Episciences.org
Not a member yet
6707 research outputs found
Sort by
Adapting vs. Pre-training Language Models for Historical Languages
No full textAs large language models such as BERT are becoming increasingly popular in Digital Humanities (DH), the question has arisen as to how such models can be made suitable for application to specific textual domains, including that of 'historical text'. Large language models like BERT can be pretrained from scratch on a specific textual domain and achieve strong performance on a series of downstream tasks. However, this is a costly endeavour, both in terms of the computational resources as well as the substantial amounts of training data it requires. An appealing alternative, then, is to employ existing 'general purpose' models (pre-trained on present-day language) and subsequently adapt them to a specific domain by further pre-training. Focusing on the domain of historical text in English, this paper demonstrates that pre-training on domain-specific (i.e. historical) data from scratch yields a generally stronger background model than adapting a present-day language model. We show this on the basis of a variety of downstream tasks, ranging from established tasks such as Part-of-Speech tagging, Named Entity Recognition and Word Sense Disambiguation, to ad-hoc tasks like Sentence Periodization, which are specifically designed to test historically relevant processing
Actris 2.0: Asynchronous Session-Type Based Reasoning in Separation Logic
No full textMessage passing is a useful abstraction for implementing concurrent programs.For real-world systems, however, it is often combined with other programmingand concurrency paradigms, such as higher-order functions, mutable state,shared-memory concurrency, and locks. We present Actris: a logic for provingfunctional correctness of programs that use a combination of the aforementionedfeatures. Actris combines the power of modern concurrent separation logics witha first-class protocol mechanism -- based on session types -- for reasoningabout message passing in the presence of other concurrency paradigms. We showthat Actris provides a suitable level of abstraction by proving functionalcorrectness of a variety of examples, including a channel-based merge sort, achannel-based load-balancing mapper, and a variant of the map-reduce model,using concise specifications. While Actris was already presented in aconference paper (POPL'20), this paper expands the prior presentationsignificantly. Moreover, it extends Actris to Actris 2.0 with a notion ofsubprotocols -- based on session-type subtyping -- that permits additionalflexibility when composing channel endpoints, and that takes full advantage ofthe asynchronous semantics of message passing in Actris. Soundness of Actris2.0 is proven using a model of its protocol mechanism in the Iris framework. Wehave mechanised the theory of Actris, together with custom tactics, as well asall examples in the paper, in the Coq proof assistant
Лексико-грамматические сдвиги в узусе беларуского языка в постсоветское время
No full textÉpisciences - SlovoУтварэнне самастойнай дзяржавы Рэспубліка Беларусь (1991) выклікала змены і ў узусе беларускай мовы, скіраваныя перш за ўсё на яе дывергенцыю з рускай моваю, асабліва ў галіне лексікі і граматыкі. Канкурэнцыя двух варыянтаў беларускай літаратурнай мовы (афіцыйнага – наркамаўкі і неафіцыйнага – тарашкевіцы) вядзе да ўзаемапранікнення іх нормаў і стыхійнаму фарміраванню ў актыўных карыстальнікаў своеасаблівага кампраміснага падыходу, які акумулюе ў сабе найбольш рацыянальныя рысы абодвух варыянтаў.The formation of an independent state of the Republic of Belarus (1991) has also brought about changes in the Belarusian language, primarily aimed at its divergence with the Russian language, especially in the field of vocabulary and grammar. The competition of two variants of the Belarusian literary language (the official one – narkamaǔka – and the unofficial one – taraškevica) leads to the interpenetration of their norms and the spontaneous formation of a kind of compromise among active users, which accumulates the most rational features of both variants.La proclamation de l’indépendance de la république du Bélarus (1991) a provoqué des changements dans la langue bélarussienne, visant principalement à diverger de la langue russe, en particulier dans le domaine du vocabulaire et de la grammaire. La concurrence de deux variantes du bélarussien littéraire (l’officielle est la narkamaŭka et la non-officielle est la taraškievica) conduit à l'interpénétration de leurs normes et à la formation spontanée auprès des bélarussophones actifs d’une approche de compromis dans leur usage de la langue qui concilie les traits les plus probants des deux variantes.Образование самостоятельного государства Республика Беларусь (1991) вызвало изменения и в узусе беларуского языка, направленные прежде всего на его дивергенцию с русским языком, особенно в области лексики и грамматики. Конкуренция двух вариантов беларуского литературного языка (официального – наркамавка и неофициального – тарашкевица) ведет к взаимопроникновению их норм и стихийному формированию у активных пользователей своеобразного компромиссного подхода, который аккумулирует в себе наиболее рациональные черты обоих вариантов
Quotients, inductive types, and quotient inductive types
