Chrysanne Di Marco
Using OWL for rhetorical ontologies
The Web ontology language, OWL, presents opportunities and challenges for the cognitive representation of rhetorical figures. OWL is a description-logic–based knowledge representation language in which subsumption (“IsA”= hypernymy/hyponomy) is the organizing relationship. However, OWL’s reliance on subsumption as the primary semantic relation forces different sorts of relations to be translated into subsumption relations, but other common relations, e.g., sequence, can’t be represented in any natural way. Our research team is developing innovative ways, through linked data and SWRL rules, to meet the challenges while capitalizing on the opportunities provided by OWL’s automated classification and consistency-checking.
Di Marco will consider two key movements (1) Linked Linguistic Open Data (LLOD): the linking of linguistic and semantic resources in a common Semantic Web format (e.g., Web Ontology Language (OWL)); and (2) Combining ontologies with Machine Learning in Natural Language Processing. She will address two related questions: (1) Whether the RhetFig Ontology, which uses OWL format, might be usefully linked with other LLOD resources, and (2) Whether Machine Learning might one day be integrated with annotated rhetorical corpora (when such corpora eventually exist).
Marie Dubremetz
Syntax Matters for Rhetorical Structure: The Case of Chiasmus
The chiasmus is a rhetorical figure involving the repetition of a pair of words in reverse order, as in “all for one, one for all”. Previous work on detecting chiasmus in running text has only considered superficial features like words and punctuation. During this presentation, we explore the use of syntactic features as a means to improve the quality of chiasmus detection. Our results show that taking syntactic structure into account may increase average precision from about 40 to 65% on texts taken from European Parliament proceedings. To show the generality of the approach, we also evaluate it on literary text and observe a similar improvement and a slightly better overall result.
Randy Allen Harris
Cognition, computation, and chiasmus; Chiasmus, computation, and cognition
This talk will outline the chiastic suite of rhetorical figures, a family of figures built on reverse repetitions, as in the following:
- All for one and one for all.
- A place for everything and everything in its place.
- The right to bear arms is slightly less ridiculous than the right to arm bears.
- Let us never negotiate out of fear, but let us never fear to negotiate.
- Ask not what your country can do for you—ask what you can do for your country.
The chiastic suite is compelling. In the first order, it reveals in the clearest possible terms the neurocognitive dimensions of rhetorical figures. It appeals aesthetically and registers mnemonically by activating our natural affinities for symmetry, opposition, and repetition. In the second order, it shows the most immediate promise of figuration for computational applications. It has an easily detectable pattern linked to a small range of communicative purposes—an iconicity of balance and completeness and relative order that gives rise to simple rhetorical functions (think, for instance, of the chiastic representation of the law of commutation, m+n=n+m, which evinces the rhetorical function, irrelevance of order). And, in the third order, it exhibits combinatoric tendencies with other figures that narrow and enhance those communicative functions.
Daniel Hromada
Fast & Frugal Detection of Chiasm-like Protofigures in English Subsection of CHILDES Corpus
Intervention shall focus on identification and extraction of instances of repetition-based rhetorical figures (e.g. epanaphor, epiphore, chiasm, antimetabole etc.) from English language sections of Child Language Data Exchange System (CHILDES). CHILDES is the biggest publicly available repository of language acquisition data and the size of its English section is quite exhaustive: 1841729 motherese utterances and 1673351 utterances produced by children between 0 and 10 years of age. After introducing the corpus, we shall introduce a fast & frugal method, exploiting the backreference faculty of Perl Compatible Regular Expressions (PCREs) allowing us to identify repetition-involving utterances in complete CHILDES in just few seconds. We shall subsequently assess frequencies of occurrence of chiasm-like expressions and estimates their distributions for different age groups. We shall conclude with discussion aiming to elucidate cognitive and ontogenetic aspects of surface, significant-encoded figures of speech. All experiments shall be immediately reproducible by workshop participants having an access to standard UNIX shell and GNU tools (wc, sort, grep, uniq and perl).
Ashley Kelly
The Problems of Prolepsis
Currently the Cognitive Ontology of Rhetorical Figures does not systematically include nebulous rhetorical figures, such as the figure known as prolepsis (forecasting and countering an argument before it has been offered). While great gains have been made with those rhetorical figures we often call schemes (figures dependent upon relatively tight lexical or syntactic structures or pairings) in the ontology project at UWaterloo, the failure to incorporate the broader range of rhetorical figures raises many questions of the nature of these figures and their historical inclusion in handbooks and other catalogues of devices. These figures do indeed follow many of the same processes of their more formalized kin, but in an effort to investigate these alignments, numerous obstacles must be overcome. My work is an attempt to integrate less formalized figures by charting prolepsis in ontological terms, chiefly in terms of information structure and the classical canon of arrangement.
