arXiv subjects

Djelloul Ziadi

Publications and source records attributed to Djelloul Ziadi.

16 recordsLinked to original sources

arXiv2023

Parallel Tree Kernel Computation

Tree kernels are fundamental tools that have been leveraged in many applications, particularly those based on machine learning for Natural Language Processing tasks. In this paper, we devise a parallel implementation of the sequential algorithm for the computation of some tree kernels of two finite sets of trees (Ouali-Sebti, 2015). Our comparison is narrowed on a sequential implementation of SubTree kernel computation. This latter is mainly reduced to an intersection of weighted tree automata. Our approach relies on the nature of the data parallelism source inherent in this computation by deploying the MapReduce paradigm. One of the key benefits of our approach is its versatility in being adaptable to a wide range of substructure tree kernel-based learning methods. To evaluate the efficacy of our parallel approach, we conducted a series of experiments that compared it against the sequential version using a diverse set of synthetic tree language datasets that were manually crafted for our analysis. The reached results clearly demonstrate that the proposed parallel algorithm outperforms the sequential one in terms of latency.

cs.CL↗

arXiv2023

New Linear-time Algorithm for SubTree Kernel Computation based on Root-Weighted Tree Automata

Tree kernels have been proposed to be used in many areas as the automatic learning of natural language applications. In this paper, we propose a new linear time algorithm based on the concept of weighted tree automata for SubTree kernel computation. First, we introduce a new class of weighted tree automata, called Root-Weighted Tree Automata, and their associated formal tree series. Then we define, from this class, the SubTree automata that represent compact computational models for finite tree languages. This allows us to design a theoretically guaranteed linear-time algorithm for computing the SubTree Kernel based on weighted tree automata intersection. The key idea behind the proposed algorithm is to replace DAG reduction and nodes sorting steps used in previous approaches by states equivalence classes computation allowed in the weighted tree automata approach. Our approach has three major advantages: it is output-sensitive, it is free sensitive from the tree types (ordered trees versus unordered trees), and it is well adapted to any incremental tree kernel based learning methods. Finally, we conduct a variety of comparative experiments on a wide range of synthetic tree languages datasets adapted for a deep algorithm analysis. The obtained results show that the proposed algorithm outperforms state-of-the-art methods.

cs.CL↗

arXiv2021

Bottom-Up Derivatives of Tree Expressions

In this paper, we extend the notion of (word) derivatives and partial derivatives due to (respectively) Brzozowski and Antimirov to tree derivatives using already known inductive formulae of quotients. We define a new family of extended regular tree expressions (using negation or intersection operators), and we show how to compute a Brzozowski-like inductive tree automaton; the fixed point of this construction, when it exists, is the derivative tree automaton. Such a deterministic tree automaton can be used to solve the membership test efficiently: the whole structure is not necessarily computed, and the derivative computations can be performed in parallel. We also show how to solve the membership test using our (Bottom-Up) partial derivatives, without computing an automaton.

cs.FL↗

arXiv2020

Mining Frequent Itemsets: a Formal Unification

It is generally well agreed that developing a unifying theory is one of the most important issues in Data Mining research. In the last two decades, a great deal of work has been devoted to the algorithmic aspects of the Frequent Itemset (FI) Mining problem. We are motivated by the need for formal modeling in the field. Thus, we introduce and analyze, in this theoretical study, a new model for the FI mining task. Indeed, we encode the itemsets as words over an ordered alphabet, and state this problem by a formal series over the counting semiring $(\mathbb{N},+,\times,0,1)$, whose range constitutes the itemsets and the coefficients are their supports. This formalism offers many advantages in both fundamental and practical aspects: the introduction of a clear and unified theoretical framework through which we can express the main FI-approaches, the possibility of their generalization to mine other more complex objects, and their incrementalisation or parallelisation; in practice, we explain how this problem can be seen as that of word recognition by an automaton, allowing an efficient implementation in $O(|Q|)$ space and $O(|\mathcal{F}_L||Q|])$ time, where $Q$ is the set of states of the automaton used for representing the data, and $\mathcal{F}_L$ the set of prefixial longest FI.

cs.DB↗

arXiv2018

The Bottom-Up Position Tree Automaton, the Father Automaton and their Compact Versions

