Search

Interest in the evaluation techniques for Information Retrieval Systems has significantly increased with the commercial use of information retrieval technologies in the everyday life of the millions of users of the Internet. Until 1993 the evaluations were done primarily by academicians using a few small, well known corpora of test documents or even smaller test databases created within academia. The evaluations focused primarily on the effectiveness of search algorithms. The creation of the annual Text Retrieval Evaluation Conference (TREC) sponsored by the Defense Advanced Research Projects Agency (DARPA) and the National Institute of Standards and Technology (NIST) changed the standard process of evaluating information systems. Conferences have been held every year, starting from 1992, usually in the Fall months. The conference provides a standard database consisting of gigabytes of test data, search statements and the expected results from the searches to academic researchers and commercial companies for testing of their systems. This has placed a standard baseline into comparisons of algorithms. Although there is now a standard database, there is still debate on the accuracy and utility of the results from use of the test corpus. Section 10.2 introduces the measures that are available for evaluating information systems. The techniques are compared stressing their utility from an academic as well as a commercial perspective. Section 10.3 gives examples of results from major comparisons of information systems and algorithms. [ABSTRACT FROM AUTHOR]

Three classical techniques for text retrieval techniques have been defined for organizing items in a textual database, for rapidly identifying the relevant items and for eliminating items that do not satisfy the search. The techniques are full text scanning (streaming), word inversion and multi-attribute retrieval (Faloutsos-85, Salton-83). In addition to using the indexes as a mechanism for searching text in information systems, streaming of text was frequently found in the systems as an additional search mechanism. Streaming of text is a sequential search of the text. This technique is used to complete a query by searching for query terms that could not be satisfied by the index (e.g., imbedded search terms). It is also frequently used to locate the search terms for highlighting in the retrieved item prior to display. [ABSTRACT FROM AUTHOR]

The primary focus on Information Retrieval Systems has been in the areas of indexing, searching and clustering versus information display. This has been due to the inability of technology to provide the technical platforms needed for sophisticated display, academicśs focusing on the more interesting algorithmic based search aspects of information retrieval, and the multi-disciplinary nature of the human-computer interface (HCI). The core technologies needed to address sophisticated information visualization have matured, supporting productive research and implementation into commercial products. The commercial demand for these technologies is growing with availability of the "information highway." System designers need to treat the display of data as visual computing instead of treating the monitor as a replica of paper. Functions that are available with electronic display and visualization of data that were not previously provided are (Brown-96): modify representations of data and information or the display condition (e.g., changing color scales)use the same representation while showing changes in data (e.g., moving between clusters of items showing new linkages)animate the display to show changes in space and timeenable interactive input from the user to allow dynamic movement between information spaces and allow the user to modify data presentation to optimize personal preferences for understanding the data.Create hyperlinks under user control to establish relationships between data [ABSTRACT FROM AUTHOR]

Chapter 5 introduced indexing associated with representation of the semantics of an item. In all of the techniques discussed in Chapter 5, our information database can be viewed as being composed of a number of independent items indexed by a series of index terms. This model lends itself to two types of clustering: clustering index terms to create a statistical thesaurus and clustering items to create document clusters. In the first case clustering is used to increase recall by expanding searches with related terms. In document clustering the search can retrieve items similar to an item of interest, even if the query would not have retrieved the item. The clustering process is not precise and care must be taken on use of clustering techniques to minimize the negative impact misuse can have. These issues are discussed in Section 6.1 along with some general guidelines of clustering. [ABSTRACT FROM AUTHOR]

Chapter 3 introduced the concept and objectives of indexing along with its history. This chapter focuses on the process and algorithms to perform indexing. The indexing process is a transformation of an item that extracts the semantics of the topics discussed in the item. The extracted information is used to create the processing tokens and the searchable data structure. The semantics of the item not only refers to the subjects discussed in the item but also in weighted systems, the depth to which the subject is discussed. The index can be based on the full text of the item, automatic or manual generation of a subset of terms/phrases to represent the item, natural language representation of the item or abstraction to concepts in the item. The results of this process are stored in one of the data structures (typically inverted data structure) described in Chapter 4. Distinctions, where appropriate, are made between what is logically kept in an index versus what is physically stored. [ABSTRACT FROM AUTHOR]

