22nd Estonian Winter School in Computer Science (EWSCS)
XXII Eesti Arvutiteaduse Talvekool

Palmse, Estonia, March 5 - 10, 2017

Student Talks 2017 (abstracts)

Syntactic analysis as an algebraic effect

Georgy Lukyanov
Southern Federal University, Institute of Mathematics, Mechanics and Computer Science, Russian Federation

It is known that parsing may be represented as effectful computation, combining effect of state with exceptions/ambiguous result and, possibly, other effects. Most parser combinators libraries (like parsec, attoparsec, etc.) use monadic approach and, in particular, monad transformers to structure effectful computation. During last time, algebraic effects and effects handlers --- an alternative approach to control computational effects has emerged and gained a lot of attention. As a part of my Master thesis, I study application of modern experimental programming language Frank (presented on POPL'17 by S. Lindley, C. McBride and C. McLaughlin) to parsers library design. I plan to tell a little bit about problem of computational effects control in general, then give a brief account on Frank, and, finally, present current state of implementation of parsers combinators library using Frank and its special features for programming with algebraic effects and handlers.

Source code of mentioned library is available on GitHub: https://github.com/geo2a/frankoparsec

Privacy-preserving stylometry

Alisa Pankova
Cybernetica AS, Estonia

Natural language processing techniques can be used to derive information about the author's background from the style of his written text. Since the writing style depend mostly on the particular writer, the authorship is the easiest to identify. Some other background of the
author, such as his home country or age, which may also affect the style, are also possible. This task is non-trivial, since even the same person uses different styles when writing a paper or an e-mail. This kind of text analysis is called stylometry.

In practice, machine learning techniques are used to classify pieces of text based on previous training on large data. It is easy to use publicly available data for training (e.g published papers), but classification of e-mails may be more problematic due to the privacy issues. In this work, we propose algorithms for different steps of privacy-preserving text classification algorithms.

Partiality and non-termination in type theory

Niccol├▓ Veltri
Tallinn University of Technology, Dept. of Software Science, Estonia

We present a general technique for representing partial functions in type theory, which corresponds to a type-theoretic reformulation of ideas by Rosolini, Mulry, Hyland, Taylor, etc. More precisely, we show how to translate the notions of dominance and partial map classifier into the language of type theory. Our main example of partial map classifier is the partiality monad, employed in type theory for the representation of possibly non-terminating computations. The dominance that specifies the partiality monad is the Sierpinski set. We show how the latter can be constructed as a standard higher inductive type and how our construction of relates to previous implementations by Chapman et al and by Altenkirch et al.

Slides from the talk [pdf]

Valid CSS! Valid XHTML 1.0 Strict Last changed November 26, 2014 19:55 Europe/Helsinki (GMT +02:00) by local organizers, ewscs17(at)cs.ioc.ee
EWSCS'17 page: http://cs.ioc.ee/ewscs/2017/