NLP for Clinical Data Analysis: Handling the Unstructured Clinical Information

Background

As the population is increasing at an enormous rate the need for doctors took an exponential rise. As better and skilled doctors had to handle higher volume of patients they always needed assistive technologies which can provide better processing of information and hence better treatment of the ailing patient. With the emergence of Artificial Intelligence and its dominant role in the health care sector the most specialized area that finds the highest utility is the natural language processing. Hence this work will focus on the broader aspect of NLP and clinical data analysis. Since the work focuses on handling unstructured information of clinical domain hence we will try to look at some of the previous work done in this domain. In the work stated in (Iyyer, M et al, 2014) the author provides a neural network based approach for handling the unstructured data. Way back in 1961 a module was developed for extracting information related to the game of base ball as stated in (Green, B et al, 1961). In 1973 a comprehensive amount of work has been done as stated in (Woods W.A, et al, 1973). The authors in (Androutsopoulos, I et al, 1994) proposed an interface for natural language input and the standard SQL backend. The work stated in (Androutsopoulos, I et al, 1995) also discusses in the similar direction of creation of natural language interface of between a traditional RDBMS and the natural language input information. The work of these interfaces is to reduce the complexity involved in totally removing the RDBMS. The work proposed in (Yan-hong, F et al, 2018) highlights the utility of the skip gram model instead of Named Entity Recognizer (NER). The work stated in (Banerjee PS et al, 2019) is a state of the art technology for information retrieval from totally unstructured corpus by using question and answering. In this work the authors have removed the use of SQL at the back end for information extraction and instead of that the information is retrieved using questions asked in the form of natural language text. In (Banerjee PS et al, 2015) the author have worked on the semantic as well as the syntactic level of linguistic analysis so that any natural language information can be processed in a better way. The authors in (Sharma J.A., et al, 2020) have the option of the debugger for better query handling in case of natural language. In most of cases as in (Ranjan, P., & Balabantaray, R. C, et al, 2020) or (Ma, R., Zhang, et al, 2018) the stress is on factoid output of the data that will be processed hence its applicability will be higher in case of clinical data handling. The work in (Yulianti, E, et al, 2018) and in (Iyyer, M, Boyd-Graber, et al, 2014) clearly mentions about the summarization of information also which will be discussed in the later part of the chapter also. Even though a lot of work has been stated but in this chapter the aspect that will be addressed will be the clinical records mostly handwritten. To somehow manage these kind of data the best solution is to make use of natural language understanding.