Abdelghani Bellaachia

NE-Rank: A Novel Graph-Based Keyphrase Extraction in Twitter

The massive growth of the micro-blogging service Twitter has shed the light on the challenging problem of summarizing a collection of large number of tweets. This paper attempts to extract topical key phrases that would represent topics in tweets. Due to the informality, noise, and short length of tweets, such research is nontrivial. We tackle such challenges with extensive preprocessing approach. Followed by, introduction of new features that improve topical key phrase extraction in Twitter. We start by proposing a novel unsupervised graph-based keyword ranking method, called NE-Rank, that considers word weights in addition to edge weights when calculating the ranking. Then we introduce a new approach of leveraging hash tags when extracting key phrases. We have conducted a set of experiments showing the potential of both approaches with 16% to 39% improvement for NE-Rank and 20% improvement for hash tag enhanced extraction.

Enhanced Query Expansion in English-Arabic CLIR

Arabic is a language with a particularly large vocabulary rich in words with synonymous shades of meaning. Modern Standard Arabic, which is used in formal writings, is the ancient Arabic language incorporated with loanwords derived from foreign languages. Different synonyms and loanwords tend to be used in different writings. Indeed, the Arabic composition style tends to vary throughout the Arab countries (Abdelali, 2004). Relevant documents could be overlooked when the query terms are synonyms or related to the ones used in the document collection. This could deteriorate the performance of a cross lingual information retrieval (CLIR) system. Query expansion (QE) using the document collection is the usual approach taken to enrich translated queries with context related terms. In this study, QE is explored for an English-Arabic CLIR system in which English queries are used to search Arabic documents. A thesaurus-based disambiguation approach is applied to further optimize the effectiveness of that technique. Indeed, experimental results show that QE enhanced by disambiguation gives an improved effectiveness.

Fast searching over compressed text using a new coding technique: tagged sub-optimal code (TSC)

In this paper, a new coding technique called tagged sub-optimal code (TSC) is proposed. TSC is a variable-length sub-optimal code that supports minimal prefix property. TSC technique is beneficial in many types of applications: speeding up string matching over compressed text, speeding decoding process, robustness of error detection and recovery during transmission, as well as in general-purpose integer representation code. The experimental results show that TSC is 8.9 times faster than string matching over compressed text using Huffman encoding, and 3 times faster in the decoding process.

SFLOSCAN: A biologically-inspired data mining framework for community identification in dynamic social networks

In this paper we present the first biologically inspired framework for indentifying communities in dynamic social networks. Community detection in a social network is a complex problem when interactions among members change over time. Existing community identification algorithms are limited to evaluating a snapshot of a social network at a specific time. Our algorithm evaluates social interactions as they occur over time. The user can see the detected communities at any given time. We propose a relatively simple, scalable, and novel artificial life-based algorithm named “SFloscan”. This algorithm is based on the natural phenomena of bird flocking. We model a social network as an artificial life where members flock together in a virtual two-dimensional space to form communities. We demonstrate empirically that our algorithm outperforms and overcomes the limitations of the algorithms used for community detection. We analyze the performance of SFloscan using datasets widely used in the real world.

Performance analysis of four routing protocols in sensor networks

The efficiency of sensor networks strongly depends on the routing protocol used. In this paper, we analyse four different types of routing protocols: LEACH, PEGASIS, HIT-M, and DIRECT. Sensor networks were simulated using TOSSIM simulator. Several experiments were conducted to analyse the performance of these protocols including the power consumption and overall network performance. The experimental results show that HIT-M outperforms all other protocols while PEGASIS has better performance than LEACH and DIRECT. LEACH and DIRECT have similar performance. This paper also shows the power consumption for all protocols. On the average, DIRECT has the worst power consumption.

