Logo des Repositoriums

Auflistung nach Autor:in "Chen, Jinbo"

1 - 2 von 2
  • Konferenzbeitrag
    Improving search results in life science by recommendations based on semantic information
    (Datenbanksysteme für Business, Technologie und Web (BTW 2015) - Workshopband, 2015) Colmsee, Christian; Chen, Jinbo; Schneider, Kerstin; Scholz, Uwe; Lange, Matthias
    The management and handling of big data is a major challenge in the area of life science. Beside the data storage, information retrieval methods have to be adapted to huge data amounts as well. Therefore we present an approach to improve search results in life science by recommendations based on semantic information. In detail we determine relationships between documents by searching for shared database IDs as well as ontology identifiers. We have established a pipeline based on Hadoop allowing a distributed computation of large amounts of textual data. A comparison with the widely used cosine similarity has been performed. Its results are presented in this work as well.
  • Konferenzbeitrag
    Information retrieval in life sciences: The LAILAPS search engine
    (INFORMATIK 2012, 2012) Lange, Matthias; Chen, Jinbo; Scholz, Uwe
    Retrieval and citation of primary data is the important factor in the approaching e-science age. Solving the challenge of building a flexible but homogeneous bioinformatics information retrieval infrastructure to access and query the world life science databases is a crucial factor for an efficient building bioinformatics infrastructure. In this contribution, we demonstrate the use of nine features, which are determined per database entry, in combination with a neural networks as relevance approximator, a novel approach to increase the quality of information retrieval in life science. The implementation of this concept is the LAILAPS search portal. It was designed to support scientist to extract relevant records in a set of millions entries come from private or public databases. In order to consider the fact that data relevance is highly subjective, we support use specific training of several relevance predicting neural networks. In order to make the neural networks working, a continuously training of the networks is performed in background. Here, the system use the user feedback, eighter by conclusions from the user interaction with the query result browser or by manual rating the data quality. Featured by an intuitive web frontend, the user may search over millions of integrated life science data records. The web frontend comprise a browser for relevance ordered query result, a keyword based query system supporting auto completion, spelling suggestions and synonyms. A data browser is provided to inspect and rate matching data records, and finally a recommender system to suggest closely related records. The system is available at http://lailaps.ipk-gatersleben.de