Information Extraction and Named Entity Recognition for Surgical Data

Abstract

In the field of information extraction, based name entity recognition is one of the major challenges for the word processing operation. NER involves the handling of structured and unstructured data and processing them in terms of entities, people (patient s name), diseases, corresponding doctor and other previous history. NER is instinctively simple for human beings, but most of the name entities are the exact times and capitalized letters with respect to English language and which can be easily recognizable in usual language. There will be many problems which are tedious for the recognition process by human beings for handling these ambiguities with list of name which can be an added advantage in human based recognition process. newlineThe thesis is an effective NER system proposed and built with supervised approach in order to provide better result for NER approach as we trust the globally accepted English language as the input medium based on hospital and clinical database. This research study investigate the criteria for the enhanced performance of NER on clinical test case, for the best practices of NER we have selected two major approach such as LSTM(Long Short term Memory) based on Recurrent Neural Network approach. newlineThe major contributions are initially from the clinical data provided and identified with various entities which are required for processing input data into various predefined categories further followed by the division of data into training and testing such a way that 70% and 30% better accuracy is based on the neural network approach, the processing is made based on the calculation of the Term frequency (TF) and Inverse Document frequency (IDF) on a numerical statics on the corpus data. These calculations are done based on the identification of names from the input data, finally named entities are detected from the sentence using machine learning approach. The thesis also provides a hybrid approach which can build a novel NER system developed on python platform.