Corpus Approaches to the Language of Literature 2008
- 1. Corpus Approaches to the Language of Literature Martin Wynne Oxford Text Archive University of Oxford [email protected]
- 2. OTA
- 3. PALA
- 4. PALA 2008
- 6. 6 What is corpus stylistics? The use of the resources, tools and methodologies of corpus linguistics to carry out literary analysis on the basis of the language of literature.
- 7. 7 Corpus Stylistics - Methods Examining and analysing texts and corpora Comparing texts and corpora Building and annotating resources
- 8. 8 “I'm just going out to commit certain deeds.” In an episode of The Simpsons, Homer has planned with Moe to steal Moe's car and drive it into the water, so that Moe can claim the insurance money. Before Homer goes out to steal the car, he is eating dinner with the family, and is trying to act innocently, as if it is a normal evening. He makes various mistakes, and when he gets up to leave, he says, “I'm just going out to commit certain deeds.”
- 9. 9 Consult a corpus to see how a word / phrase / construction / collocation 'normally' occurs. For example, we can look at 'commit' and 'deeds' in the British National Corpus, and try to answer questions like “Why is this funny?”, and “Why are commit and deeds the wrong words to use here?” Requirements: access to a general reference corpus and analysis tools (preferably online) for concordance, collocation, cluster, distribution, word frequency lists
- 10. 10 Analyse an electronic version of a literary text, using text analysis tools. How does author X use expression Y? How often does she use Y? Does she prefer another expression in certain contexts? In what parts of the novel / play / poem does she tend to use Y? Requirements: (reliable) electronic version of the text (in an appropriate format), plus relevant tools (preferably online)
- 11. 11 Analysing a literary corpus Ask questions like those above, but across the oeuvre of an author, or across a literary genre or time period. Furthermore, analyse variation in an author's work (e.g. compare one novel with the rest) Requirements: a relevant corpus, plus tools that allow for internal comparisons
- 12. 12 Analysing an author's work Clusters >4 words in Dickens - among the top 25: AS IF HE HAD BEEN IN THE COURSE OF THE A QUARTER OF AN HOUR AT THE BOTTOM OF THE WHAT DO YOU THINK OF IN THE MIDDLE OF THE AS IF IT HAD BEEN AT THE TOP OF THE ON THE OTHER SIDE OF AT THE END OF THE AS A MATTER OF COURSE THE OTHER SIDE OF THE UP AND DOWN THE ROOM – Names and labels 21 – Speech 16 – “As if” 6 – Body parts 12 – other 22 Categorisation of cluster types (more than 5 words): Mahlberg, M. “Corpus stylistics: bridging the gap between linguistic and literary studies” In M. Hoey, M. Mahlberg, M. Stubbs, W. Teubert. Text, Discourse, and Corpora. London: Continuum. 2007.
- 13. 13 Making internal comparisons within a text Comparing the speech of one character with the rest, e.g. Romeo and Juliet . Comparing one act or scene with the rest. Comparing the style of one section of a novel with the rest. Requirements: text processing tools to separate text elements, or markup to tag text structure and markup-aware tools, plus keywords software
- 14. 14 Comparing a text to a reference corpus Compare the frequency, distribution and usage of words in the text with a reference corpus. E.g. A Conneticut Yankee in King Arthur's Court by Mark Twain, compared to the British National Corpus (BNC) Requirements: many reference corpora, literary and non-literary, different languages, genres, time periods, etc.
- 15. 15 Comparing texts and corpora I, AND, SIR, KING, YE, IT, MY, LAUNCELOT, ME, WAS, KNIGHTS, MERLIN, KNIGHT, ARMOR, CLARENCE, THING, SANDY, HIM, MARHAUS, THAT, UPON, TOWARD, MORDRED, GAWAINE, CAMELOT, SAGRAMOR, SO, DOWLEY, YES, COULDN'T, MILRAYS, THEN, BUT, THEY, HUNDRED, PRESENTLY, KING'S, ARTHUR'S, WOULD, MAN, HAD, WE, ALL, YONDER, THOU, SLAVE, MIRACLE, OUT, ARTHUR, GOOD, UNTO, COULD, AH, HATH, MYSELF, ERRANTRY, LET, SMOTE, ALONG, WELL, MAGICIAN, NOBLE, HIS, GOT, WHEREFORE, SWORD, HE, EVERYBODY, THEE, SPEAR, YOU, ABBOT, PERADVENTURE, OFFENSE, HERMIT, THEM, PROCESSION, STRAIGHTWAY, A, YET, MONKS, KAY, EVER, GUENEVER
- 16. 16 Comparing a literary corpus to a general reference corpus Identifying and characterizing an author's style, e.g. comparing all of Mark Twain's work with US fiction in the period 1870-1910; Identifying and characterizing literary style (of a period, or genre, etc), e.g. comparing a corpus of US fiction with a corpus of non-fiction from the same period, or comparing dramatic dialogue in plays with real conversation in a spoken corpus. Requirements: More literary corpora, more reference corpora, more computing power!
