International Journal of Advance Foundation and Research in Science & Engineering (IJAFRSE) Volume 1, Issue 2, July 2014.

Transcription

1 A B S T R A C T International Journal of Advance Foundation and Research in Science & Engineering (IJAFRSE) Information Retrieval Models and Searching Methodologies: Survey Balwinder Saini*,Vikram Singh,Satish Kumar. National Insititute of Technology, Kurukshetra, India. me7saini@gmail.com *, viks@nitkkr.ac.in, er.satishkumar149@gmail.com Information retrieval is a paramount research area in the field of computer science and engineering. Information retrieval (IR) is mainly concerned with the probing and retrieving of cognizance-predicated information from database. In this paper, we represent the different models and techniques for information retrieval and we are additionally describing sundry indexing methods for decrementing search space and different (searching) probing techniques for retrieving information. Index Terms: Information Retrieval (IR), Indexing, IR model, Searching, Vector-Space model, Probabilistic IR model. I. INTRODUCTION Information retrieval is mainly considered as a component of computer science that deals with the representation, storage, and access of information [1]. Information retrieval is pertained with the organization and retrieval of information from immensely colossal database Sources [2]. Information Retrieval (IR) is the technique by which a sizably voluminous accumulation of data is represented, stored, and fetched for the purport of cognizance revelation as a result to a utilizer's request or query [3]. This process demand sundry stages initiate with representing data and exiting with returning germane information to the perspective utilize. Middle stage includes searching, filtering, matching and ranking operations. The main aim of information retrieval model (IR) is to finding relevant knowledge-base information or a document that fulfill user needs. To achieve this aim, IR usually employ following processes: In indexing process, the documents are represented in restate content form. In filtering process, all the stop words and common words are filtered out. Searching is the main process of IR. There are different techniques for retrieving information that match with users need. There are two elementary measures for assessing the pronominal of information retrieval [2]. Precision (positive predictive value): Division of retrieved documents that is relevant to the users need. Recall (sensitivity): Division of the documents that are relevant to the user query that are successfully retrieved , IJAFRSE All Rights Reserved

2 There are three primary processes an IR model has to support: the representation of the information of the documents, the interpretation of the user's information need, and the comparison of these two representations. The general IR processes are shown in Figure 1. In the figure 1, square shape box represent data and round shape box represent processes. Representing the documents in a summarized way is usually called the indexing process. Indexing process implemented off-line, means; client of the information retrieval system is not directly involved in this process. Indexing process result in a representation of the document [3] [4]. Users do not search irrelevant information; they have a need for only relevant information. The process of representing relevant information need to the given user is called as the query formulation process. The resulting representation is called query [4]. Documents Information Indexer Interpretation Indexed User Query Comparison/Matching Retrieved Documents Figure. 1 A general Process of IR System Comparing the two different representations is called as the matching process and retrieval of relevant documents is the result of this process. The organization of this survey paper is as follows. A brief overview of IR models is presented in Section II, followed by indexing or ranking method in section III, come after by searching methods in Section IV, Followed by IR usage area in section V, Finally, Section VI address conclusions. II. IR MODELS An IR model describes the elaboration of the document representation, the user query representation and the retrieval mechanism or process [4]. The basic IR models can be divided into three models like: Boolean model, vector-space model, and probabilistic model [1], [2], and [3]. This section briefly describes these models. A. Boolean model: The Boolean model uses set theory, that is, Boolean algebra and its three components AND, OR and NOT, for query formulation, but it has one major drawback: a Boolean system is fail to rank the result list of retrieved documents [4]. In the Boolean model, all documents are associated with a set of distinct words or key-words and User Queries are also represented by expressions of keywords separated by AND, OR, or NOT. The retrieval function of Boolean model takes a document as either relevant or irrelevant [4]. B. Vector-Space model: The vector-space model is the best model because its attempt to rank documents by some similarity value between the user query and each document [1], [2], [3], and [4]. In the Vector Space Model (VSM), , IJAFRSE All Rights Reserved

