Data mining locative thematic narratives: analysing twitter and POI data to extract precise spatial themes
Data mining locative thematic narratives: analysing twitter and POI data to extract precise spatial themes
This thesis produces a novel framework for constructing thematic narratives of space. Existing research into extracting information from Twitter data is fraught with inconsistencies and a general lack of attention to metadata. This thesis conducts five experiments to delve deeply into how to extract precise location-based themes from Twitter data, with a focus on under-researched metadata fields and exploration of external APIs to enrich the location data of previously assumed ‘precisely’ geolocated tweets. The analysis is based on Twitter data from June 2016 to August 2018. A key argument is made for the analysis of third-party sources and the stratification of tweet granularity to better understand the scale of location-based narratives. These experiments form a framework that is tested in a case study, then validated and evaluated with qualitative interviews with HM Treasury. This framework is the main contribution of the thesis.
University of Southampton
Bennett, Nicholas
7dc9ad84-3507-4595-9b0c-98c2a26be84e
June 2020
Bennett, Nicholas
7dc9ad84-3507-4595-9b0c-98c2a26be84e
Millard, David
4f19bca5-80dc-4533-a101-89a5a0e3b372
Martin, David
e5c52473-e9f0-4f09-b64c-fa32194b162f
Morley, Jeremy
21b99007-07ee-4ece-843a-22d720b149b3
Bennett, Nicholas
(2020)
Data mining locative thematic narratives: analysing twitter and POI data to extract precise spatial themes.
University of Southampton, Doctoral Thesis, 197pp.
Record type:
Thesis
(Doctoral)
Abstract
This thesis produces a novel framework for constructing thematic narratives of space. Existing research into extracting information from Twitter data is fraught with inconsistencies and a general lack of attention to metadata. This thesis conducts five experiments to delve deeply into how to extract precise location-based themes from Twitter data, with a focus on under-researched metadata fields and exploration of external APIs to enrich the location data of previously assumed ‘precisely’ geolocated tweets. The analysis is based on Twitter data from June 2016 to August 2018. A key argument is made for the analysis of third-party sources and the stratification of tweet granularity to better understand the scale of location-based narratives. These experiments form a framework that is tested in a case study, then validated and evaluated with qualitative interviews with HM Treasury. This framework is the main contribution of the thesis.
Restricted to Repository staff only
Text
PDThesis form Bennett - SIGNED
Restricted to Repository staff only
More information
Published date: June 2020
Identifiers
Local EPrints ID: 448269
URI: http://eprints.soton.ac.uk/id/eprint/448269
PURE UUID: a4ef901f-b72a-46f2-9f82-0fded00cbc1c
Catalogue record
Date deposited: 16 Apr 2021 16:34
Last modified: 06 Jun 2024 01:37
Export record
Contributors
Author:
Nicholas Bennett
Thesis advisor:
David Millard
Thesis advisor:
Jeremy Morley
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