Computational Linguistics for Literature
Call for papers
Fifth Workshop on Computational Linguistics for Literature
San Diego, California, USA, June 16, 2016
co-located with NAACL HLT 2016
https://sites.google.com/site/clfl2016/
Contact: clfl2016@googlegroups.com
For better or for worse, on-screen reading is now a reality. There are machine-readable texts from libraries, collections and e-book stores, and more and more "live" literature, such as e-zines, blogs or self-published e-books. We need tools to help navigate, visualize and perhaps even appreciate the high volume of available literature.
Literary texts differ from technical and other formal documents. The primary mode is the narrative rather than the exposition. A story may be cognitively less demanding than a scientific document, but it is a challenge for NLP tools and methods. For example, literary prose lacks overt discourse markers often present in expository genres. Also, even conventional literary texts exhibit far less unity of time, space and topic than most formal discourse. Learning to handle such challenges in literary data may help move past heavy reliance on surface clues in general.
Literature as a genre is unique because of the needs of its typical audience. Just contrast reading or browsing literature online with searching for news on a particular topic. Looking for literature would require certain abstract criteria: literary style, similarity to another work or a point of view. When a summary or a digest is required, knowing or visualizing a text's broad characteristics may be preferred to an outline of the plot. In particular, usual extraction-based summarization methods seldom work for literary narrative.
Large amounts of digitized literary data also enable synchronic and diachronic analysis, and give insight into both history of literature and contemporary trends. Literary theories used to arise from an in-depth study of relatively few works. This form of analysis is detailed, informed and necessarily subjective. Modern, large-scale quantitative studies cannot replace close reading, but they offer broad-coverage, objective insights into a corpus or a genre of literature. The specifics of literary writing require a treatment different than other types of data; it is important to outline the differences and perhaps propose solutions. We will invite papers which touch upon these areas, but we will be open to other ideas which promote the processing of the literary narrative or related forms of discourse.
The topics of interest (papers on related topics will be most welcome):
the needs of the readers and how those needs translate into meaningful NLP tasks;
differences between literature and other types of writing as relevant to NLP;
identification and analysis of literary genres;
finding similar books;
discourse structure in literature;
emotion analysis for literature;
profiling and authorship attribution;
building and analyzing social networks of characters;
computational modelling of narrative / narratology / folkloristics;
modelling dialogue literary style for generation;
searching for literature;
recommendation systems for literature;
summarization of literature;
generation of literary narrative, dialogue or poetry.
We will accept regular papers, describing experimental methods or theoretical work, and we will gladly welcome position papers. The NLP community does not study literature often enough, so it is important to discuss and formulate the problems before proposing solutions. Largely unrestricted argumentative papers will meet that need.