3 documents and user query are both represented as a Vector and the angle between the two vectors are calculated using special function, that is, cosine function. Cosine function defines the similarity values between two given vectors and it can also be defined as: Where,,. Documents and queries are represented as vectors. $!,"!,#!%& $ * ' /)!,"!%& $ * )!,#!%& " &,", *,",..,," &,#, *,#,..,,# Vector Space Model have been introduce term-weight scheme known as tf-idf weighting. These weights have a term frequency (tf) factor measuring the frequency of occurrence of the terms in the document or query texts and an inverse document frequency (idf) factor measuring the inverse of the number of documents that contain a query or document term [10 ]. C. Probabilistic model: The most crucial or essential function of the probabilistic model is its initiate to rank documents by their probability of relevance given a user's query [5]. Both documents and user queries are represented by vectors ~d and ~q these are binary vectors, each binary vector component show whether a given document attribute/component or term occurs in the document or query, or not. Instead of probabilities, for the probabilistic model the index term weight variables are all binary that is,,! - {0, 1}, #! {0, 1} [2, 3]. A user query q is a subset of index terms. Let R be the set of documents known to be relevant. Let R be the complement of R (that is, the set of non-relevant documents). Let./be the probability that the document d is relevant to the query q and. R/d be the probability that d is non-relevant to q. The similarity (sim) sim (d, q) of the document d to the query q is defined by the ratio:,. /.0 R d 1 Using Bayes rule, this equation defined above can be expressed as follows:,.2 3./. R'P5 R6 P (d/r) stands for the probability of randomly selecting the document d from the set R of relevant documents. In addition P(R) is a probability of document that is randomly selected from the complete set of documents is relevant. The meanings attached to P (d/ ( R )) and P ( R) are analogous and complementary. P(R) and P ( R) are the same for all the documents in the collection in other words they are constants. Therefore,,~. /. R' , IJAFRSE All Rights Reserved

4 P (d/r) stands for the probability of randomly selecting the document d from the set R of relevant documents. In addition P(R) is a probability of document that is randomly selected from the complete set of documents is relevant. The meanings attached to P (d/ ( R )) and P ( R) are analogous and complementary. P(R) and P ( R) are the same for all the documents in the collection in other words they are constants. Therefore,,~. - 8 /. - R '. III. INDEXED TECHNIQUES There are various common information retrieval indexing techniques [6, 7], including signature files and inverted index. A. Signature file: In signature file indexing technique each document return a bit of string, (that is, signature) using hashing method on its text and superimposed coding. The final output of document signatures are stored in a special way, that is sequentially in a separate file and this file is called as signature file. The signature file is much smaller than the original file, and it can provide high search rate. B. Inverted index: Each document can be represented by a list of some reference words called keywords which depict the contents of the document for retrieval purpose. Fast retrieval can be obtained if we invert on those keywords. All the reference words are stored alphabetically in a file called index file. For each keyword we keep a list of pointers to the characterize documents in the postings file. This method is mostly used by all the commercial systems. IV. SEARCHING TECHNIQUES There are different searching techniques, including linear search, brute force search and binary search and these searching techniques are describe as follows: A. Linear search technique: It is a basic technique of finding a particular word or keyword from a list of words or array that checks presence of every element in list, one at a time and in a sequence. This search technique is a simplest search technique. Disadvantage of linear search is its searching speed is very poor or slow especially in case of ordered list. This type of search is also called as sequential search. B. Brute force search technique: It is a very common problem-solving technique that consists of consistently itemize all possible participants for the solution and determine whether each participant gratify the problem s statement. This searching technique is simple to apply and it will always return a solution if it exist , IJAFRSE All Rights Reserved

