gms | German Medical Science

Introduction: The COVID-19 pandemic has increased the relevance of information retrieval methods as new insights have to be extracted from hundreds of new publications per day. In addition to the opportunity of using well-known and established literature search engines such as Livivo, PubMed and Scopus, a new problem arose: A lot of works are published as a so-called preprint – a publication form that does not undergo peer-review. Preprints are usually not covered by the aforementioned search engines and should also be treated with caution because of the lack of quality control. However, searching information in different preprint servers, such as medRxiv and preprints.org is time-consuming and exhausting as they all have a different search interface and functionalities. Therefore, in order to support researchers in finding relevant COVID-19 related information, we built a semantic search engine, publicly available under https://preview.zbmed.de [1].

Methods: We developed a semantic search engine in close cooperation with the user community. Thereby, the data sources, the functionalities and the software itself have been continuously improved in order to fit the user needs [2]. In its current version, preVIEW comprises COVID-19 related data of seven different preprint servers. In addition, semantic search is supported by automatic text mining methods. Besides the recognition of diseases and human genes and proteins, we developed SARS-CoV-2 specific terminologies to tag virus proteins and, later, variants of concern. Based on a user request, the next step will be the development of a classifier for long-Covid related articles. Despite the semantic concepts, we tackled different further challenges, e.g. the removal of duplicates so that that only the newest preprint version is shown. Moreover, we developed a filter option for those preprints that are already published in a journal. As we realized that this is not completely covered in the metadata, we also developed an algorithm that searches for a publication in PubMed.

Results: preVIEW is a freely accessible preprint search engine that actually comprises more than 37,000 preprints. We provide a user interface that – according to a usability test – is easy to handle [3]. Moreover, an application programming interface is provided in order to allow for programmatic access.

Conclusions: Corona crisis shed new light on the importance of preprints as they enable rapid communication of new findings. Even though there is a need for applications that allow for information retrieval, this publication type is unlikely to be indexed by experts due to the lack of quality control - which increases the need for automatic text mining solutions. With preVIEW, we built a COVID-19 specific service in which we have explored new methods that will be integrated into our literature search Livivo in the long run.

Keywords: COVID-19, semantic search, information retrieval, preprints