Learning from Twitter Hashtags: Leveraging Proximate Tags to Enhance Graph-Based Keyphrase Extraction

In the micro-blogging service Twitter, the sparseness of text messages is an enormous obstacle in extracting key phrases from tweets. However, regardless of the sparseness in text, tweets include an abundant number of links in the form of hash tags. This paper investigates the possibility of leveraging hash tags in tweets to enhance the graph-based key phrase extraction. By using an auxiliary set of tweets found in hash tags, we show that we can improve extracting key phrases from tweets by augmenting the graph with a wider knowledge context. Specifically, we propose two different approaches for choosing the best hash tags links to use for enhancing graph-based key phrase extraction by either using a frequency approach or a hybrid approach that uses multiple methods for cleverly choosing the best hash tags. Experiments on the proposed approaches showed an improvement in the range of 9% to 37% over the case of ignoring the hash tag links.

Speeding up string matching over compressed text on handheld devices using tagged sub-optimal code (TSC)

Tagged suboptimal code (TSC) is a new coding technique presented in this paper to speed up string matching over compressed databases on PDAs. TSC is a variable-length suboptimal code that supports minimal prefix property. It always determines its codeword boundary without traversing a tree or lookup table. TSC technique may be beneficial in many types of applications: speeding up string matching over compressed text, speeding decoding process, as well as any general-purpose integer representation code. Experimental results show that TSC is 8.9 times faster than string matching over compressed text using Huffman encoding, and 3 times faster in the decoding process. On the other hand, the compression ratio of TSC is 6% less than that of Huffman encoding. Additionally, TSC is 14 times faster than byte pair encoding (BPE) compression process, and achieves better performance than searching over compressed text using BPE scheme on handheld devices.

Short text keyphrase extraction with hypergraphs

Graph-based ranking for keyphrase extraction has become an important approach for measuring saliency scores in text due to its ability to capture the context. By modeling words as vertices and the co-occurrence relation between words as edges, the importance of words is measured from the whole graph. However, graphs by nature can only capture the pair-wise relation between vertices. Therefore, it is not clear if graphs can capture high-order relations of more than two words. In this paper, we propose to use a hypergraph to capture high-order relations appearing in short documents, and use such information to infer better ranking of words. Additionally, we model the temporal and social attributes of short documents and discriminative weights of words into the hypergraph as weights which give us the ability of capturing recent and topical keyphrases. Furthermore, to rank vertices in the proposed hypergraph, we propose a probabilistic random walk that takes into account weights of both vertices and hyperedges. We show the effectiveness of our approach by conducting extensive experiments over two different data sets which demonstrate the robustness of the proposed approach.

Multi-document Hyperedge-based Ranking for Text Summarization

In a multi-document settings, graph-based extractive summarization approaches build a similarity graph out of sentences in each cluster of documents then use graph centrality approaches to measure the importance of sentences. The similarity is computed between each pair of sentences. However, it is not clear if such approach captures high-order relations among more than two sentences or can differentiate between descriptive sentences of the cluster in comparison with other clusters. In this paper, we propose to model sentences as hyperedges and words as vertices using a hypergraph and combine it with topic signatures to differentiate between descriptive sentences and non-descriptive sentences. To rank sentences, we propose a new random walk over hyperedges that will prefer descriptive sentences of the cluster when measuring their centrality scores. Our approach outperform a number of baseline in the DUC 2001 dataset using the ROUGE metric.

Using the maximum Mutual Information criterion to textural Feature Selection for satellite image classification

This paper presents and evaluates the use of the maximum mutual information criterion to textural feature selection for satellite image classification. Our approach is based on a recent work of Mutual Information Feature Selector Algorithm. The effectiveness of the proposed approach is evaluated on real data. In fact, the textural features are extracted using the cooccurrence matrix from two forest zones of SPOT HRV(XS) image in the region of Rabat, Morocco. The experimental tests of this study prove that the proposed approach gives a better performance for satellite image classification than classical methods such as principal components analysis (PCA) and linear discriminant analysis (LDA). The classifier used in this work is the support vectors machine (SVM).