Faculty of Arts and Social Sciences
Lancaster University

Home
About FASS
- Home page
- Read the Dean's Welcome
- Contact Us
Departments
Departments
- Find departments, websites and department heads.
- Go to Department list
People
Faculty Office
- Contact Deans, admin, research support, finance and IT staff via the Faculty Office
Find all Academic and Support Staff
- Academic and support staff in departments
- General Contact information
Study Here
Undergraduate Study
- Introduction to undergraduate study
- Studying for a degree
Postgraduate Study
- Introduction to Postgraduate Study
- Go to the Graduate School Website for information about funding, research training, view student profiles etc.
Research
Our Research
Publications
- Download publications from Lancaster's Publications Repository.
- Recent Books
See also
- Business and Enterprise
- Knowledge Transfer
Business & Enterprise
Business and Enterprise
See also
- Research
- Knowledge Transfer
News & Events
FASS Events
- All Forthcoming Events
- Events Archive
FASS News
- Keep up-to-date with our News
- View News Archive
FASS Intranet
FASS Intranet
- LU Staff only. Policies/guidelines, funding, research support, committees etc.
- Go to FASS Intranet

You are here: Home >

Events

News

Subscribe to News and Events

UCREL CRS: Wrangling large-scale data for specialised corpora

Date: 7 March 2013 Time: 2.00-3:00 pm

Venue: FASS Meeting Room 3

UCREL Corpus Research Seminar

Wrangling large-scale data for specialised corpora

Andrew Hardie (LAEL, Lancaster University)

With vast amounts of text now readily accessible via the web, a "specialised corpus" need not be a "small corpus". However, the immensity of web resources presents challenges. Automatically-spidered data comes with none of the structure that characterises carefully-constructed corpora; when the research goal is to approach language of a very specific type, a "flat" corpus of this kind will typically not be satisfactory.

The ESRC-funded project "Metaphor in End-of-Life Care" (MELC) aims to examine metaphoricity in language associated with terminal illness - the language not only of patients but also of carers and medics. We mass-downloaded message-boards amounting to 50 million words, but then faced a number of problems: (1) structuring the data in a way that reflects the conceptual divisions of the original message-board; (2) allowing analysts routes of access into this dataset; (3) labelling the different classes of participant.

By using a bespoke spidering program, rather than an off-the-shelf mass-download tool, we made a single message-board thread correspond to a single corpus text. Within each thread/text, the mark-up identifies different posts, as well as the user responsible for each. A relational database was created in which all threads, users and posts are represented and cross-linked. A web-interface to this database allowed us to annotate user types - identifying users as patients, carers etc. by examining how they identify themselves in their first posts. We can then extract specified "slices" of the corpus for detailed analysis.

These techniques illustrate how a very large web-derived corpus can be made tractable as a resource for detailed analysis such as the investigation of metaphor, whilst respecting the conceptual structure of the original online resource.

Event website: http://ucrel.lancs.ac.uk/crs

Contact:

Who can attend: Anyone