The Time Machine: Capturing Worlds across Time in Texts

The Time Machine: Capturing Worlds across Time in Texts
Juuso, Ilkka, University of Oulu, Finland, ilkka.juuso@ee.oulu.fi
Opas-Hänninen, Lisa Lena, University of Oulu, Finland, lisa.lena.opas-hanninen@oulu.fi
Johnson, Anthony, University of Oulu, Finland, anthony.johnson@oulu.fi
Seppänen, Tapio, University of Oulu, Finland, tapio.seppanen@oulu.fi
Introduction
This paper describes a number of ways in which a temporally-sensitive electronic dictionary resource, the Historical Thesaurus of the Oxford English Dictionary (2 vols; Oxford, 2009 [=HTOED]), may be employed in the automatic dating of words and entire texts. We investigate how the text captures time: most expressly, how the residue of the present (or the different ‘presents’ of language history) have managed to become trapped in the linguistic matrix of a narrative so that we sense, for instance, the difference between a period being represented and the narrator’s temporal positionality, or even the gap between an author and his or her narrative stance. Through computer-assisted means we analyze the impact of later historical and linguistic events on the reporting of earlier events. To this end, we have developed an automatic system for retrieving dating information and a colour-coded browsing interface for searching and viewing the time-coded text, calling it the ‘Time Machine’.

Novels capture worlds, but however disparate the materials that may go into them, something of the space-time in which they have been written remains as a residue. This, in part, is a function of language itself: the instabilities, changes and, above all, affordances at any one moment of that linguistic mesh that Lotman (1990) might have called the semiosphere. In part, too, it is a function of what, within Cultural Imagology, might more concretely be called the texture of the iconosphere (Johnson 2005, 2006): the distinctiveness given to the world at any particular moment by the concatenation of signifying objects present within it. This is why it is an attractive idea to apply a tool such as a time-coded dictionary to novels written at time t purporting to convey events taking place at a time t-x. Within this frame of thought, the case of the historical novel is a particularly pronounced one. By definition, the genre tries to capture something of the iconosphere of a world that has passed us by (even though its semiosphere may remain that of a contemporary reader). And even in cases where the linguistic texture of the semiosphere is deliberately archaized – or localized by the use of dialect forms – the residue of the present remains. As a test case, we examine Diana Gabaldon’s Cross Stitch (1991): a text which flaunts traditional temporal typologies by figuring a protagonist who crosses from the iconosphere of the mid-twentieth century to that of the seventeenth century and becomes trapped there: perceiving the past in a lexis and syntax which palpably belong to a different age.

Time has previously been explored within documents in several ways. Some work has concentrated on identifying expressions of time within text in an attempt to build models of the succession of events. This has been particularly fruitful in the case of, for example, medical discharge records and road accident reports, where the sequence of events is of great importance (Hirschman 1981, Kayser and Nouioua 2009). Other work has used a training set of time-associated words and a Naïve Bayes Classifier to detect temporal concepts in blogs (Noro et al. 2006). While this work is promising in analyzing writings about daily life in a compact time frame, it seems ill-equipped for investigating iconospheres that deal with spans constituting years or even decades. Thus a tool that can retrieve time-related information from the HTOED automatically offers a very promising way forward for the literary and linguistic scholar.

Methodology
Using the ‘Time Machine’, we map out the iconospheric precision with which Gabaldon represents different characters in her fiction (not to mention the humour generated by the gradual blending of their discourses as the novel progresses). But beyond this, by linking our tool with the powerful additional resources which the HTOED has now opened up for those studying the ‘external’, ‘mental’ and ‘social’ worlds of the novel from a historical and etymological perspective, the project hopes to facilitate the achievement of a more nuanced understanding of the interrelationship between ‘real’ and ‘fictional’ time in the historical novel than has been possible before.

In order to better study iconospheres, we sought to develop a tool that would automatically look up dating information and definitions for words, processing entire texts at a time, thereby removing the need for manual queries using a dictionary. Furthermore, we wanted the tool, on the one hand, to enable users to specify time periods of interest for closer inspection while, on the other hand, it left them free to browse the material through diverse visualization schemes in order to discover trends or new time periods of interest.

At present the tool is a prototype, running inside a web browser, in order to enable rapid experimentation with new visualization schemes using CSS. We use a local SQL database to store the HTOED data. Texts can be uploaded via a browser interface and are processed in any user-defined units tagged in the text, e.g. page by page or speaker by speaker, or in the text as a whole. The tool reads both XML and plain text. To finish verifying that the visualization schemes we have chosen are useful, we wish to bring the tool to the digital humanities community, in addition to the poetics and linguistics community (Johnson et al. 2010). Following this, we intend to develop the final tool in Java for inclusion within the LICHEN toolbox.