Knowledge of data structures used in Information Retrieval Systems provides insights into the capabilities available to the systems that implement them. Each data structure has a set of associated capabilities that provide an insight into the objectives of the implementers by its selection. From an Information Retrieval System perspective, the two aspects of a data structure that are important are its ability to represent concepts and their relationships and how well it supports location of those concepts. This chapter discusses the major logical data structures that are used in information retrieval systems. The implementation of a data structure (e.g., as an object, linked list, array, hashed file) is discussed only as an example. [ABSTRACT FROM AUTHOR]

Previous chapters defined the concept of indexing and the data structures most commonly associated with Information Retrieval Systems. Chapter 5 described different weighting algorithms associated with processing tokens. Applying these algorithms creates a data structure that can be used in search. Chapter 6 describes how clustering can be used to enhance retrieval and reduce the overhead of search. Chapter 7 focuses on how search is performed. To understand the search process, it is first necessary to look at the different binding levels of the search statement entered by the user to the database being searched. The selection and ranking of items is accomplished via similarity measures that calculate the similarity between the userśs search statement and the weighted stored representation of the semantics in an item. Relevance feedback can help a user enhance search by making use of results from previous searches. This technique uses information from items judged as relevant and non-relevant to determine an expanded search statement. Chapter 6 introduces the concept of representing multiple items via an single averaged representation called a "centroid." Searching centroids can reduce search computation, but there is an associated risk of missing relevant items because of the averaging nature of a centroid.Hyperlinked items introduce new concepts in search originating from the dynamic nature of the linkages between items cr] [ABSTRACT FROM AUTHOR]

The first two chapters of this book presented the architecture of a total Information Storage and Retrieval System and the basic functions that apply to it. One of the most critical aspects of an information system that determines its effectiveness is how it represents concepts in items. The transformation from the received item to the searchable data structure is called Indexing. This process can be manual or automatic, creating the basis for direct search of items in the Document Database or indirect search via Index Files. Rather than trying to create a searchable data structure that directly maps to the text in the input items, some systems transform the item into a completely different representation that is concept based and use this as the searchable data structure. The concept weighting schemes have demonstrated the capability to find items that the traditional weighted and non-weighted data structures have missed. Systems that use a specialized hardware text search processor do not require the searchable data structure, but search the original standardized documents (see Chapter 9). [ABSTRACT FROM AUTHOR]

This chapter discusses the major functions that are available in an Information Retrieval System. Search and browse capabilities are crucial to assist the user in locating relevant items. The search capabilities address both Boolean and Natural Language queries. The algorithms used for searching are called Boolean, natural language processing and probabilistic. Probabilistic algorithms use frequency of occurrence of processing tokens (words) in determining similarities between queries and items and also in predictors on the potential relevance of the found item to the searcher. Chapter 4 discusses in detail the data structures used to support the algorithms, and Chapters 5 and 7 describe the algorithms. The majority of existing commercial systems are based upon Boolean query and search capabilities. The newer systems such as TOPIC, RetrievalWare, and INQUERY all allow for natural language queries. [ABSTRACT FROM AUTHOR]

This chapter defines an Information Storage and Retrieval System (called an Information Retrieval System for brevity) and differentiates between information retrieval and database management systems. Tied closely to the definition of an Information Retrieval System are the system objectives. It is satisfaction of the objectives that drives those areas that receive the most attention in development. For example, academia pursues all aspects of information systems, investigating new theories, algorithms and heuristics to advance the knowledge base. Academia does not worry about response time, required resources to implement a system to support thousands of users nor operations and maintenance costs associated with system delivery. On the other hand, commercial institutions are not always concerned with the optimum theoretical approach, but the approach that minimizes development costs and increases the salability of their product. This text considers both view points and technology states. Throughout this text, information retrieval is viewed from both the theoretical and practical viewpoint. [ABSTRACT FROM AUTHOR]