Shashi Shekhar

Semantic Based Image Retrieval using multi-agent model by searching and filtering replicated web images

The volume of web has increased tremendously, as a result retrieving useful and relevant information in terms of textual information or visual information has become a difficult task. The current search engines still depend on textual descriptions for retrieving images from the web. It is difficult to search for the relevant images using commercial search engines. The results from commercial image search engines are often mixed up with irrelevant and redundant images extracted from the web. This paper on SBIR(Semantic Based Image Retrieval) proposes a method to find user intended images from the web using proposed image crawling mechanism. The proposed framework overcomes the two major problems in case of retrieving user centric images from the web: freshness problem and redundancy problem. The proposed framework can also be used as personalized image search engine which effectively extract the text information on the web to semantically describe the retrieved images. Experiments are designed and conducted to test the performance of proposed image crawling approach. The experimental results illustrate substantial improvement in the image crawling strategy, especially when the search strings is combined with low level image features.

An Architectural Framework of a Crawler for Retrieving Highly Relevant Web Documents by Filtering Replicated Web Collections

As the Web continues to grow, it has become a difficult task to search for the relevant information using traditional search engines. There are many index based web search engines to search information in various domains on the Web. By using such search engines the retrieved documents (URLs) related to the searched topic are of poor quality also as the amount of Web pages is growing at a rapid speed, the issue of devising a personalized Web search is of great importance. This paper proposes a method to reduce the time spend on browsing search results by providing a personalized Web Search Agent (MetaCrawler). In the proposed technique of personalized Web searching, Web pages relevant to user interests will be ranked in the front of the result list, thus facilitating the user to get a quick to get access those links ranked in the front of the list. An experiment was designed and conducted to test the performance of proposed Web-Filtering approach. The experimental results suggest substantial improvement in the crawling strategy, especially when the search strings are small.

Opinion Mining Classification Based on Extension of Opinion Mining Phrases

Opinion mining is the field of study that analyses people’s thoughts, sentiments, emotions and attitude towards entities, product, services, issues, topics, events and their attributes. There are many different tasks such as opinion extraction, sentiment mining, emotional analysis, review mining etc. The important aspect of opinion minion is to gather the information from reviews, blogs, etc. and then finding out the behavior of that information, i.e. the information is related to either positive or negative context. The positive and negative reviews or blogs deal with a numerical value. The value is to be calculated using SentiWordNet 3.0. The opinion words are mainly adjective words such as “good,” “better,” “awesome.” But there arises several problems because identifiers negation words and the extension of the opinion words such as “very very good” are not considered. In this paper, details about opinion mining, how the polarity value deals with positive and negative and how to deal with Roman language reviews and blogs is discussed.

A WEBIR Crawling Framework for Retrieving Highly Relevant Web Documents: Evaluation Based on Rank Aggregation and Result Merging Algorithms

Finding relevant information on the web is an ongoing problem. Commercial search engines like Google rely on sophisticated algorithms to index huge collection of web pages to make them accessible to user queries. Users, however, are still frequently overloaded with irrelevant results. The required information is available in replicated manner scattered in various disjoint databases. For effective web information retrieval, user need to consult several commercial search engines working on different architecture and principles. Rank aggregation and Result merging is the key component of a crawling mechanism used by the commercial search engines. Once the results from various search engines are collected, they need to be merged into a single unified ranked list. The effectiveness of any crawling mechanism is closely related to the rank aggregation and result merging algorithm it employs. In this paper, we investigate a variety of rank aggregation and result merging algorithms based on a wide range of available information. The effectiveness of these algorithms is then compared experimentally to our proposed crawling framework based on queries from the TREC Web track and 3 most popular general-purpose search engines. Our experiments yield two important results. First, simple result merging strategies can outperform Google, Yahoo and MSN Live. Second, Proposed Content Based Result Aggregation (CBRA) algorithm outperforms other existing content based merging algorithms based on full document content.

An object based image retrieval framework based on automatic image annotation

Automatic image annotation is the process of assigning relevant keywords to the images. It is considered to be potential research area in current scenario. Annotation to an image can be defined as the information which could describe an image by considering three ways i.e. when these images were taken, what are the different objects available in these images and finally the images belongs to whom. For solving the problem of automatic image annotation, many algorithms have been proposed. Efforts are going on to develop more efficient algorithms. In this paper we have proposed an object based image retrieval algorithm for automatic image annotation. The proposed algorithm considers selection of objects with in an image. This object selection helps in dividing the image into different set of groups on the basis of present objects in an image. Thus, we do not need to extract the whole features from the images when a new image comes, rather we extract features from the objects and matches those features against the different groups of images for the feature matching and effective retrieval based on object selection.

An on demand routing protocol AODV with end to end reliability and backward route information

AODV protocol is a comparatively more popular on demand routing protocol in mobile ad hoc networks. However, reliability in terms of data communication is not much satisfactory in traditional AODV. This paper presents an AODV with end to end reliability (AODV-EER). The main idea of proposed modification in AODV is to find the route with lowest drop rate from source to destination. We also propose a backward route entry mechanism in order to initiate repair action after primary route breaks. We analyze the performance of proposed protocol on NS2 in terms of packet delivery ratio, normalized routing load, and end to end delay. The experiments clearly indicate that proposed protocol performs better than the traditional protocol AODV.

A web link retrieval framework for retrieving non redundant description based web links

This paper explores the measuring similarity between the web objects which are one of the fundamental task in information retrieval domain. This paper proposes a framework for improved and efficient web object search based on search domain. The concept of proposed approach defines the similarity between two objects (object can be a link or text content) and retrieve the related links with their content descriptor according to users interest. The proposed framework can retrieve the similar web objects in effective and faster way by eliminating the redundant objects. The proposed algorithm is experimentally tested and the results shows improvement in fast and effective retrieval of web links. The proposed algorithm automatically extracts web links based on user interests. Furthermore, in accordance with user interests web objects mined and feedbacks of users are assessed for effective retrieval with an idea of dynamically adjusting the ranking scores of Web Pages based on similarity measures.