No full textThis paper introduces an expressive class of indexed quotient-inductivetypes, called QWI types, within the framework of constructive type theory. Theyare initial algebras for indexed families of equational theories with possiblyinfinitary operators and equations. We prove that QWI types can be derived fromquotient types and inductive types in the type theory of toposes with naturalnumber object and universes, provided those universes satisfy the WeaklyInitial Set of Covers (WISC) axiom. We do so by constructing QWI types ascolimits of a family of approximations to them defined by well-foundedrecursion over a suitable notion of size, whose definition involves the WISCaxiom. We developed the proof and checked it using the Agda theorem prover
Robustly Self-Ordered Graphs: Constructions and Applications to Property Testing
No full textA graph is called self-ordered (a.k.a asymmetric) if the identitypermutation is its only automorphism. Equivalently, there is a uniqueisomorphism from to any graph that is isomorphic to . We say that is robustly self-ordered if the size of the symmetric differencebetween and the edge-set of the graph obtained by permuting using anypermutation is proportional to the number of non-fixed-points of. In this work, we initiate the study of the structure, construction andutility of robustly self-ordered graphs. We show that robustly self-ordered bounded-degree graphs exist (inabundance), and that they can be constructed efficiently, in a strong sense.Specifically, given the index of a vertex in such a graph, it is possible tofind all its neighbors in polynomial-time (i.e., in time that ispoly-logarithmic in the size of the graph). We also consider graphs of unbounded degree, seeking correspondinglyunbounded robustness parameters. We again demonstrate that such graphs (oflinear degree) exist (in abundance), and that they can be constructedefficiently, in a strong sense. This turns out to require very different tools.Specifically, we show that the construction of such graphs reduces to theconstruction of non-malleable two-source extractors (with very weak parametersbut with some additional natural features). We demonstrate that robustly self-ordered bounded-degree graphs are usefultowards obtaining lower bounds on the query complexity of testing graphproperties both in the bounded-degree and the dense graph models. One of theresults that we obtain, via such a reduction, is a subexponential separationbetween the query complexities of testing and tolerant testing of graphproperties in the bounded-degree graph model.Comment: Slightly modified and revised version of a CCC 2021 paper that also appeared on ECCC 27: 149 (2020). TheoretiCS 1 (2022), Article 1, 92 page
Enumerating Answers to First-Order Queries over Databases of Low Degree
No full textA class of relational databases has low degree if for all , all butfinitely many databases in the class have degree at most , where is the size of the database. Typical examples are databases of boundeddegree or of degree bounded by . It is known that over a class of databases having low degree, first-orderboolean queries can be checked in pseudo-linear time, i.e.\ for all in time bounded by . We generalize this result byconsidering query evaluation. We show that counting the number of answers to a query can be done inpseudo-linear time and that after a pseudo-linear time preprocessing we cantest in constant time whether a given tuple is a solution to a query orenumerate the answers to a query with constant delay
On the -domination number of cylinders with small cycles
No full textDomination-type parameters are difficult to manage in Cartesian productgraphs and there is usually no general relationship between the parameter inboth factors and in the product graph. This is the situation of the dominationnumber, the Roman domination number or the -domination number, among others.Contrary to what happens with the domination number and the Roman dominationnumber, the -domination number remains unknown in cylinders, that is, theCartesian product of a cycle and a path and in this paper, we will compute thisparameter in the cylinders with small cycles. We will develop two algorithmsinvolving the matrix product that will allow us to compute thedesired values of , with and .We will also pose a conjecture about the general formulae for the-domination number in this graph class.Comment: 15 pages, 1 figur
Arkady Onishchik: on his life and work on supersymmetry
No full textSelected stories about the life of A. L. Onishchik, and a review of hiscontribution to the classification of non-split supermanifolds, in particular,supercurves a.k.a. superstrings; his editorial and educational work. A briefoverview of his and his students' results in supersymmetry, and their impact onother researchers. Several open problems growing out of Onishchik's research are presented, someof them are related with odd parameters of deformations and non-holonomicstructures of supermanifolds important in physical models, such as Minkowskisuperspaces and certain superstrings.Comment: 24p
A generalization of certain associated Bessel functions in connection with a group of shifts
No full textConsidering the kernel of an integral operator intertwining two realizationsof the group of motions of the pseudo-Euclidian space, we derive two formulasfor series containing Whittaker's functions or Weber's parabolic cylinderfunctions. We can consider this kernel as a special function. Some particularvalues of parameters involved in this special function are found to coincidewith certain variants of Bessel functions. Using these connections, we alsoestablish some analogues of orthogonality relations for Macdonald and Hankelfunctions
Local derivations of semisimple Leibniz algebras
No full textWe prove that every local derivation on a complex semisimplefinite-dimensional Leibniz algebra is a derivation