emargin.bcu.ac.uk
Research & DevelopmentUnit for English Studies
Andrew Kehoe & Matt Gee
A collaborative textual annotation tool
• Corpus Linguistics: developing software to build and analyse large text collections: crawling, indexing, annotation, search.
Background
• Our own large-scale search engine for linguistic study.
• 10bn words of web text (part-of-speech tagged).
• Includes collections of news and blogs.
• Lets users extract examples of words/phrases in context, monitor change across time, etc.
www.webcorp.org.uk
New Audiences• Bringing Corpus Linguistic techniques to new
audiences:i. School (A-Level) English studentsii. Literary colleagues (teachers/researchers/critics)
• A move toward literary texts and Corpus Stylistic approaches
New Corpora• Literary collections, including:
– Novels of Charles Dickens– Works of Thomas Carlyle– Works of James Joyce– Works of Samuel Beckett– Poems of Percy Bysshe Shelley – Restoration Drama– Science Fiction
• Downloaded and processed whole of Project Gutenberg (23,484 texts; 1.6 billion words)
Colleagues’ Own ExamplesThe doctor seemed especially troubled by the fact of the robbery having been unexpected, and attempted in the night-time; as if it were the established custom of gentlemen in the housebreaking way to transact business at noon, and to make an appointment, by post, a day or two previous. (Oliver Twist)
But there was no hitch in the conversation nevertheless; for one gentleman, who travelled in the perfumery line, exhibited an interesting nick-nack, in the way of a remarkable cake of shaving soap which he had lately met with in Germany;
(Martin Chuzzlewit)
130 instances of ‘in the * [way|line]’
“Dickens is known for a rich range of writing styles-indignant, ironical, melodramatic, and sentimental, all of which appear in David Copperfield. To set the nostalgic tone for this novel, he also uses certain words like "little" and "old" more than usual, so his language seems especially sentimental.”
(Barron’s Book Notes: David Copperfield, 1985, p.32)
Testing Intuitions
• Literary scholars saw benefits of corpus linguistic techniques but concerned about straying too far from the text.
• Literary language is highly creative/variable.
• Corpus Linguistic techniques work best with exact repetitions, not so good at finding paraphrases in fully automated way.
• Difficult to pick up themes/motifs without human input.
Limitations
“corpus stylistics can make an important contribution to the investigation of the interplay between conventional, idiosyncratic and creative patterns of language use. Corpus stylistics also highlights that intuition and automatic processes should work together”
(Mahlberg, 2007:224)
A collaborative textual annotation tool
Literary StudyHow do you study a literary text?
‘Close Reading’: detailed study of short text extracts down to individual word level.
An Established Tradition• Can be traced back to 11th Century.
Martin Luther: Lectures on Romans (1515)
Glossae: student’s notes in the margins
Image from:Cummings, B. (2002) The Literary Culture of the Reformation (Oxford: OUP).
• (re-)read the text
• underline important words
• make notes in margin
• colour-code
• draw out themes/motifs
• (re-)read the text
• underline important words
• make notes in margin
• colour-code
• draw out themes/motifs
• Text quickly becomes cluttered with underlining/ notes on each re-reading
• Annotations tied to printed copy of text
• Difficult to share / combine in class
• Annotations not archivable / searchable
• Text quickly becomes cluttered with underlining/ notes on each re-reading
• Annotations tied to printed copy of text
• Difficult to share / combine in class
• Annotations not archivable / searchable
Increasing emphasis on e-texts
but surprising lack of software to
support close reading.
Difficult to annotate
Difficult to share annotations
Not fine-grained enough for
academic study
Increasing emphasis on e-texts
but surprising lack of software to
support close reading.
Difficult to annotate
Difficult to share annotations
Not fine-grained enough for
academic study
• ‘Book Lovers Fear Dim Future for Notes in the Margins’, New York Times, Feb 20 2011:
–writing comments alongside passages…is a rich literary pastime, sometimes regarded as a tool of literary archaeology, …but it has an uncertain fate in a digitalized world
Limitations of Traditional Model
Our Solution• Web-based collaborative annotation system operating
down to word level.
• Initial prototype late-2007 allowing basic highlighting/ commenting.
• Classroom trials at BCU and Leicester.
Pilot Study• Structured feedback collected from 25 Leicester students
across 3 modules (2 BA, 1 MA).
– 96% found word-level commenting useful.
– 88% found highlighting useful.
– 92% agreed that “reading others’ comments helped me formulate my own ideas”.
– 96% found prototype ‘easy’ to use.
• Pilot study suggested which features of most use.
• JISC Learning & Teaching Innovation grant (June 2011–May 2012) to build fully-functioning,open-source system.
Demonstration of Features
http://emargin.bcu.ac.uk/
Try it yourself for free at:
HIGHLIGHT CLICKED
MULTIPLE ANNOTATIONS
MOUSE DRAGGED
COMMENT ENTERED
ANNOTATION SAVED
ANNOTATION ADDED TO TEXT,
AVAILABLE TO OTHER USERS
ANNOTATION OPENED
REPLY ENTERED
REPLY SAVED, AVAILABLE TO OTHER USERS
NEW HIGHLIGHT
COLOUR CHOSEN
TAG ENTERED
ANNOTATION SAVED
TAG SAVED IN ANNOTATION
TAG CLOUD
LOOK UP HIGHLIGHTED TEXT
IN THE OED(FOR EXAMPLE)
Case Study: Student Projects• Individual research projects on 3rd year BA Narrative
Analysis module at BCU.
• Making connections between literary and linguistic study by examining narrative theories.
• Example: April’s study of newspaper narratives
– 10 articles each from The Sun and The Guardian
– Analysed using 3 narrative models: Labov (1972), White (1997), Hoey (2001).
– In eMargin: used a different colour for each model and tags to indicate the different stages of the model.
– Shows that eMargin can be used individually as a well as collaboratively.
Future Plans• Separate layers of annotation
• Retain text layout and formatting
• Import and Export
Future Plans• Integrate linguistic analysis features
– Corpora
– Tools
• Concordancing
• Wordlists
• Keywords
• Collocation
Phase 2: 2012-13• JISC Embedding Benefits funds for integration
with Virtual Learning Environments (VLEs) using IMS Learning Tools Interoperability specification:
– Single sign-in for seamless transition from VLE to eMargin
– Easier group management - import class lists from VLE
– Compatible with all major VLEs (Moodle, Blackboard Learn, WebCT, etc.)
– Explore potential of eMargin as an e-assessment tool
Beyond English• English Literature in first instance but transferable to
any text-based discipline: Law, Social Sciences, Theology, Languages (and potential beyond text…)
• Trialled at Birmingham School of Acting
• Collaborative research/editing tool
• Beyond HE: United World College of SE Asia
• Working to increase uptake across disciplines
emargin.bcu.ac.uk