Regular tree language recognition with static information - CDuce
Recommend Documents
words (or keywords) were strongly indicative of a specific language. For example, the ..... and bottom 600 ranking keywords for each target language in the.
May 16, 2017 - Accompanying the report is a blog [2], discussing the process of the ..... cultural differences, a solution could be to customize the system to use ...
tern algorithms and for the construction and minimization of finite automata. Both ... spondence between regular tree grammars and finite tree automata.
We show that a tree substitution grammar (TSG) induced with a collapsed ... of
groups have had success parsing with tree substitution grammars (TSGs) that.
Jan 10, 2014 - ... to verify the validity of the continuity equation: (εu. µ. );µ = 0, where the semicolon denotes covariant derivative. The conservation equation, T.
level using key-value pair concept error rate as the evaluation metric. When word confidence scores were integrated into the understanding component, the ...
on word and character uni-gram probabilities. 2 Task 1.5: Entity ... gorithm[1], as it was implemented in the C Minimal Perfect Hashing Library. (CMPH) V2.02 in ...
AbstractâSign language recognition systems suffer from the problem of signer .... the fixed number of iterations that the messages' changes fall below a ...
First, we replace the phone tokenizer by a powerful speech-to-text sys- tem. Second, we use a discriminative support vector machine for language modeling.
scene recognizer and evaluated it with a dataset of 20 scenes and 100+ objects. ..... and the number of times when scene S occurs in Gigaword,. #(S). Then we ...
Nov 13, 2014 - [34â. 38], with a focus on λ but not on the full magnetic response function. arXiv:1411.3630v1 [cond-mat.mtrl-sci] 13 Nov 2014 ...
Nov 13, 2014 - term and a âBerry curvatureâ correction, which we believe ... sea like term since it involves an energy integration up to the Fermi energy of a ...
Aug 30, 2017 - Boolean finite automata, which allow for the direct and natural ... deterministic finite automata (NFA) [20] and then use automata operations to.
tions in a representation called Spatial Pattern Templates (SPT), which ... on their relative location and classifier output. Ideally, such ..... ICCV Workshops (2011)
Consistency is preserved if the spectral radius of matrix is smaller than one (Wetherell, 1980). It is important to remark that two stochastic languages are identical ...
the cross domain European project DictaSign [23] aims to produce a multi-lingual data set suitable for ..... both extended motion history images and Non-rigid Registration using Free-Form ..... DictaSign Project: Dictasign project website (2010).
In this paper we investigate Indian spoken Language recognition for 3 languages namely Kannada, Tamil and Malayalam with the usage of RASTA-PLP for.
ence and Software Engineering, The University of Western Australia, Craw- ley, WA 6009, Australia, ... Almost all existing hyperspectral face recognition tech- niques perform ...... Pakistan and the M.S. degree in Electronics and. Computer ...
Nov 2, 2009 - model-checking [BMV03] ) and many tools (ProVerif [Bla01], Athena ... The Proverif tool [Bla01] allows an unbounded number of sessions but ...
Bounded methods (e.g., Hampi [14], Kudzu [20], and CFGAnalyzer [1]) unroll the constraints to a given length bound, encode the unrolled problem as a set of ...
Research Scholar, Mother Teresa Women's University, India. E-mail:[email protected]. Dr. R.S. Rajesh. Reader, Department of Computer Science and ...
Page 1 ... have shown that the grammar created with this approach consistently outperformed the ..... The location constraints significantly reduce the fan-out at.
2.Net Speech Group. Microsoft Corporation, One Microsoft Way, Redmond, WA 98052, USA ... for example, a âShowFlightâ command like âList the flights from Seattle to .... checks the character set defined by the atom of the piece. If it is the dig
Regular tree language recognition with static information - CDuce
Indeed, this information allows in many cases to compute the result (that ... cannot take this kind of static knowledge into ac- ... But if we use the information given by the function in- .... The set of trees is written V . ... propagate information downwards: the context of a loca- .... NUA does not need to be fully built at compile time.
Regular tree language recognition with static information Alain Frisch D´epartement d’Informatique ´ Ecole Normale Sup´erieure, Paris, France [email protected]
Abstract
tern matching in CDuce1 . To simplify the presentation, the paper studies only a restricted form of pattern matching, without capture variable and with a very simple kind of trees. Of course, our implementation handles capture variables and the full set of types and patterns constructors in CDuce. In the simplified form, the pattern matching problem is a recognition problem, namely deciding whether a tree v belongs to a regular tree language X or not.
This paper presents our compilation strategy to produce efficient code for pattern matching in the CDuce compiler, taking into account static information provided by the type system. Indeed, this information allows in many cases to compute the result (that is, to decide which branch to consider) by looking only at a small fragment of the tree. Formally, we introduce a new kind of deterministic tree automata that can efficiently recognize regular tree languages with static information about the trees and we propose a compilation algorithm to produce these automata.
1
1. If the regular language is given by a tree automaton, a top-down recognition algorithm may have to backtrack, and the recognition time is not linear in the size of the input tree.
Introduction
2. It is well-known that any tree automaton can be transformed into an equivalent bottom-up deterministic automaton, which ensures linear execution time; the size of the automaton may be huge even for simple languages, which can make this approach unfeasible in practice.
Emergence of XML has given tree automata theory a renewed importance[Nev02]. Indeed, XML schema languages such as DTD, XML-Schema, Relax-NG describe more or less regular languages of XML documents (considered as trees). Consequently, recent XML-oriented typed programming languages such as XDuce [Hos00, HP02], CDuce [BCF03, FCB02], Xtatic [GP03] have type algebras where types denote regular tree languages. The type system of these languages relies on a subtyping relation, defined as the inclusion of the regular languages denoted by types, which is known to be a decidable problem (new and efficient algorithms have been designed for this purpose, and they behave well in practice). An essential ingredient of these languages is a powerful pattern matching operation. A pattern is a declarative way to extract information from an XML tree. Because of this declarative nature, language implementors have to propose efficient execution models for pattern matching. This paper describes our approach in implementing pat-
3. The static type system of the language provides an upper approximation for the type of the matched tree v, that is some regular language X0 such that v is necessarily in X0 . Taking this information into account, it should be possible to avoid looking at some subtree of v. However, classical bottom-up tree automata are bound to look at the whole tree, and they cannot take this kind of static knowledge into account. Let us give an example to illustrate the last point. Consider the following CDuce program: 1 CDuce is available for download at http://www.cduce.org/.
1
of information from trees. More precisely, a NUA will skip a subtree if and only if looking at this tree does not bring any additional relevant information, considering the initial static information and the information already gathered during the beginning of the run.