Enroller: A Grid-based Research Platform for English and Scots Language

Enroller: A Grid-based Research Platform for English and Scots Language
Anderson, Jean, University of Glasgow, Jean.Anderson@glasgow.ac.uk
Alexander, Marc, University of Glasgow, Marc.Alexander@glasgow.ac.uk
Green, Johanna, University of Glasgow, Johanna.Green@glasgow.ac.uk
Sarwar, Muhammad, National e-Science Centre, UK, Muhammad.Sarwar@glasgow.ac.uk
Sinnott, Richard, University of Melbourne, rsinnott@unimelb.edu.au
Summary
This paper describes a collaboration between eScientists and humanists; specifically Grid scientists and language and literature scholars, working together to create a repository of data sets and tools, combining our most advanced knowledge in all areas.

Language and literature scholars make use of variety of language resources to conduct their research. Such resources include dictionaries, thesauri, corpora, images, audio and video collections. At present most of these resources are distributed, non-interoperable and license protected. As a result researchers typically conduct their research through direct access to independent data sets using multiple browser windows and multiple authorisations. This approach results in non-scalable and less productive research, and often leads to incomplete and/or non-verifiable results.

The JISC funded project, Enhancing Repositories for Language and Literature Researchers (Enroller) is addressing these issues through development of a targeted eResearch environment. This paper presents the current state of progress and outlines how secure access to distributed data resources with targeted analysis and collaboration tools is supported. In the full paper we will also describe how Enroller is exploiting high performance computing infrastructures such as the UK National Grid Service and ScotGrid, and discuss a problematic issue that has arisen through the differing working practices of humanists and scientists.Consider a typical language and literature scenario where a researcher wants to search for a word such as "bonny" in the dictionary to find its meaning; in a thesaurus to look up the concepts and categories it is found in and in a corpus to find the documents containing it. The user may also want to see the concordances and word frequency of the word in each found document. In undertaking this process, they might want to save the different results; share them with others and possibly perform a comparison between many different resultant data sets. This scenario becomes more challenging when multiple dictionaries, thesauri and text corpora need to be cross-searched simultaneously, for example when the researcher wants to lookup the word "bonny" in the Oxford English Dictionary, in the Scottish National Dictionary, and in the Dictionary of the Older Scottish Tongue. The researcher may also want to search for the word in the Historical Thesaurus of English to look up synonyms and related concepts and categories and then search for all of the matching concepts in further corpus resources.

Researchers will want to use the standard text analysis tools: e.g. concordances, word frequencies, collocation clouds. They may well wish to save and download the results for further analysis or use targeted tools to investigate, e.g. variant spellings of the word 'bonny'. The problem is further exacerbated if the researcher decides to search for multiple, possibly hundreds, of words at once and do all of the mentioned tasks at once. Most of the language and literature data environments do not permit scholars to do any of these activities, instead researchers are left with individual level data sets, coded differently (e.g. with different metadata and data formatting), accessible through individual web-based resources with individual access codes and methods.

The challenge for the project is to maintain the data integrity of its collaborators and the security of access-limited data, while facilitating research across and between each dataset for the benefit of researchers in multiple fields. Enroller provides an interactive research infrastructure that provides seamless, secure access to all of the different language and literature data sets in a user-oriented environment. Furthermore, since many of the searching and analysis efforts can be computationally intensive – especially when bulk searching or building of indexes is required - we provide secure access to high performance computing infrastructures such as the UK e-Science National Grid Service (http://www.ngs.ac.uk) and ScotGrid (http://www.scotgrid.ac.uk). In this project, language and literature researchers, including an associated network of international scholars, are now able to access large amounts of language and literature data and analysis services from a single, easy-to-use portal. Enroller is currently working with the following data sets:

The EPSRC and AHRC funded Scottish Corpus of Text and Speech is a collection of text and audio files covering a period from 1945 to the present. The SCOTS corpus is available in TEI-compliant XML. (http://www.scottishcorpus.ac.uk)
The AHRC funded Corpus of Modern Scottish Writing offers a collection of texts and manuscript images from the period 1700 to 1945. (http://www.scottishcorpus.ac.uk/cmsw/)
The AHRC funded Newcastle Electronic Corpus of Tyneside English provides a corpus of dialect speech from Tyneside in Northeast England. The NECTE corpus is encoded in TEI-compliant XML. (http://www.ncl.ac.uk/necte)
The Dictionary of the Scots Language encompasses two major Scottish language dictionaries: the Scottish National Dictionary and the Dictionary of the Older Scottish Tongue. DSL data is available in XML format. (http://www.dsl.ac.uk/dsl)
The Historical Thesaurus of English contains more than 750,000 words from Old English to the present in 250,000 categories. HTE was published in print form by Oxford University Press in 2009 and is a new and significant development for historical language studies. (http://libra.englang.arts.gla.ac.uk/historicalthesaurus/)
The Oxford English Dictionary is a commercial resource published by Oxford University Press and is the authority on English language vocabulary. It is accessible through our search interface by API. (http://www.oed.com)
The inclusion of other data sets is underway, e.g. we have incorporated Hansard, early C19th to late C20th, and are negotiating for the 1755 Dictionary of Samuel Johnson. Many scholars have no platform or assistance to put texts online and make them accessible to others, far less can they make them interoperable with other relevant data sets. Enroller provides a place where users can deposit their own texts. Texts are wrapped in a basic, TEI minimal XML envelope and indexed. The user can choose whether a text is available to all or private. Once deposited a text can be searched from the portal with the other data sets.

The project has data at its heart, but it is also exploiting state of the art high-performance computing and security solutions. In particular the project is employing a Virtual Organisation Membership Service (VOMS)-based solution in accessing the NGS where pooled Enroller accounts are used by researchers accessing these resources through a targeted project portal. This includes use and exploitation of the Workload Management System (WMS) to provide resource broking based job scheduling across all of the NGS nodes. This job scheduling is targeted currently to supporting large-scale searching based upon the Google MapReduce application.

The full paper will describe the Enroller project in more detail and outline the capabilities that are supported and our experiences so far in implementing this infrastructure. This includes a description of how the portal has been made accessible through the UK Access Management Federation; the easy to use search system and the reasons for its human-computer interface choices; the advanced Grid-based information retrieval system, capable of executing linguistic workflows, taking advantage of HPC facilities such as NGS and ScotGrid; how the system supports large-scale data indexes for fast searching; support for tools for linguistic analysis, including concordance, collocation and word frequency analysis; support for seamless secure access to licensed data; support for data deposition and automated indexing services; documented analysis of our user experiences in using of the infrastructure provided thus far. We will outline the plans for future adoption by the wider research community and end with our reflections on this eScience/humanities collaboration.

References:
J. Watt .O. Sinnott T. Doherty J. Jiang 2008 “Portal-based Access to Advanced Security Infrastructures, ” UK e-Science All Hands Meeting , Edinburgh, September 2008 17

M.S. Sarwar R.O. Sinnott 2010 “Towards a Virtual Research Environment for Language and Literature Researchers, ” IEEE e-Science 2010 , Brisbane, Australia, December 2010 8

Java Database Connectivity (JDBC) API, 15/03/2011 (link)

The Internet2 Shibboleth framework , 15/03/2011 (link)

Enroller, 15/03/2011 (link)

Oxford English Dictionary , 15/03/2011 (link)

Scotish Language Dictionaries , 15/03/2011 (link)

The Scottish Corpus of Texts and Speech, 15/03/2011 (link)

Newcastle Electronic Corpus of Tyneside English, 15/03/2011 (link)

The Historial Thesaurus of English , 15/03/2011 (link)