ESSLLI 2014 will feature four public evening lectures. We are happy to announce the following invited speakers:
August 12th - 7.00 pm: Hannes Leitgeb (München)
The Humean Thesis on Belief
I am going to make precise, and assess, the following thesis on (all-or-nothing) belief and degrees of belief: It is rational to believe a proposition just in case it is rational to have a stably high degree of belief in it. I will start with some historical remarks, which are going to motivate calling this postulate the "Humean thesis on belief". Once the thesis has been formulated in formal terms, it is possible to derive conclusions from it. Three of its consequences that I will highlight in particular are: doxastic logic; an instance of what is sometimes called the Lockean thesis on belief; and a simple qualitative decision theory
August 14th - 7.00 pm: Uwe Mönnich (Tübingen)
A Model-Theoretic perspective on Chomsky-Schützenberger theorems
The development of model-theoretic syntax in terms of a logical speciffication language that is both expressive and manageable presents problems for mono-level approaches. All these approaches suffer from a lack of expressive power in that the family of regular tree languages properly includes all other language families that are captured by the logical formalisms that have been considered in model-theoretic syntax. The talk sketches a solution to these problems by integrating a formally unified notion of grammar morphism into the framework of model-theoretic syntax. The approach we present follows Courcelle's extension of Rabin's method of model-theoretic interpretation. This extended version of the classical method of semantic interpretation serves to carve out the exact position of recent linguistic theories within the family of mildly context-sensitive languages. In the process of determining this position we rely on the linguistically significant affinities between restrictions on the logical formulae defining the relations in the target structures of logical transductions and the specific rule formats of these linguistic theories. These affinities provide then the basis for extending the generator characterization of context free languages due to Chomsky and Schüutzenberger to more expressive accounts of natural language syntax.
August 19th - 7.15 pm: Stefan Evert (Erlangen)
A multivariate approach to linguistic variation and distribution
Most linguistic features exhibit considerable variability in their frequency distribution across different time periods, language varieties, regions, speakers, text types, topic domains, etc. Such language variation is of great interest to corpus linguistics, sociology, dialectology, historical linguistics, and many other fields of research. However, it can also be a confounding factor for studies that focus on a specific contrast, obscuring or distorting the relevant differences, and needs to be accounted for in a statistical analysis.
In many cases, several related or unrelated linguistic features follow a similar pattern of variation. Such correlations can be analysed with multivariate statistical techniques – such as factor analysis, principal component analysis and correspondence analysis – in a completely unsupervised manner, revealing insights about the dimensions of linguistic variation as well as the distribution of individual features. A well-known example of this inductive approach is Doug Biber's multidimensional register analysis. Computational linguists use similar techniques in order to identify topic domains and obtain a distributional representation of word meaning.
In my lecture, I will present case studies of multivariate analyses, highlighting both their common core and individual differences between the approaches. I will then focus on several important methodological problems, in particular (i) the possibility of introducing researcher bias through the choice of text samples and features and (ii) the difficulty of assigning a meaningful interpretation to the identified dimensions of variation. Recent findings suggest that a minimal amount of supervision – e.g. in the form of language-external categories – can help to guide the multivariate analysis towards interpretable dimensions representing specific aspects of language variation.