Print

Print


*Apologies for cross-posting *

 

 

Dear all,

 

Huge thanks to everyone for their help and suggestions in locating a suitable dataset.

There has been a great response and your assistance has been invaluable. It’s very much appreciated.

 

Dr Proeme has now submitted a project proposal using image data from the Wallace Collection.

 

[But  he remains open to finding that elusive "large" dataset of either measurements or standard-format images – although finding consistent-format data on large numbers of swords is probably a pipe-dream!]

 

Very best regards, and thanks once again,

 

Catrina

 

Catrina Hey

OA & Repository Support

The Library

University of Sussex BN1 9QL

01273 872957

 

 

 

From: Discussion list for UK Research Repository Administrators [mailto:[log in to unmask]] On Behalf Of Catrina Hey
Sent: 01 November 2017 11:57
To: [log in to unmask]
Subject: Seeking large dataset relating to swords

 

·         Apologies for cross-posting *

 

 

Dear all,

 

Can anyone help?

 

I am in search of digital datasets related to swords - the larger the better (the dataset that is, not the sword) - and would be grateful for any leads that list members may be able to provide.

 

This is for a data science MSc project at the University of Edinburgh. Please feel free to contact the supervisor, Dr Arno Proeme, directly at [log in to unmask].

Further details below.

 

Many thanks!

 

Catrina

 

Catrina Hey

OA & Repository Support

The Library

University of Sussex BN1 9QL

01273 872957

 

 

 

 

Further details:

----------

I'm looking for one or more relatively "large" datasets for an MSc student to perform analysis on. This could include numerical sword measurements such as dimensions, weights, points of balance, period/dating, but also non-numerical data such as geographical provenance. For the approaches we have in mind it is more important to have large number of records in the dataset even if each record has only one or two fields, say, rather than a small number of records each with many fields.

 

It's feasible for us to do image analysis, ideally on standardised image formats (consistent size, resolution, lighting, angle of shot) such as one might find in standardised catalogue listings. This could be used for example to see if typology categories arise naturally from either a "dumb" or guided clustering algorithm, to discern temporal evolution of parameters / forms or other correlations including geographical localisation/spread.

 

Although the student's dissertation would become public, I do not believe it would be a requirement for the data on which findings are based to themselves be made public (although this is considered good practice as part of open, transparent and reproducible research and increasingly emphasised by research funding bodies). We have plenty of experience with projects where data is considered confidential (e.g. due to commercial / industrial interests). I should stress that the project would not be primarily interested in answering any particular research question, but instead would serve for the student to put into practice what they will have learned about in the MSc and serve as an opportunity to learn new techniques and technical approaches and to gain experience dealing with real data.

 

The project would take place at the Edinburgh Parallel Computing Centre (EPCC) at the University of Edinburgh. We host supercomputers used by UK academia including the current national service ARCHER (http://www.archer.ac.uk), provide expertise and training in high-performance scientific computing within academia, and engage in software development and hardware projects with academia and industry. We also run MSc programmes in High-Performance Computing and High-Performance Computing with Data Science with an annual cohort of 15-30 students.

 

 

 

Contact the list owner for assistance at [log in to unmask]

For information about joining, leaving and suspending mail (eg during a holiday) see the list website at https://www.jiscmail.ac.uk/cgi-bin/webadmin?A0=archives-nra