Improving searchability of datasets
Improving searchability of datasets
Data is one of the most important digital assets in the world thanks to its business and social value. As is becoming increasingly available online, in order to use it effectively, we need tools that allow us to retrieve the most relevant datasets that match our information needs. Web search engines are not well suited for this task as they are designed for documents, not data. In recent years several bespoke search engines have been proposed to help with finding datasets, such as Google Dataset Search crawling the whole web or DataMed focused on creating an index of biomedical datasets. In this work we look closer into the problem of searching for data on the example of Open Data platforms. We first applied a mixed-methods approach aimed at deepening our understanding of users of Open Data portals and types of queries they issue while searching for datasets accompanied by analysis of search sessions over one of these data portals. Based on our findings we look into a particular problem of dataset interpretation - meaning of numerical columns. We propose a novel approach for assigning semantic labels to numerical columns. We conclude our work with the analysis of the future work needed in the field in order to potentially improve the searchability of datasets on the web.
University of Southampton
Kacprzak, Emilia, Magdalena
fdc38ad7-6879-4769-ad65-5d3582690af2
Kacprzak, Emilia, Magdalena
fdc38ad7-6879-4769-ad65-5d3582690af2
Ibanez Gonzalez, Luis
65a2e20b-74a9-427d-8c4c-2330285153ed
Kacprzak, Emilia, Magdalena
(2022)
Improving searchability of datasets.
University of Southampton, Doctoral Thesis, 146pp.
Record type:
Thesis
(Doctoral)
Abstract
Data is one of the most important digital assets in the world thanks to its business and social value. As is becoming increasingly available online, in order to use it effectively, we need tools that allow us to retrieve the most relevant datasets that match our information needs. Web search engines are not well suited for this task as they are designed for documents, not data. In recent years several bespoke search engines have been proposed to help with finding datasets, such as Google Dataset Search crawling the whole web or DataMed focused on creating an index of biomedical datasets. In this work we look closer into the problem of searching for data on the example of Open Data platforms. We first applied a mixed-methods approach aimed at deepening our understanding of users of Open Data portals and types of queries they issue while searching for datasets accompanied by analysis of search sessions over one of these data portals. Based on our findings we look into a particular problem of dataset interpretation - meaning of numerical columns. We propose a novel approach for assigning semantic labels to numerical columns. We conclude our work with the analysis of the future work needed in the field in order to potentially improve the searchability of datasets on the web.
Text
Emilia_Kacprzak_PhD_WAIS_27_March
- Version of Record
Text
Permission to deposit thesis - form
Restricted to Repository staff only
More information
Submitted date: March 2022
Identifiers
Local EPrints ID: 457260
URI: http://eprints.soton.ac.uk/id/eprint/457260
PURE UUID: 2bf7dc65-d906-4790-8a4b-ede58674f877
Catalogue record
Date deposited: 30 May 2022 16:34
Last modified: 17 Mar 2024 03:39
Export record
Contributors
Author:
Emilia, Magdalena Kacprzak
Thesis advisor:
Luis Ibanez Gonzalez
Download statistics
Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.
View more statistics