EMRs) is the core of medical information system in hospital. EMRs arises from the medical institution, and large amount of clinical data generated every day. Due to the EMRs contains many medical entities and clinical information of the patient, by analyzing and mining the texts data, medical knowledge which closely related to patients or certain diseases can be obtained. In this paper, we ues the EMRs of patients with Esophageal Carcinoma(EC), which include the clinical symptoms of the patients, tests they have undergone, results of the examinations, and the diagnosis and treatment plan. In this paper, word vector training is carried out for large-scale electronic medical record corpus by means of the skip model of word2vec deep learning tool. Then applying the features data to the k-means clustering algorithm to identifying medical entities about EC by word clustering. As the first step of medical knowledge mining, medical entity recognition is the premise of extracting the semantic relationship implied in medical text and structuring EMRs.