Mostafa Aref

Web document clustering approach using wordnet lexical categories and fuzzy clustering

Web mining is defined as applying data mining techniques to the content, structure, and usage of Web resources. The three areas of Web mining are commonly distinguished: content mining, structure mining, and usage mining. In all these areas, a wide range of general data mining techniques, in particular association rule discovery, clustering, classification, and sequence mining, are employed and developed further to reflect the specific structures of Web resources and the specific questions posed in Web mining. In this paper, we introduced a Web document clustering approach that uses WordNet lexical categories and fuzzy c-means algorithm to improve the performance of clustering problem for Web document. Experiments show that fuzzy c-means algorithm achieves great performance optimization with comparison with the recent algorithms for document clustering.

Fuzzy Document Clustering Approach using WordNet Lexical Categories

Text mining refers generally to the process of extracting interesting information and knowledge from unstructured text. This area is growing rapidly mainly because of the strong need for analysing the huge and large amount of textual data that reside on internal file systems and the Web. Text document clustering provides an effective navigation mechanism to organize this large amount of data by grouping their documents into a small number of meaningful classes. In this paper we proposed a fuzzy text document clustering approach using WordNet lexical categories and Fuzzy c-Means algorithm. Some experiments are performed to compare efficiency of the proposed approach with the recently reported approaches. Experimental results show that Fuzzy clustering leads to great performance results. Fuzzy c-means algorithm overcomes other classical clustering algorithms like k-means and bisecting k-means in both clustering quality and running time efficiency.

Extracting N-gram terms collocation from tagged Arabic corpus

Information Extraction (IE) is one of the most important Natural Language Processing (NLP) applications, which extracts information such as Named-Entities (NE) and collocation of terms from the corpus. Collocation is a sequence of terms that co-occur together in the corpus. In Arabic Information Extraction, there are many problems because of the complex of Arabics grammar and ambiguity. In general, in linguistics research, the more efficient corpus is the one annotated by Part of Speech Tagging (POST). In this paper, we propose a prototype that extracts collocation of N-gram words (from 2-6 gram) based on the sequence of POST from Arabic Quran corpus. This approach extracts the collocation of N-gram words by matching the input structured pattern of Arabic language versus the Part of Speech Tagging of Quran corpus. The system enables users to select a sequence of tags (2-6 gram) and scope of the corpus source (whole Quran Corpus or specific Surah). To show how the system is beneficial for linguistic research, a set of experiments has been conducted in different scenarios.

Generic opponent modelling approach for real time strategy games

One of the fundamental and challengeable research areas in Real Time Strategy (RTS) games is opponent modelling. Most current approaches to opponent modelling pretended inefficiency. They are either computationally expensive or required a numerous amount of online gameplays to start learn successful models. Unfortunately, most successful approaches also were game specific. They mainly depend on the experts knowledge of the game. In this paper, a generic and adaptive opponent modelling approach for RTS games is proposed. It is a completely automated approach for learning the highly informative features of the opponents behavior of any RTS game. Inspired by the case-based reasoning technique, a case base of different opponent models is constructed in the approach offline phase. The online phase (during gameplay) utilizes only this model base for opponent classification. To better cope with opponents that switch strategies, the approach keeps track of the performance after classification. To show how the proposed approach is beneficial, a case study called SPRING game case-study is presented.

Intelligent online case-based planning agent model for real-time strategy games

Research in learning and planning in real-time strategy (RTS) games is very interesting in several industries such as military industry, robotics, and most importantly game industry. A recent published work on online case-based planning in RTS Games does not include the capability of online learning from experience, so the knowledge certainty remains constant, which leads to inefficient decisions. In this paper, an intelligent agent model based on both online case-based planning (OLCBP) and reinforcement learning (RL) techniques is proposed. In addition, the proposed model has been evaluated using empirical simulation on Wargus (an open-source clone of the well known RTS game Warcraft 2). This evaluation shows that the proposed model increases the certainty of the case base by learning from experience, and hence the process of decision making for selecting more efficient, effective and successful plans.