The conversion of a given regular tree expression into a tree automaton has been widely studied. However, classical interpretations are based upon a Top-Down interpretation of tree automata. In this paper, we propose new constructions based on the Gluskov's one and on the one of Ilie and Yu one using a Bottom-Up interpretation. One of the main goals of this technique is to consider as a next step the links with deterministic recognizers, consideration that cannot be performed with classical Top-Down approaches. Furthermore, we exhibit a method to factorize transitions of tree automata and show that this technique is particularly interesting for these constructions, by considering natural factorizations due to the structure of regular expression.

cs.FL↗

arXiv2017

Hierarchical Classification for Spoken Arabic Dialect Identification using Prosody: Case of Algerian Dialects

In daily communications, Arabs use local dialects which are hard to identify automatically using conventional classification methods. The dialect identification challenging task becomes more complicated when dealing with an under-resourced dialects belonging to a same county/region. In this paper, we start by analyzing statistically Algerian dialects in order to capture their specificities related to prosody information which are extracted at utterance level after a coarse-grained consonant/vowel segmentation. According to these analysis findings, we propose a Hierarchical classification approach for spoken Arabic algerian Dialect IDentification (HADID). It takes advantage from the fact that dialects have an inherent property of naturally structured into hierarchy. Within HADID, a top-down hierarchical classification is applied, in which we use Deep Neural Networks (DNNs) method to build a local classifier for every parent node into the hierarchy dialect structure. Our framework is implemented and evaluated on Algerian Arabic dialects corpus. Whereas, the hierarchy dialect structure is deduced from historic and linguistic knowledges. The results reveal that within {\HD}, the best classifier is DNNs compared to Support Vector Machine. In addition, compared with a baseline Flat classification system, our HADID gives an improvement of 63.5% in term of precision. Furthermore, overall results evidence the suitability of our prosody-based HADID for speaker independent dialect identification while requiring less than 6s test utterances.

cs.CL↗

arXiv2015

Bottom Up Quotients and Residuals for Tree Languages

In this paper, we extend the notion of tree language quotients to bottom-up quotients. Instead of computing the residual of a tree language from top to bottom and producing a list of tree languages, we show how to compute a set of k-ary trees, where k is an arbitrary integer. We define the quotient formula for different combinations of tree languages: union, symbol products, compositions, iterated symbol products and iterated composition. These computations lead to the definition of the bottom-up quotient tree automaton, that turns out to be the minimal deterministic tree automaton associated with a regular tree language in the case of the 0-ary trees.

cs.FL↗

arXiv2015

Rational Kernels for Arabic Stemming and Text Classification

In this paper, we address the problems of Arabic Text Classification and stemming using Transducers and Rational Kernels. We introduce a new stemming technique based on the use of Arabic patterns (Pattern Based Stemmer). Patterns are modelled using transducers and stemming is done without depending on any dictionary. Using transducers for stemming, documents are transformed into finite state transducers. This document representation allows us to use and explore rational kernels as a framework for Arabic Text Classification. Stemming experiments are conducted on three word collections and classification experiments are done on the Saudi Press Agency dataset. Results show that our approach, when compared with other approaches, is promising specially in terms of Accuracy, Recall and F1.

cs.CL↗

arXiv2015

Efficient Geometric-based Computation of the String Subsequence Kernel

Kernel methods are powerful tools in machine learning. They have to be computationally efficient. In this paper, we present a novel Geometric-based approach to compute efficiently the string subsequence kernel (SSK). Our main idea is that the SSK computation reduces to range query problem. We started by the construction of a match list $L(s,t)=\{(i,j):s_{i}=t_{j}\}$ where $s$ and $t$ are the strings to be compared; such match list contains only the required data that contribute to the result. To compute efficiently the SSK, we extended the layered range tree data structure to a layered range sum tree, a range-aggregation data structure. The whole process takes $ O(p|L|\log|L|)$ time and $O(|L|\log|L|)$ space, where $|L|$ is the size of the match list and $p$ is the length of the SSK. We present empiric evaluations of our approach against the dynamic and the sparse programming approaches both on synthetically generated data and on newswire article data. Such experiments show the efficiency of our approach for large alphabet size except for very short strings. Moreover, compared to the sparse dynamic approach, the proposed approach outperforms absolutely for long strings.

