Mining Information from Collections of Papers: Illustrative Analysis of Groundwater and Disease

Disease surveillance using online news: Dengue and zika in tropical countries

BACKGROUND

Around the world in tropical areas, certain vector-borne diseases have become endemic and hyperendemic. Among the developing nations, there are common difficulties in establishing the incidences of various diseases, especially vector-borne diseases with complex etiologies and a broad spectrum of presentations. One alternative approach to characterization of the disease outbreaks examines the possibilities of developing proxy information from online news articles. Such sources are being evaluated for applications to disease surveillance, early outbreak detection, and epidemiology research. Our study here looks to examine the potential of news articles in elucidating outbreaks of dengue in India and zika disease in Brazil.

OBJECTIVE

This study is designed to assess the potential usefulness of news articles in tracking case numbers of dengue and zika through an improved understanding of how news outlets report on disease. We specifically examine the possibilities of providing near real-time reporting on the development of outbreaks of dengue and zika.

METHODS

Newspaper articles related to dengue fever and zika disease in India and Brazil, respectively were extracted from the LexisNexis database. We targeted news articles available from five popular international news sources and two local newspapers in each country. The news articles were processed to provide yearly and weekly time series in the number of articles concerned with dengue and zika to test their potential suitability as proxies for disease prevalence. The collections of articles were analyzed using a text mining tool-kit that subdivides a collections of news articles into smaller clusters to study the topical focus of articles and their relevance to tracking diseases.

RESULTS

For dengue fever in India, the local newspapers provide a better source of information than international newspapers. The multi-year analysis (2010-2016) suggests that the numbers of dengue cases are strongly correlated with the numbers of news reports, with an R² value of 0.88. For zika disease in Brazil, the news reports provided useful information on the timing of the zika outbreak. Reporting increase sharply at the beginning of 2016, peaked in weeks 5 to 8, and decreased sharply. The numbers of articles remained low for the remainder of 2016 and 2017. Comparisons with reported case again show article numbers to be a useful proxy of prevalence of zika in Brazil.

CONCLUSIONS

The paper describes a strategy that applies newspaper as proxies to monitor outbreaks of infectious diseases and to study the epidemiology. It has potential applicability in some developing countries and regions with relatively poor medical infrastructures and records. Clearly, large national newspapers in India provide a better source of information on diseases than international outlets. This approach has potential with selected diseases in a few selected countries. Article numbers internationally appear to vary in proportion to the perceived health impact.

Disease surveillance using online news: an extended study of dengue fever in India

BACKGROUND

The study demonstrates the potential in using newspaper information as a proxy for monitoring dengue fever outbreaks in India. Online newspapers are being considered as sources of information on disease surveillance, early outbreak detection, and epidemiology research. Our objective is to understand the complex dengue epidemiology and discover inter-relationships between dengue fever and local social-environmental factors by mining information from local Indian news articles.

RESULTS

We search and extract articles from the newspaper database, LexisNexis. News articles related to dengue fever in India are analyzed together with local environmental, climate, and population data in both temporally and spatially to study disease epidemiology. We also examine the influence of newsworthiness for constructing a disease surveillance system. In terms of temporal aspects, dengue outbreaks follow consistent patterns every year. However, for many areas, this application is frustrated by the relatively small numbers of news articles.

CONCLUSIONS

The study has advanced capabilities in producing approaches that provide for richer interpretations of textual information provided in newspaper articles. Such approaches appear particularly well suited for developing countries with relatively poor medical infrastructures and records.

Visual Exploration of Neural Document Embedding in Information Retrieval: Semantics and Feature Selection

Neural embeddings are widely used in language modeling and feature generation with superior computational power. Particularly, neural document embedding - converting texts of variable-length to semantic vector representations - has shown to benefit widespread downstream applications, e.g., information retrieval (IR). However, the black-box nature makes it difficult to understand how the semantics are encoded and employed. We propose visual exploration of neural document embedding to gain insights into the underlying embedding space, and promote the utilization in prevalent IR applications. In this study, we take an IR application-driven view, which is further motivated by biomedical IR in healthcare decision-making, and collaborate with domain experts to design and develop a visual analytics system. This system visualizes neural document embeddings as a configurable document map and enables guidance and reasoning; facilitates to explore the neural embedding space and identify salient neural dimensions (semantic features) per task and domain interest; and supports advisable feature selection (semantic analysis) along with instant visual feedback to promote IR performance. We demonstrate the usefulness and effectiveness of this system and present inspiring findings in use cases. This work will help designers/developers of downstream applications gain insights and confidence in neural document embedding, and exploit that to achieve more favorable performance in application domains.

What's Next Now That the Boom in Contaminant Hydrogeology Has Busted?

This paper shows how important the historical contaminant boom, 1975 to 2005, has been to hydrogeology. Its role in this respect has never been well documented or appreciated. We use text mining and machine learning to extract information describing research trends in subsurface hydrology (SH), groundwater plus the vadose zone. The analysis, using articles from Water Resources Research (WRR) and Environmental Science and Technology, shows how SH grew from a small niche area within broader hydrologic sciences in the 1960s to become the dominant area of research by 1995. Subsequently, the performance of SH has declined, as other areas of hydrologic sciences have emerged. These patterns reflect the growth and decline of research related to contaminant topics in SH. Thus, the decline is economic-less money leads to less research. The boom in contaminant hydrogeology was effectively a 1000-year flood of money for SH. It let us outcompete the non-SH areas for a while. From 1991 to 2000, SH research in WRR was dominated by papers on contaminant-related themes. More recently, these themes and all other areas of SH except one have declined significantly in WRR. All major topics related to non-SH areas are growing except one. This evidence points to a continuing decline in competitiveness of basic research in SH. The fact that critical problems or social needs for research still exist has not changed this basic trend. The focus of research has clearly shifted away from SH to areas of hydrology associated with climate change, water cycle modeling and similar themes.