Ontology-based Query Expansion for Arabic Text Retrieval

The semantic resources are important parts in the Information Retrieval (IR) such as search engines, Question Answering (QA), etc., these resources should be available, readable and understandable. In semantic web, the ontology plays a central role for the information retrieval, which use to retrieves more relevant information from unstructured information. This paper presents a semantic-based retrieval system for the Arabic text, which expands the input query semantically using Arabic domain ontology. In the proposed approach, the search engine index is represented using Vector Space Model (VSM), and the Arabics place nouns domain ontology has been used which constructed and implemented using Web Ontology Language (OWL) from Arabic corpus. The proposed approach has been experimented on the Arabic Quran corpus, and the experiments show that the approach outperforms in terms of both precision and recall the traditional keyword- based methods.

A model for generating Arabic text from semantic representation

Text Generation is a challenging task in Natural Language Processing (NLP). Its purpose is to generate grammatically correct text from machine representation source such as a knowledge base. One of the most recent semantic representation is Rich Semantic Graph (RSG). It exploits not only the semantic representation techniques but also the Language structure and writing styles. Our work is a part of an ongoing research to create an abstractive summary for a single input document in the Arabic Language. The abstractive summary is generated through three modules; converting the input Arabic text into an RSG, then performing Graph Reduction, and finally generating the summarized text from the reduced graph. This is achieved with the aid of a domain Ontology. In this paper, we are illustrating the architecture of the third module, which works on generating Arabic text from RSG using Ontology.

Ontology-based model for Arabic lexicons: An application of the Place Nouns in the Holy Quran

Ontology is an explicit specification of conceptualization. It defines the terms with specified relationships between them and can be interpreted by both humans and computers. In general, there are scare semantic resources for Arabic language especially in Arabic ontologies. These semantic resources are very essential components in both Information Retrieval and Natural Language Processing applications like search engines, question answering, machine translation, etc. In recent years, many researchers are interested in building Arabic sematic resources, which are then can be exploited by others to build Arabic sematic applications. Recently, a proposed ontological model for “Time Nouns” vocabulary in the Holy Quran was introduced. To share towards building an integrated and unified ontology for Arabic language, in this paper, we are proposing an ontology-based model for Arabic language vocabulary associated with “Place Nouns” in the Holy Quran. This ontology is represented by the Web Ontology Language (OWL), which is the standard language for the semantic web. The ontology will be useful in the knowledge of the Islamic learning, linguistics researches, and Semantic Web applications.

An efficient filtering technique for super-graph query processing

Recently, a lot of applications depend on data modeled by graphs. Efficient query processing over graph databases serves these applications. Having a graph query q, super-graph query processing finds all the graphs g in a database of graphs D where g is contained in q (gsubeq). Because graph databases contain a lot of graphs and because sub-graph isomorphic tests are NP-complete, an indexing-based technique should be adopted. In this paper we propose an efficient technique for processing supergraph queries. The technique consists of an index called eIndex and a query processing algorithm. Given a query graph, the database is filtered to generate candidate graphs. While filtering, the proposed technique takes into consideration the full structure of the database graphs besides considering the frequent fragments. Through polynomial time algorithms, the proposed technique reduces the subgraph isomorphism tests required for query processing and hence the total processing time is reduced. The results show that building eIndex takes much less time and space than other methods.

A Convolutional Neural Network Model for Emotion Detection from Tweets.

Sentiment analysis and emotion recognition are major indicators of society trends toward certain topics. Analyzing opinions and feelings helps improving the human-computer interaction in several fields ranging from opinion mining to psychological concerns. This paper proposes a deep learning model for emotion detection from short informal sentences. The model consists of three Convolutional Neural Networks (CNNs). Each CNN contains a convolutional layer and a max-pooling layer, followed by a fully-connected layer for classifying the sentences into positive or negative. The model employs the word vector representation as textual features, which works on random initialization for the word vectors, and are set to be trainable and updated through the model training phase. Eventually, task-specific vectors are generated as the model learns to distinguish the meaning of words in the dataset. The model has been tested on the Stanford Twitter Sentiment dataset for classifying sentiment into two classes positive and negative. The presented model achieved to record 80.6% accuracy as a prove that even with randomly initialized word vectors, it can work very well in text classification tasks when trained with CNNs.