A large amount of data produced by governmental organizations is accessible in the form of tables encoded as CSV files. Semantic table interpretation (STI) strives to transform them into linked data in order to make them more useful. As significant portion of the tabular data is of statistical nature, and therefore comprises predominantly of numeric values, it is paramount to possess effective means for interpreting relations between the entities and their numeric properties as captured in the tables. As the current general-purpose STI tools infer the annotations of the columns almost exclusively from numeric objects of RDF triples already present in the linked data knowledge bases, they are unable to handle unknown input values. This leaves them with weak evidence for their suggestions. On the other hand, known techniques focusing on the numeric values also have their downsides. Either their background knowledge representation is built in a top-down manner from general knowledge bases, which do not reflect the domain of input and in turn do not contain the values in a recognizable form. Or they do not make use of context provided by the general STI tools. This causes them to mismatch annotations of columns consisting from similar values, but of entirely different meaning. This thesis addresses the...
Identifer | oai:union.ndltd.org:nusl.cz/oai:invenio.nusl.cz:393117 |
Date | January 2019 |
Creators | Brodec, Václav |
Contributors | Nečaský, Martin, Svoboda, Martin |
Source Sets | Czech ETDs |
Language | English |
Detected Language | English |
Type | info:eu-repo/semantics/masterThesis |
Rights | info:eu-repo/semantics/restrictedAccess |
Page generated in 0.0018 seconds