- 17. 17 Tracing historical change Diachronic studies of the language of literature, studying language change, changes in style, genre, etc. Requirements: sets of historical literary corpora of various time periods, or a diachronic corpus which allows internal comparisons, or a collection of texts (with dates) which can be cross-searched
- 18. 18 Annotating and manually analyzing texts and corpora Can be used to test, refine and develop theories about the language of literature. Theories are forced to demonstrate textual evidence, account for all textual phenomena. Frequencies and relevant frequencies can be calculated. Requirements: lots of time, money and expertise!
- 19. 19 Building and Annotating The Speech, Thought and Writing Presentation Corpus Elena Semino, Mick Short, Martin Wynne et al Lancaster University Identifying, categorising and analysing the functions of all occurrences of reported speech, thought and writing (e.g. direct speech, indirect speech, free indirect speech, direct thought, etc.) in a small corpus of fictional and non-fictional texts (and later also speech)
- 20. 20 Building and annotating (2) VICI Free University of Amsterdam Gerard Steen et al Identifying and categorising metaphorical expressions in a subset of the BNC corpus; analysing usage and distributions across text types and modes
- 21. 21 Further types of analysis More levels of annotation: parsing, semantic tagging, etc. Stylometry Text mining Multilingual, parallel, comparable, translation corpora Socio-cultural and historical investigations in literary corpora But note, please, that you don't need annotation for many useful techniques! Requirements: various!
- 22. 22 A new type of Shakespeare dictionary: Jonathan Culpeper A proposal for a dictionary of the language of Shakespeare, involving better integration of linguistic description, frequency information and non-linguistic information. − How often does X occur? − How often do the particular meanings of X occur? − What kind of words does X tend to co-occur with? − How often do the particular ‘grammatical categories’ of X occur? − What kinds of register does X co-occur with? − What kinds of speaker/addressee does X co-occur with? − Is X part of a particular lexical field (semantic category) and how does that field distribute across the plays? − How can the above help differentiate X word from Y word? − Etc. (1) a particular theoretical approach to meanings, (2) a particular methodology ….. enter Corpus Linguistics
- 23. 23 Using large-scale literary corpora For example, Matthew Jockers, Sarah Allison and others at Stanford University, using large collections of literary texts, from commercial providers, applying corpus linguistic and data mining techniques to address literary research questions e.g. Joe Shapiro comparing quantity of narrative v. descriptive passages in US 19th Century literature Perhaps, particular potential for historical literary and linguistic studies
- 24. 24 Basic methods: summary 1. Examine the norms in a general reference corpus 2. Perform text analysis on an electronic literary text 3. Make internal comparisons in a literary text 4. Analyse a literary corpus 5. Make internal comparisons in a literary corpus 6. Compare a text to a reference corpus 7. Compare a literary corpus to a non-literary corpus 8. Compare different literary corpora with each other 9. Build and annotate corpora 10. Others!
- 25. 25 Methods: conclusion It is becoming increasingly possible to test empirically claims about the language of literature, to search for and provide evidence from texts, and to establish the norms of literary and non-literary style. Stylistics typically makes use of a toolkit of linguistic techniques, methods and resources. Corpus stylistics will become a powerful addition to this toolkit in the future.
- 26. 26 Resources for Corpus Stylistics What do we need? ● Reliable electronic editions of literary texts ● Relevant reference corpora ● Analysis tools ● Interoperability ● Shared access ● Sustainability ● Methodology ● Expertise
- 27. 27 Research Infrastructure The vision is for a set of relevant texts, corpora and tools, hosted in various locations around the world, available online from the user's desktop, via a single sign-on; all the resources and tools working together using high- speed connections and high-performance computing. Plus tools for showing, sharing and collaborating in a virtual workspace. CLARIN is working to build this infrastructure for the use of language resources and technologies across the humanities and social sciences.
- 28. 28 Links Oxford Text Archive (OTA) http://www.ota.ox.ac.uk/ PALA Corpus Stylistics Special Interest Group http://www.pala.ac.uk/sigs/corpus-style/ Corpus-style mailing list http://www.jiscmail.ac.uk/lists/corpus-style.html Speech, Thought and Writing Presentation Project http://bowland-files.lancs.ac.uk/stwp/ British National Corpus http://www.natcorp.ox.ac.uk/ Brigham Young University Corpora from Mark Davies http://corpus.byu.edu/