The best legal additional source for ordinary way of finding data is browsing internet. Internet has become the most flexible and easily reachable source of getting information within a single click. Web mining is a process of fetching most appropriate, trustworthy information after analyzing the structure of the data has included and the usage of that data. Numerous methodologies for improving web search mechanism have been established in last decades. But the most of the prevailing approaches are restricted, mainly because of the lesser amount of user preferences for manipulating web search. With the development of the technology, Artificial intelligence become successful approach for web search. With the use of AI, web search model could be included legal reasoning, augmentation ability for retrieving information from the web. The paper will show that the different methods for retrieve and rank information according to the user query in web search with the aid of Artificial Intelligence and critically evaluate each approach.
AI is a huge field which has been rapidly used in today’s technology and by today AI has become an important and necessary topic for each and every sector. When it comes to the internet, internet is the main source which people are trying to obtain most of the solutions for their problems and the best place for the knowledge seekers. Therefore people keep trust on what they have received from typing their issue or matter at the search bar. At the previous stage of the internet era, search results were intrusive. Web browsers such as yahoo, AOL have done their best to interpret what they believed users wanted. But new companies like Google and Bing started using Artificial Intelligence (AI) to intensify the searching algorithms to find better and accurate search results. By today Google company has provided Google voice search option to make more feasible in browsing web search engines.
WEB SEARCH MECHANISM
Not like data mining, web mining is considerable because web is consist of unorganized data and their changes are rapid. Since web mining process consists 3 steps such as web structure mining, web usage mining and web content mining. Further Web structure mining is about discovering information using the hyperlinks in web. Web content mining grab the essential data from the web pages which can be categorized based on the topics, patterns concerning after users’ opinion can also be identified. Opinion mining is considered both data mining techniques and natural language processing techniques. Patterns in behaviors and interactions are recognized in web usage mining. It identifies the patterns and flows in user preferences against resources among web pages. Even though there were new searching strategies with the development of the technology in last decades, still there are some limitations in searching without having proper criteria for improving user preferences and ranking searching results. This report has been included several solutions which implemented to deploy filtering and ranking process in an effective manner. Multi agent technology and Swarm Intelligence are two methods that uses advanced AI methods for ranking retrieved queries and sort appropriate information according to the user requirements. In 1989 Swarm intelligence (SI) was introduced, by analyzing the real scenario that the group behavior of a swarm helps to react intelligently to an emergency situation. But both multi agent technology and swarm intelligence methods were analyzed in qualitative manner, not comprehensively discussed about the implementations and algorithms.This article has been noted a recommender system for classifying obtained results according to user preference order. Traditional web search engines and semantic web search engines based on user keyword with documents, terms, texts, entities, which correlate with user queries. Image based searching strategy with web ranking mechanism has formulated. This integrated approach uses both semantic information of user query and image appearing within web pages to retrieve more relevant information to the user. Thereafter Web scraping is initially used to convert the information collected from unstructured sources to structured data. Other the HTML pages, information can be found as pdf files, PowerPoint presentations, excel files, images, videos and audios. Due to the inconvenience in recognizing the proper terms to be used in searches and to verify the relevant results in human approach, metasearch engine was built. Users’ preferences can be determined by metasearch engine. A metasearch engine is a tool which is able to implement the same query on different search engines. Moreover, it organizes and ranks the collected information before presenting the results to the users. The article has observed Google, Bing and DuckDuckGo search engines. Even though this approach does not pay attention to the images as an important source of information for fetching web documents, image contains vital details for a web search query. The issue occurs while extracting query from images. It is a challenging and computationally expensive task. To identify the way images matches with geometrical transformation with the enhancement of image recognition, computer vision and object recognition. Another issue in web search mechanism is all the trivial keyword based search engines display the result without considering the meaning of the keywords. Therefore, the result set may vary and might also lack relevance with respect to a user. Ontology Driven Web search celebration system attempts to remove this drawback with the help of ontology. Proposed details about semantic aspects of traditional search system. The system utilizes an extended version of inverted index and provides a search technique based on same. By such an update, the system becomes capable of answering the user’s query by considering semantic side of a query. This is another technique used to obtain repositories using the earlier experiences of user preferences. CBR has an Intelligent Text analysis for retrieving information from unorganized data from web pages. They have followed Dynamic System Development Model (DSDM) with iterative approach with parallel to incremental approach based on incessant user participation. Requirement Change Management (RCM) phase in which used for handling the modifications based on the preferences of the users. Case based reasoning technique is vitally used to manipulate the requests. And also they have used text mining technique with the purpose of obtaining the cases from the repository. With the usage of CBR it was able to successfully analyze user preferences when it comes to the further browsing. With the aid of natural language processing methods to identify information based on human language processing ability with the knowledge, speech process Word for word approach, semantic transfer approach ,syntactic transfer approach, and Interlingua approach are four approaches considering machine translations by Manning and Schutze.. Word-for-Word Approach is the simplest, but most inapplicable, with two major problems, first one is being the ambiguity problem. The second problem is the order of words which differs from one language to another, with the meaning of the order which did not confirm the linguistics rules.. The Syntactic Approach analyzed the ranking issue, but does not resolve the first problem in the Word-for-Word-Approach. The Semantic Approach consider the semantic meaning more comprehensive, but it still faces literal meaning problem. The Interlingua Approach is not dependent on the language presents the meaning. To overcome these problems the proposed based on Corpus-based Approach. Thesaurus-based Approach. Thesaurus-based Approach provides more details on searching process. This is a powerful tool which visualized hints for a multilingual dictionary. Proposed system uses an extension of thesaurus- based approach which with two languages (English and Spanish). Query translation, Document translation, and Mixture of Query and Document translation are the phrases used. Whenever a user enters a query in the semantic search interface, based on the importance and the relevance fetched the links and ranked them based on the scoring algorithm and spontaneously show two semantic results in both languages. Secondly this approach is used Vector Space Model to recognize and optimize the query term with the indexed documents. For example when user inputs a query either in English or Spanish, search engine translates the query to the intermediate language using dot product between the documents in the repository and the translated query. After retrieval relevant documents will be ranked based on the score. Using Smart Voice search Engine, you just need to talk and it will process other functions at all. Smart voice search engine consists of three main sections such as speaker recognition, automatic speech recognition and the browsing engine. When it comes to the implementation section they have used MFCC features with GMM and HMM. Eventually they have conducted an evaluation for the system. Among all variables system was able to obtain more than 50% accuracy. And also there was a comparison between SVSE and Google voice engines. Even though people have used search engines by typing what they want to search or speak using the microphone button. But this scenario is not done in a useful manner. The new system was optimized for lesser running time for retrieving information. Therefore the new approach can be considered as a worth solution for web based search applications. The search engine researchers have focused on providing more accurate and fast search engines. Therefore In 2008 Google Mobile App (GMA) was released for IPhones. In 2009 Google Maps introduced new methodology with taking inputs with speech or text and output maps with adding voice commands, and street view support as the input and output maps with adding voice commands and traffic reports. Click through data structure consider logs of the interactions between users and search engines. Based on the past preferences that user has gone through and that candidates are used for alternate items to retrieve new query information. System decides whether the preference feedback helps to improve search results or not. The executed query q and the ranking results r are recorded when ranked document is displayed to the user. IP address, the time, experimental condition, a session identifier browser and a query identifier are the main elements that obtained for recording purposes. Click through data was captured using middle server which records the user clicks information without having overhead of the user in search, which reduces the processing time. This is a better way for web search engines for enhance their performance and attract user attention to their approaches.
Commenti