Towards Implementation of an Information Dissemination Tool for Health Publications: Case of a Developing Country

Health related publications have been growing at a considerable rate over the years. They are archived at public or private repositories for research or decisionmaking. In developing countries there is an increase in the need for technological infrastructure and funding towards research that enables the voluminous unstructured data to be effectively identified, stored, analysed and visualized to enable prompt decision making. Analysing data in real-time will assist knowledge seekers and researchers in timely access hence quick approach to solutions. We propose a web based, low-cost and user-friendly health information dissemination tool based on machine learning algorithms that analyses full-text publications sequentially and cluster related documents for ease of access. Information retrieval aspect of the model is enhanced through use of a semi-supervised approach that optimises topic selections during search operations. As future works, we propose to scale-up the prototype for bulky data processing and apply it to big data environments.

1. Introduction

Health information dissemination involves strategies aimed at spreading knowledge and their associated evidence-based interventions on a wide scale within or across geographic locations, social or other networks of end-users such as patients and health care providers [1]. Development and implementation of health information dissemination tools may vary depending on the target users and the conveyed information. These target users may be in a broad network whose dissemination of information involves use of web technologies and tools as described in [2] or, a closed group of organisations or community, like in [3], [4]. [3] uses short messaging services (SMS) to enhance community health information system while DHIS2 [4] deals with a large collection of public health data which through an API can allow pre-subscribed health entities and research institutions access to specific resources. This study proposes a web-based, cost effective, user friendly health information dissemination tool based on machine learning algorithms for full text search, information retrieval, classification/clustering and data visualization. We target health domain because of the rapid increase in biomedical research, [5] states that the increase is at a rate of several thousand per week. These published researches are archived in both public and private repositories either for research or as reference for decision-making [6]. Utilization of evidence contained in the publications is challenging due to the magnitude of health publications and the varied publication journals used. This limits the efficiency and effectiveness of knowledge seekers and researchers in finding relevant and related documents in time for decision making. There is therefore need for efficient techniques to discover hidden structures in a collection of documents to counter the challenges that arise in the large and complicated health data [7]. The challenges may include:-(i) quick insight of what is contained in the collection of documents (ii) discovery of data relationships (iii) how much has been researched in a particular topic (iv) the trend of topics (v) research trend in a particular country (what affects it), and (vi) how quick and easy we can access related data to this topic.