Jelena Mitrović
Processing of Rhetorical Figures for Serbian – Tools, Lexical resources and Implementation
Research related to rhetorical figures and their automatic processing in Serbian started with building the Ontology of Rhetorical Figures for Serbian which gives a formal description of 98 rhetorical figures and allows for automatic processing of these figures. I will give an overview of the way we have built and evaluated this ontology – OWL2 language was used for modelling in Protege 4.2 tool, and we used SPARQL queries for validation. Automatic detection of irony in tweets is our newest attempt in this regard. We have collected a corpus of tweets using specific rules and we managed to achieve high precision of automatic recognition of tweets using a specific set of features in the process of machine learning, i.e. a classification task. Adding new semantic relations to WordNet, based on the Simile rhetorical figure – I will give an outline of the research performed for the purpose of my PhD thesis, where we developed an automatic method of extracting relevant Adjective-Noun constructs, related to the Simile rhetorical figure (one possible form of this figure).
Cliff O’Reilly
Ontological engineering for rhetorical figures and beyond
The LaRheto application takes text input and automatically generates a knowledge base of meaning representations that are augmented with world knowledge. Inference is undertaken on the knowledge base using a set of Descriptive Rules that, finally, outputs the original text tagged with rhetorical figures. LaRheto is described, with special emphasis on ontological engineering from the perspective of the design process, following through to application. The use of linked ontologies is highlighted and a description of these methods, used within the LaRheto application, is analysed in more detail, including multi-namespace RDF knowledge bases. Linked ontologies are considered further, both from a large-scale engineering perspective and also in relation to extending domains of reference. By example, more recent work on an alternate ontology – designed to cover, more generally, the domains of cognition and knowledge – is shown, both in relation to ontology design and also to links with ontologies of rhetorical figures.
Chris Reed and John Lawrence
Structural Techniques for Argument Mining
Argument mining — the generalised, automated recognition of the structure of human reasoning expressed linguistically — is an enormously challenging task. The large and active philosophical research field of argumentation theory demonstrates the breadth of the issues that are apposite.
Our work in this area aims to combine a range of approaches, looking at the kinds of features common in natural language that enable the audience to understand the points being made and the relations between them. These methods cover linguistic features, changes in the topic being discussed and the identification of argumentation schemes, patterns of human reasoning which have been detailed extensively in philosophy and psychology. We have demonstrated that the structure of such schemes can provide rich information to the task of automatically identify complex argumentative structures in natural language text. By training a range of classifiers to identify the individual proposition types which occur in these schemes, it is possible not only to determine where a scheme is being used, but also the roles played by its component parts.
Whilst these methods focus mainly on the logical structure, this can be strengthened by also working with the rich and extensive domain of rhetorical figures. It seems likely that at least some rhetorical figures can be easily recognised automatically. If the occurrence of rhetorical figures can be correlated with some aspects of argument structure, then a new and potentially extremely powerful set of techniques can be brought on stream for substantially improving the performance of argument mining techniques.
We have, so far, focused on a small set of rhetorical figures that can be easily identified, compared instances of these figures with pre-annotated argument structures, and determined that in many cases connections can be found. Although this work is still at an early stage, requiring expansions of both the datasets used for comparison and the figures considered, it is a promising early sign that any one of the types of rhetorical figure could be interestingly challenging to identify and highly correlated with some aspect of argumentative structure. With over 700 to choose from, we could be at the outset of an extremely rich seam for argument mining.
Michael Ullyot
Toward an Augmented Criticism
The Augmented Criticism lab is detecting features of repetition and variation in the works of Shakespeare and his contemporaries (starting with drama). We’ve begun with rhetorical figures that repeat lemmas (heed, heedful, heeding) or morphemes (heeding, wringing, vexing). We use natural-language processing to discover if these unnatural formulations signal natural habits of thought. The interpretive payoff is our ability to make more definitive arguments not just about these figures, but also about the cognitive processes they signal.
This paper will describe our process and our corpus, and present a range of our results with an initial corpus (the Folger’s Digital Anthology texts) before we expand to the billion words in the EEBO-TCP corpus (1473-1700).
Ying Yuan
On Argumentative Functions of Litotes
Litotes, avoiding a directly affirmative claim via denying its opposite, is a powerful figure, pervasive in both English and Chinese discourses. Previous studies in these two languages have examined its types and functions in brief, but failed to realize that these types convey specific argumentative functions. Based chiefly upon Chaim Perelman and Lucie Olbrechts-Tyteca’s notion of argumentative figure, and Jeanne Fahnestock’s extension to Figural Logic, we investigate the shared active litotes types in English and Chinese texts, and their corresponding argumentative functions. The findings expect to be drawn from the abundant data analysis of corpora for Chinese and English classic works. We argue that litotes, its major argumentative functions in particular, deserves increased attention both in and outside the discipline of rhetoric.