Mohammad Hajjar

Efficient super-peer-based queries routing

Peer-to-peer (P2P) Data-sharing systems now generate a significant portion of internet traffic. P2P systems have emerged as a popular way to share huge volumes of data. Requirements for widely distributed information systems supporting virtual organizations have given rise to a new category of P2P systems called schema-based. In such systems each peer is a database management system in itself, exposing its own schema. In such settings, the main objective is the efficient search across peer databases by processing each incoming query without overly consuming bandwidth. The usability of these systems depends on effective techniques to find and retrieve data; however, efficient and effective routing of content-based queries is an emerging problem in P2P networks. In this paper, we propose an architecture, based on (super-)peers, and we focus on query routing. Our approach considers that (super-)Peers having similar interests are grouped together for an efficient query routing method. In such groups, called knowledge-super-peers (KSP), super-peers submit queries that are often processed by members of this group. A KSP is a specific super-peer which contains knowledge about: 1. its super-peers and 2. the others super-peers. Knowledge is extracted by using data mining techniques (e.g. decision tree algorithms) starting from queries of peers that transit on the network. The advantage of this distributed knowledge is that, it avoids to making semantic mapping, between heterogeneous data sources owned by (super-)peers, each time the system decides to route query to other (super-)peers. The set of KSP improves the robustness in queries routing mechanism and scalability in P2P Network. Compared with a baseline approach, our proposal shows a better performance with respect to important criteria such as response time, precision and recall.

A Global Knowledge for Information Retrieval in P2P Networks

In traditional P2P networks, such as Gnutella, peers propagate query messages towards the resource holders by flooding them through the network. However, it is a costly operation since it consumes node and link resources excessively, which are often unnecessarily. There is no reason, for example, for a peer to receive a query message if the peer has no matching resource or is not on the path to a peer holding a matching resource. However, how to quickly discover the right resource in a large-scale P2P network without generating too much network traffic and with minimum possible time remain highly challenging. In this paper, we propose a new peer-to-peer (P2P) search method aiming at exploiting data mining concepts (Decision Tree) to improve search performance for information retrieval in P2P network. We use a PDMS system, which aims to combine a Super-Peer (SP) based network with the capability of managing a data model attached to the peers in the form of relational, xml, or object schemes. Each SP is connected to a Global-Knowledge-Super-Peer (GKSP) that operates with an index (decision tree), to predict the relevant domains (super-peers), to answer a given query. Compared with a super peer-based approach, our proposal architectures show the effect of the data mining with better performance with respect to response time, number of messages, precision and recall.

An Improved Structured and Progressive Electronic Dictionary for the Arabic Language: iSPEDAL

In this article, we propose an improved structured and progressive electronic dictionary for the Arabic language (iSPEDAL) which can be presented in the form of a relational database or in the form of an XML document which can be easily exploitable using suitable query languages. Indeed, many Arabic dictionaries are found but are not structured and not directly exploitable since they are in flat textual files form. iSPEDAL doesn’t contain any duplicated data (roots, prefixes, suffixes, the infixes, the patterns and the derived words). Moreover, for a given word, it provides links to its root, to their associated affixes, and to its patterns. iSPEDAL is supplied automatically from one or several traditional textual dictionaries and is enriched permanently with any Arabic textual corpus using system that we built. This system is composed of a Parser, a Selector, a Classifier, an Extractor, a Comparator, an Analyzer, and a Validator. The Parser allows the transformation of a textual source (dictionary or textual corpus) into a set of words. The Selector determines if a word is new or already exists in iSPEDAL. The Classifier allows to classify a given word and to add it to iSPEDAL as a root or as a derived word. The Extractor uses the Arabic extraction method to deduce the root of all words arriving to this component without their root or any indication about their root. The Comparator permits to avoid duplication of roots, affixes or patterns in iSPEDAL. The Analyzer allows the extraction of the affixes and the pattern from a derived word and of its root. The Validator can validate the information (word, root, patterns, and affixes) before adding to iSPEDAL database. This dictionary can be used to evaluate the information extraction methods from an Arabic document, given that; the vocabulary of the Arabic language is essentially built from the roots.

Data mining effect in peer-to-peer queries routing

Data mining has been used to extract hidden information from large databases. In peer-to-peer context, a challenging problem is how to find the appropriate peer to deal with a given query without overly consuming bandwidth? Different methods proposed routing strategies of queries taking into account the p2p network at hand. We consider an unstructured P2P system based on an organization of peers around super-peers that are connected to meta-super-peer according to their semantic domains. This paper integrates decision trees in P2P architectures for predicting Query-Suitable super-peers representing a community of peers where one among them is able to answer the given query. In fact by analyzing the queries log file, we construct a predictive model that avoids flooding queries in the p2p network by predicting the appropriate super-peer, and hence the peer to answer the query. A challenging problem in a schema-based peer-to-peer (P2P) system is how to locate peers that are relevant with respect to a given query. In this paper, we propose an architecture, based on (super-) peers, and we focus on query routing. Our approach considers that (super-) peers having similar interests are grouped together for an efficient query routing method. In such groups, called Meta-Super-Peers (MSP), super-peers submit queries that are often processed by members of this group. A MSP is a specific super-peer which contains knowledge about: 1. its super-peers and 2. The others MSP. Knowledge is extracted by using data mining techniques (e.g. decision tree algorithms) starting from queries of peers that transit on the network. The advantage of this distributed knowledge is that, it avoids making semantic mapping, between heterogeneous data sources owned by (super-)peers, each time the system decides to route query to other (super-)peers. The set of MSP improves the robustness in queries routing mechanism and scalability in P2P Network. Compared with a baseline approach, our proposal architectures show the effect of the data mining with better performance with respect to response time and precision.

Towards the Enrichment of Arabic WordNet with Big Corpora.

This paper presents a method aiming to enrich Arabic WordNet with semantic clusters extracted from a large general corpus. The Arabic language being poor in open digital linguistic resources, we built such a corpus (more than 7.5 billion words) with ad-hoc tools. We then applied GraPaVec, a new method for word vectorization using automatically generated frequency patterns, as well as state-of-the-art Word2Vec and Glove methods. Word vectors were fed to a Self Organizing Map neural network model; the clusterings produced were then compared for evaluation with Arabic WordNet existing synsets (sets of synonymous words). The evaluation yields a F-score of 82.1% for GrapaVec, 55.1% for Word2Vec’s Skipgram, 52.2% for CBOW and 56.6% for Glove, which at least shows the interest of the context that GraPaVec takes into account. We end up by discussing parameters and possible biases.

Survey and Classification of Methods for Building a Semantic Annotation

Though Arabic is one of the five most spoken languages, little work has been done on building Arabic semantic resources. Currently, there is no agreed-upon method for building such a reliable Arabic semantic resource. The purpose of this paper is to present a comprehensive survey of different methods for building or enriching Arabic semantic resources; to study and analyze each method; and to categorize the methods according to their properties. This work should contribute to the definition of new methods and help researchers on Arabic semantics to fit their work in the panel of existing ones.