Results
What the HTOED is able to offer to the ‘Time Machine’ is the ability to isolate different experiential modes within particular iconospheres at the same time as it reveals the range of etymological meanings open to the reader at any given moment. (This, of course, is an invaluable aid for critics who wish to avoid anachronism in their own readings.) In our preliminary development of the ‘Time Machine’ we concentrated on its capacity for isolating different lexical categories within a given iconosphere and indicating the etymological choices available for particular readings. At the top of the screen, the Source section allows the user to choose either an entire text or some part of it (see Figure 1 below: a case in which the speaker Jamie has been chosen). In the Filter section, the user can choose to narrow the search down to one particular word, or to all words that were in use at a particular time (choosing either first use date, last use date or both). To produce the present screen we started by choosing all the words that entered the language after 1742, in other words after the time period in which Jamie speaks. The Colour-coded text section then highlighted all those words which entered the language after the given date, as did our Wordlist section.

Our initial investigation found that the tool is able to pick out swathes of temporal incongruities from this playful text or, further, search out instances relating more specifically to the ‘external’, ‘mental’ and ‘social’ worlds of the novel. It spots moments when the eighteenth-century clansman Jamie seems prescient (mentioning ‘aesthetics’ for instance, or re-circulating the word ‘sadist’, which has been bandied to him by his twentieth-century wife). It detects instabilities not only in the iconosphere of the 1700s – which Gabaldon has carefully researched – but also in the representation of mid-twentieth-century England (which she appears to have taken more for granted).

However, despite the manifest advantage of using even this approach to the ‘Time Machine’ to spot faultlines and incongruities within the fictional world of a novel, some teething-troubles remained: the most significant being that, unlike human readers, the prototype cannot, of course, intuit the ‘correct’ lexical choice from the range of possible meanings thrown up by a search. Accordingly, we have tweaked the search and display capability of the ‘Time Machine’ so that it can also narrow its lexical catchment area by trawling parts of speech (such as substantives) in which cultural and temporal change exhibit their highest visibility. Figure 1 demonstrates how, using these restrictions, the prototype is able to flag up the way in which Gabaldon has inadvertently endowed Jamie’s lexicon with three words stereotypically associated with Scottishness (‘Sassenach, shinty, sporran’) which were not, in fact, recorded until some time after the period in which Jamie is meant to be speaking.

In sum, our study indicates that automated access to chronological information, such as the date of first use for any given word, and full etymologies has promising applications in literary and historical research that has until now relied mostly on intuition and laborious manual methods to combine dating information and texts. And beyond this, with some adaptation, it is also clear that the ‘Time Machine’ could be of significance within areas such as forensic linguistics, collocation studies, and the study of micro-linguistic change over time in large corpora.

Figure 1

Full Size Image

References:
Gabaldon, D. Cross Stitch, Arrow Books Ltd. 1991

Historical Thesaurus of the Oxford English Dictionary, Oxford: OUP 2009

Hirschman, L. 1981 “Retrieving time information from natural-language texts, ” Proceedings of the 3rd Annual ACM Conference on Research and Development in Information Retrieval,

Johnson, Anthony W. “Notes Towards a New Imagology, ” European English Messenger, 2005 14 1 50-58

Johnson, Anthony W. “New Methodologies: Imagology, Language and English Philology, ” Linguistic Topics and Language Teaching, 2006 Oulu University Press 7-27 Antilla, H. Gear, J. Heikkinen, A. Sallinen, Riitta

Johnson, Anthony W. Opas-Hänninen, L.L. Juuso, I. “Stitches in Time and Switches in Text: Diana Gabaldon and the Historical Thesaurus of the Oxford English Dictionary, ” Paper presented at PALA2010, 2010

Kayser, D. Nouioua, F. “From the textual description of an accident to its causes, ” Journal of Artificial Intelligence, 2009 173 12-13

Lotman, Yuri M. Ann Shukman Universe of the Mind: A Semiotic Theory of Culture, 1990 173 12-13

Noro, T. Inui, T. Takamura, H. Okumura, M. 2006 “Time period identification of events in text, ” Proceedings of the 21st international Conference on Computational Linguistics and the 44th Annual Meeting of the Association for Computational Linguistics (Sydney, Australia, July 17 - 18, 2006). Association for Computational Linguistics, (link)