5 C. Binary search technique: It finds the position of a particular input value that is, (the search key) within an array sorted by some key value. For binary search technique, the given array should be arranged in some order that is, ascending or descending. In each step, this method examines the search key value as respect with the middle element key value of the given sorted array. If the value of both key's matched, then a matching item has been found and it should be indexed. Differently, if the search key value is less/greater than the middle element's key value, then the method repeats its steps on the sub-array to the left/right of the middle element. If the leftover array to be searched and it is found empty, then the search key cannot be found in this empty array and a particular bit of string is returned that is, Not Found. V. USAGE AREA OF IR SYSTEMS Information retrieval (IR) systems were initially developed to improve and manage the large amount of data or information. Many private or government universities, corporate sector, and public Libraries nowadays use IR systems to provide access to the different information like books, journals, and other documents. Now information retrieval is frequently used in so many applications [8], [9], [10] and some common applications of information retrieval system are defined as follows: A. Web Search Engine: One of the most practical applications of information retrieval system is a search engine and it is meant for retrieving relevant information from a large or big size text collections. They are best-known examples of IR system, but various searches exist, like: Desktop search, Enterprise search, Unify search, Mobile search, and Social search [7, 8]. B. Multimedia search: This type of Search can be applied by multi-modal search interfaces means this include some other type of media also for getting information which is different from textual search for example an image retrieval system is a multimedia search system in a computer for browsing, searching and retrieving images from a huge collections of digital images [6]. C. Digital Library: A digital library is a type of library in which collections are stored in digital formats and these formats are accessible by computer systems. The digital information may be stored locally, or accessed remotely via computer network systems. So finally these types of libraries are considered into the area of information retrieval system [9]. D. Information Filtering: Information filtering system consists of many tools that help user to retrieve the most relevant information and valuable information, so you can dedicate less time to read, listen and view. It is rightly directs the most relevant, interesting and valuable documents, apart from the most irrelevant information. Information filters are also used to manage and structure information in a right and intelligible way, in favor to cluster messages on the mail addressed. Information filters are very valuable , IJAFRSE All Rights Reserved

6 in the retrieved information obtained of the search engines on the Internet. The usage of information filtering improves day by day to get downloading Web documents and more efficient messages [8, 10]. VI. CONCLUSION At the end of this survey paper, we conclude that, information retrieval (IR) is a process of searching and retrieving the knowledge based information from a large collection of documents. This survey also describes the basics of the information retrieval system. In very first section, we are specifying the information retrieval system with their common attributes. After this section, we concerns with traditional IR models and also discuss about their process of indexing or ranking techniques and searching techniques. This survey paper also includes two areas that are, the area of information retrieval literature and the area of information retrieval applications. VII. REFERENCES [1] G. Salton and M.J. McGill, Introduction to Modern Information Retrieval, McGraw-Hill Book Co., New York, [2] R. Baeza-Yates and B. Ribeiro Neto, Modern Information Retrieval, Harlow: ACM Press [3] H. Dong, F.K. Husain and E. Chang, A Survey in Traditional Information Retrieval Models, IEEE International Conference on Digital Ecosystems and technologies, pp ,2008. [4] S. Raman, V. Kumar, and S. Venkatesan, Performance Comparison of Various Information Retrieval models Used in Search Engines, IEEE conference on communication, information and Computing Technology, Mumbai, India, [5] F.Silva, R. Girardi, and L. Drumond, An IR Model for the Web, IEEE International Conference on information technology, Brazil, [6] A. Lashkari, F. Mahdavi and V. Ghomi, A Boolean Model in Information Retrieval for Search Engines, IEEE International Conference on Information Management and Engineering, pp , [7] J. Hua, Study on the Performance of Information Retrieval Models, in 2009 International Symposium on Intelligent Ubiquitous Computing and Education, pp , [8] D. Wolfram, Search characteristics in different types of Web - based IR environments: Are they the same?, in Elsevier Journal of Information Processing and Management, pp , [9] J. Han, M. Kamber, and J. Pai, Data Mining: Concepts and Techniques, Morgan Kaufmann Publishers, USA, Third Edition, Elsevier, [10] M. Karthikeyan and P. Aruna, Probability based document clustering and image clustering using content-based image retrieval, in ELSEVIER journal of Applied Soft Computing, pp , , IJAFRSE All Rights Reserved