Blog Snippets Based Drug Effects Extraction System Using Lexical and Grammatical Restrictions

Blog Snippets Based Drug Effects Extraction System Using Lexical and Grammatical Restrictions

Shiho Kitajima (Graduate School of Information Science and Technology, Hokkaido University, Sapporo, Hokkaido, Japan), Rafal Rzepka (Graduate School of Information Science and Technology, Hokkaido University, Sapporo, Hokkaido, Japan) and Kenji Araki (Graduate School of Information Science and Technology, Hokkaido University, Sapporo, Hokkaido, Japan)
DOI: 10.4018/ijmdem.2014040101
OnDemand PDF Download:
List Price: $37.50


Obtaining medical information has a beneficial influence on patients' treatment and QOL (quality of life). The authors aim to make a system that helps patients to collect narrative information. Extracting information from data written by patients will allow the acquisition of information which is easy to understand and provides encouragement. Additionally, by using large-scale data, the system can be utilized for discovering unknown effects or patterns. As the first step, the purpose of this paper is to extract descriptions of the effects caused by taking drugs as a triplet of expressions from illness survival blogs' snippets. This paper proposes a method to extract the triplets using specific clue words and parsing the results in order to extract from blogs written in free natural language. Moreover, recall was improved by combining their proposed method and a baseline system, and precision was improved by filtering using dictionaries we created from existing medical documents.
Article Preview

1. Introduction

Currently, the total number of patients with serious illnesses in Japan has reached about 18,026,000 (Ministry of Health, Labour and Welfare, 2011). When someone is diagnosed with a disease or feels that they might be ill, medical information is required to solve various questions in such situations. If patients obtain the required information and they are satisfied with it, this can motivate patients in decision-making and proactive treatment, and have beneficial effects for their QOL (quality of life) and health (Setoyama & Nakayama, 2011). Most major sources of medical information for patients are medical personnel, who are recognized as a reliable source (Taniguchi, 2004; Hesse et al., 2005). However, in order to make better decisions, both “evidence information”, which shows the effects for the population stochastically, and “narrative information” which consists of individual stories from personal experience are necessary (O’Connor, 2002). Although obtaining stories of the experiences of other patients with the same illness can reduce a patient’s anxiety and encourage them (Matsumoto et al., 2005; Maeda et al., 2009), it is difficult for medical staff to provide patients with enough narrative information. Therefore, patients are using various sources such as the Internet in addition to doctors, as has been clarified by research (Setoyama & Nakayama, 2011; Walsh et al., 2010). However, it is not easy to extract and collect answers required by patients appropriately, since there is a huge amount of information on the Web, and it is written in free natural language. In addition, we can see from survey results that it is hard for patients to ask doctors or nurses certain questions (Tsuchiya & Horn, 2009; Kokubu 2008), and we can also see that patients feel that there is a lack of information. Accordingly, we aim to make a system that helps patients to obtain information about other patients with the same illness.

Currently, pharmaceuticals used in Japan consist of about 18,000 prescription drugs, which are prescribed based on a diagnosis of a doctor, and about 12,000 OTC (Over-the-Counter) drugs, which people can buy at pharmacies and drugstores without a prescription by a doctor. Although these drugs are necessary for medical treatment, side effects that occur with use are a problem. Early detection of side effects and taking countermeasures are important issues in the medical field (Mikami et al., 2013). Thus, a great deal of research to estimate the relationship between drugs and side effects has been carried out, and there are many medicine search services (NPO Narrative of Disease and Health DIPEx-Japan, n.d.; Pharmaceuticals and Medical Devices Agency, n.d.). These are based on data created by manufacturers and distributors of pharmaceuticals and information reported by doctors or pharmacists. As yet, medical institutions cannot deal with unknown side effects because additional risks of using medication, such as those not reported, often become known after a product is used by a larger number of patients for a longer period of time. Therefore, we need to collect as much information as possible in order to ascertain effects and side effects of medicines clearly. Thus, we focused on large-scale data on the Web which we can obtain in real time. We extract medical information from blogs written by patients and their families in order to acquire narrative information. Although patients’ understanding of diseases and medical care and the reliability of Internet sources are still far from satisfactory, information on the Web is extensive and instantaneous, and thus can be used to understand people’s changes and trends. Additionally, Japanese people usually disclose personal information anonymously on social media and blogs. It is considered that anonymous narratives are unreliable due to lacking a sense of responsibility. However, anonymized information on the Web gives us information that may be inaccessible in the real world, since anonymity allows users to express opinions and ideas that they may not want someone close to know, without the need to worry about appearances (Orita et al., 2007).

Complete Article List

Search this Journal:
Open Access Articles: Forthcoming
Volume 8: 4 Issues (2017)
Volume 7: 4 Issues (2016)
Volume 6: 4 Issues (2015)
Volume 5: 4 Issues (2014)
Volume 4: 4 Issues (2013)
Volume 3: 4 Issues (2012)
Volume 2: 4 Issues (2011)
Volume 1: 4 Issues (2010)
View Complete Journal Contents Listing