Waqas Anwar

A Statistical Based Part of Speech Tagger for Urdu Language

In this paper we present a pioneering step in designing n-gram based part of speech tagger for the Urdu language. In the last few years part of speech tagging work has been done in the area of supposed English, South Asian and European languages. In this paper our focus of attention is on the disambiguation problem (to assign the accurate tag for every word of a set of possible tags). Our approach employs n-gram Markov model, train from annotated Urdu corpus and assigns possible tags to text. The proposed n-gram part of speech tagger has been tested which achieved state of the art performance of 95.0%. Furthermore, we check our experiment results of two type of tagset. Along the way, we apply evaluation method that shows how significant our experiment results are. Besides, we present the error analysis (confusion matrix) and show the tagging example of Urdu tagging. We also present overview of Urdu language. The contribution of our work is an initial step of statistical based Urdu part of speech tagger.

Contextual advertising using keyword extraction through collocation

Contextual advertising is to place an ad in the web page by analyzing content of the web page. Different contextual advertising approaches are used to make a content match. We analyzed different contextual approaches in this paper to find out how the content-match is performed. Some Contextual advertising approaches use relevance factor between content of ad and web page by using machine translation strategies, machine learning techniques, while some use ontologies to find out an optimal match between an ad and web page. We proposed a strategy which is based on collocation between different words. It finds collocation between selected words in a web page and than match it with the keyword of an ad.

A Comparison of Moments-Based Logo Recognition Methods

Logo recognition is an important issue in document image, advertisement, and intelligent transportation. Although there are many approaches to study logos in these fields, logo recognition is an essential subprocess. Among the methods of logo recognition, the descriptor is very vital. The results of moments as powerful descriptors were not discussed before in terms of logo recognition. So it is unclear which moments are more appropriate to recognize which kind of logos. In this paper we find out the relations between logos with different transforms and moments, which moments are fit for logos with different transforms. The open datasets are employed from the University of Maryland. The comparisons based on moments are carried out from the aspects of logos with noise, and rotation, scaling, rotation and scaling.

Morpheme Matching Based Text Tokenization for a Scarce Resourced Language

Text tokenization is a fundamental pre-processing step for almost all the information processing applications. This task is nontrivial for the scarce resourced languages such as Urdu, as there is inconsistent use of space between words. In this paper a morpheme matching based approach has been proposed for Urdu text tokenization, along with some other algorithms to solve the additional issues of boundary detection of compound words, affixation, reduplication, names and abbreviations. This study resulted into 97.28% precision, 93.71% recall, and 95.46% F1-measure; while tokenizing a corpus of 57000 words by using a morpheme list with 6400 entries.

Two Problems of Circle in Image

Circle is an important primitive in document images, which is contained in many scenes and shapes due to the aesthetic feeling. As it is known to all, three nonlinear points determine a circle, namely, the radius and the coordinates of circle can be calculated. However, three nonlinear points cannot determine a unique desirable circle occasionally in images. Circumferences consist of many pixels, what is the relation of radius size and number of pixels on the circumferences with and without redundancy? We mainly discuss the condition about how to make a unique desirable circle with three nonlinear pixels in images and the above-mentioned relation. Finally, the relation and condition have been found and testified by experiments.

Improving the efficiency of FP tree construction using transactional patternbase

Mining frequent patterns in transaction databases has been a popular theme in data mining study. Common activities include finding patterns among the large set of data items in database transactions. The Apriori algorithm is a widely accepted method of generating frequent patterns. The algorithm requires many scans of the database and thus seriously tax resources. Some of the methods currently being used for improving the efficiency of the Apriori algorithm are hash-based itemset counting, transaction reduction, partitioning, sampling, dynamic itemset counting etc. Two main approaches for associations rule mining are: candidate set generation and test, and restricted test only. Both approaches use to scan massive database multiple times. In our study, we propose a transaction patternbase, constructed in first scan of database. Transactions with same pattern are added to the Patternbase as their frequency is increased. Thus subsequent scanning requires only scanning this compact dataset which increases efficiency of the respective methods. We have implemented this technique with FP Growth method. This technique outperforms the database approach in many situations and performs exceptionally well when the repetition of transaction patterns is higher. It can be used with any associations rule mining method.

CLUSMA: a mobile agent based clustering middleware for wireless sensor networks

This paper proposes CLUSMA: a middleware framework for communication between sensor nodes in Wireless sensor network. Our proposed technique is based on clustering technique (CLUS) and it uses Mobile Agent (MA). This paper gives a sketch of the middleware and the actual implementation of the framework are in progress. The main objective of this middleware is to utilize the networks functions and resources in an effective way. Because the proposed framework is based on MA it will work in homogenous as well as heterogeneous environment. Previous such frameworks used symmetric secret key for sensor nodes authentication. The proposed framework will provide security in heterogeneous environment by using mobile agents in a cluster which will communicate with agents in other cluster.

Error-Correction in Electronic Architectural Drawings

Recently, three-dimension reconstruction has attracted much attention. The reconstruction from electronic architectural drawing is an important as well as urgent task according to the current situation. Errors in drawings will result in inaccurate three-dimension model, part of the methods based on scanned drawings which have been used before will be incomplete and have some drawbacks, therefore this paper proposes new strategies to deal with different categories of errors in the new system based on electronic architectural drawings.

Mobile agent watermarking (MAW): security properties and attacks revisited

Providing integrity of execution is considered as the most challenging problem in Mobile Agent System (MAs). Mobile agent watermarking (MAW) is a technique that attempts to provide integrity of execution to a mobile agent. We discuss the attacks against MAW technique and how the security properties for MAs data integrity are fulfilled. We try to model the behavior of a malicious host by launching a series of passive attacks against mobile agent and then see to what extent the security properties for mobile agent can be achieved. The experimental results suggest that MAW framework can be used to protect the computations results of mobile agents

Performance analysis of prioritized and non prioritized handoff call queuing schemes in mobile cellular networks

Handoff (HO) call queuing can be successfully employed to ensure better Quality of Service (QoS) guarantees in mobile cellular networks. In this paper, we present the performance analysis of prioritized and non prioritized HO call queuing schemes in terms of blocking and dropping probabilities, channel utilization and average waiting time for HO calls in queue with heterogeneous offered traffic. Our results signify the superiority of proposed Handoff Urgency Based (HUB) queuing scheme over First-In-First-Out (FIFO) queuing as it prioritizes HO calls according to user mobility (i.e. fast or slow) and call type (i.e. voice or data calls). It is evident that HUB achieves significantly lower dropping probability and queue waiting times with better channel utilization even under high traffic conditions.