Automatic Keyword Extraction from Medical and Healthcare Curriculum

Abstract

Abstract-Medical and healthcare study programmes are quite complicated in terms of branched structure and heterogeneous content. In logical sequence a lot of requirements and demands placed on students appear there. This paper focuses on an innovative way how to discover and understand complex curricula using modern information and communication technologies. We introduce an algorithm for curriculum metadata automatic processing --automatic keyword extraction based on unsupervised approaches, and we demonstrate a real application during a process of innovation and optimization of medical education. The outputs of our pilot analysis represent systematic description of medical curriculum by three different approaches (centrality measures) used for relevant keywords extraction. Further evaluation by senior curriculum designers and guarantors is required to obtain an objective benchmark.

Towards a Keyword Extraction in Medical and Healthcare Education

“…Some of them are separated on the software layer, but we decided to use more than one physical environment due a more effective performance and manageability. Its three main parts involve an application core, a database server and an OpenCPU server [4]: (i) The application core is based on a Symfony framework 2 written in PHP, Javascript and HTML, which uses the Twig template engine 3 . Currently we work with a Symfony framework in version 3.2 and all developed modules are PHP 7.…”

“…Unfortunately, such study programmes usually involve hundreds or even thousands elements of unstructured information which need to be understood, evaluated and optimised. As a result, it is very difficult to look at any given programme, to identify the main topics across the curriculum, and to find one's way through it to see what is actually being taught and how is it done [2,3]. From the perspective of human cognition abilities, it is not possible to carefully read and remember every single detail of all learning units and outcomes including their linkages and co-dependencies.…”

Evaluación e implementación de métodos de extracción automática de palabras clave aplicados a textos cortos en español

“…Las palabras clave en un documento son aquellas que en conjunto permiten identificar sus temas principales y resumir los aspectos más importantes de su contenido. Los métodos de extracción de palabras clave están basados en técnicas automáticas que tienen la finalidad de identificar palabras que se presentan con mayor frecuencia y que tienen mayor importancia dentro de un texto (Komenda, Karolyi, Pokorná, Víta, & Kríž, 2016). La extracción automática de palabras clave puede llevarse a cabo mediante métodos supervisados y no supervisados (Lopez, 2014); los métodos supervisados se caracterizan por tener dependencia sobre un conjunto de datos de entrenamiento (lo que dificulta su aplicación en contextos diferentes), mientras que los métodos no supervisados se caracterizan por ser más flexibles, pues no existe dependencia obligatoria con los datos de entrenamiento.…”

“…Cálculo de las similitudes entre la historia clínica a clasificar y las palabras claves de cada clase, eligiendo la clase más similar. (Komenda, Karolyi, Pokorná, Víta, & Kríž, 2016) 1. Preprocesamiento 2.…”

“…Los algoritmos de extracción de palabras clave dentro de la literatura consultada relevantes para el proyecto son TF-IDF (Jean-Louis, Gagnon, & Charton, 2013), (Suhan, Zhiqiang, & Juan, 2019), (Varga, Cano, Rowe, Ciravegna, & He, 2014), (Guifen, Yanjun, Wei, & Yiou, 2018), TextRank (Zhao, Du, Chang, & Li, 2017), (Komenda, Karolyi, Pokorná, Víta, & Kríž, 2016), (Zhang, Chen, Zhang, Zuo, & Yu, 2020), (Suhan, Zhiqiang, & Juan, 2019), BERT (Chen, Yang, & Zhao, 2020), RAKE (Jean-Louis, Gagnon, & Charton, 2013), (Lynn, Lee, Choi, & Kim, 2017), (Rose, Engel, Cramer, & Cowley, 2010), (Van Staden & van der Poel, 2017) y YAKE (Campos, y otros, 2020).…”