cs.LG↗

arXiv2015

Algorithm for the k-Position Tree Automaton Construction

The word position automaton was introduced by Glushkov and McNaughton in the early 1960. This automaton is homogeneous and has (||\E||+1) states for a word expression of alphabetic width ||\E||. This kind of automata is extended to regular tree expressions. In this paper, we give an efficient algorithm that computes the \Follow sets, which are used in different algorithms of conversion of a regular expression into tree automata. In the following, we consider the k-position tree automaton construction. We prove that for a regular expression \E of a size |\E| and alphabetic width ||\E||, the \Follow sets can be computed in O(||\E||\cdot |\E|) time complexity.

cs.FL↗

arXiv2015

Construction of rational expression from tree automata using a generalization of Arden's Lemma

Arden's Lemma is a classical result in language theory allowing the computation of a rational expression denoting the language recognized by a finite string automaton. In this paper we generalize this important lemma to the rational tree languages. Moreover, we propose also a construction of a rational tree expression which denotes the accepted tree language of a finite tree automaton.

cs.FL↗

arXiv2015

Root-Weighted Tree Automata and their Applications to Tree Kernels

In this paper, we define a new kind of weighted tree automata where the weights are only supported by final states. We show that these automata are sequentializable and we study their closures under classical regular and algebraic operations. We then use these automata to compute the subtree kernel of two finite tree languages in an efficient way. Finally, we present some perspectives involving the root-weighted tree automata.

cs.FL↗

arXiv2014

K-Position, Follow, Equation and K-C-Continuation Tree Automata Constructions

There exist several methods of computing an automaton recognizing the language denoted by a given regular expression: In the case of words, the position automaton P due to Glushkov, the c-continuation automaton C due to Champarnaud and Ziadi, the follow automaton F due to Ilie and Yu and the equation automaton E due to Antimirov. It has been shown that P and C are isomorphic and that E (resp. F) is a quotient of C (resp. of P). In this paper, we define from a given regular tree expression the k-position tree automaton P and the follow tree automaton F . Using the definition of the equation tree automaton E of Kuske and Meinecke and our previously defined k-C-continuation tree automaton C, we show that the previous morphic relations are still valid on tree expressions.

cs.FL↗

arXiv2014

An Efficient Algorithm for the Equation Tree Automaton via the $k$-C-Continuations

Champarnaud and Ziadi, and Khorsi et al. show how to compute the equation automaton of word regular expression $E$ via the $k$-C-Continuations. Kuske and Meinecke extend the computation of the equation automaton to a regular tree expression $E$ over a ranked alphabet $Σ$ and produce a $O(R\cdot|E|^2)$ time and space complexity algorithm, where $R$ is the maximal rank of a symbol occurring in $Σ$ and $|E|$ is the size of $E$. In this paper, we give a full description of the algorithm based on the acyclic minimization of Revuz. Our algorithm, which is performed in an $O(|Q|\cdot|E|)$ time and space complexity, where $|Q|$ is the number of states of the produced automaton, is more efficient than the one obtained by Kuske and Meinecke.

cs.FL↗

arXiv2013

Extended to Multi-Tilde-Bar Regular Expressions and Efficient Finite Automata Constructions

Several algorithms have been designed to convert a regular expression into an equivalent finite automaton. One of the most popular constructions, due to Glushkov and to McNaughton and Yamada, is based on the computation of the Null, First, Last and Follow sets (called Glushkov functions) associated with a linearized version of the expression. Recently Mignot considered a family of extended expressions called Extended to multi-tilde-bar Regular Expressions (EmtbREs) and he showed that, under some restrictions, Glushkov functions can be defined for an EmtbRE. In this paper we present an algorithm which efficiently computes the Glushkov functions of an unrestricted EmtbRE. Our approach is based on a recursive definition of the language associated with an EmtbRE which enlightens the fact that the worst case time complexity of the conversion of an EmtbRE into an automaton is related to the worst case time complexity of the computation of the Null function. Finally we show how to extend the ZPC-structure to EmtbREs, which allows us to apply to this family of extended expressions the efficient constructions based on this structure (in particular the construction of the c-continuation automaton, the position automaton, the follow automaton and the equation automaton).

cs.FL↗

arXiv2013

On the transition reduction problem for finite automata

We are interested in the problem of transition reduction of nondeterministic automata. We present some results on the reduction of the automata recognizing the language $L(E_n)$ denoted by the regular expression $E_n=(1+\varepsilon)... (2+\varepsilon)... (3+\varepsilon)... (n+\varepsilon)$. These results can be used in the general case of the transition reduction problem.

cs.FL↗