M. Ortega, Y. Rui, K. Chakrabarti, K. Porkaew, S. Mehrotra, T.S. Huang, Supporting ranked boolean similarity queries in MARS, IEEE Trans. on Knowledge and Data Engineering 10(1998) 905-925.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....by performing range queries relative to a selected example or multiple examples. However, since different people may have different perceptions on the same set of items, using one distance function as a universal solution cannot satisfy all users. To address this issue, some systems, such as MARS [13] and Garlic [14,15] have adapted query expansion models, which can simulate a customized distance function for each user by weighting the relative importance of different features. Items in these systems are represented as different vectors extracting from the physical features, and each vector ....

....similarity values from matrix QBIC color function during the query processes. Because of the subjectivity of human perceptions, however, using one single distance function as a universal solution cannot satisfy all users. To solve this problem, other contentbased retrieval systems, such as MARS [13,11] and Garlic [14, 15] employ different distance functions, where each function corresponds to a physical feature. These systems simulate the ideal distance function through a weighted combination of the distance functions based on subjective weightings for each user. Hence, the query results ....

Ortega M, RuiY, Chakrabarti K, Mehrotra S (1998) Supporting ranked boolean similarity queries in MARS. IEEE Trans Knowl Data Eng 10:905--025

....our work, do not incorporate IR style keyword relevance. The problem of processing top k queries has attracted recent attention in a number of different scenarios. The design of the pipelined algorithms that we propose in this paper faces challenges that are related to other top k work (e.g. [14, 6, 10, 4]) However, our problem is unique (Section 5) in that we need to join (ranked) tuples coming from multiple relations in unpredictable ways to produce the final top k results. Finally, Natsev et al. 13] extend the work by Fagin et al. 6] by allowing different objects to appear in the source ....

M. Ortega, Y. Rui, K. Chakrabarti, K. Porkaew, S. Mehrotra, and T. Huang. Supporting ranked Boolean similarity queries in MARS. TKDE, 10(6):905--925, 1998.

....multimedia sources compared to traditional join techniques. However, the paper also points to intrinsic difficulties arising from heterogeneity of sources that makes establishing object identity difficult and describes the steps that were needed in Garlic to overcome these issues. Ortega et al. [29] support ranked retrieval over image databases as part of their MARS system. One of their key contributions is an adaptation of FA that has the flavor of a merge join algorithm. Top c query processing over traditional relational data has also received attention [6, 7, 10, 12] Recently, Fagin ....

M. Ortega, Y. Rui, K. Chakrabarti, K. Porkaew, S. Mehrotra, and T. S. Huang. Supporting ranked boolean similarity queries in MARS. IEEE Transactions on Knowledge and Data Engineering (TKDE), 10(6):905- 925, 1998.

....analyze the linear weighted method for feature merging and finally, in section 6 we explain how querying will be implemented in VizIR. 2. RELATED WORK Past efforts in CBIR have lead to several general purpose prototypes like QBIC ( 3] Virage ( 1] VisualSEEk ( 9] Photobook ( 5] and MARS ([4]) Next to the implemented feature classes and user interfaces these prototypes differ in their similarity measurement. Usually, CBIR similarity measurement follows the Vector Space Retrieval model and is done by measuring the distances of feature vectors with distance functions that are based on ....

.... first, the fact that not all features show a linear relationship and linear merging therefore is not a suitable method to combine such features and second, that in most systems weights have to be provided by the user who is normally overtaxed by this task ( 7] For these reasons, the authors of [4] propose the employment of the Boolean Model instead of Linear Weighted Merging. According to this model two stimuli A and B are similar for a certain feature F, if they fulfill the following condition v 1 : # # # # s s q , # is called a degree of tolerance. It is a threshold for the ....

M. Ortega, R. Yong, K. Chakrabarti, K. Porkaew, S. Mehrotra, and T.S. Huang, "Supporting Ranked Boolean Similarity Queries in MARS", IEEE Transactions on Knowledge and Data Engineering, vol. 10, no. 6, pp. 905925, November 1998.

....approach to extract spatially localized color information and provided for efficient indexing of the color regions. In their method, the large single color regions are extracted first, followed by multiple color regions. They utilize binary color sets to represent the color content. Ortega et al. [2] have used the HS co ordinates to form a two dimensional histogram. The H and S dimensions are divided into N and M bins, respectively, for a total of NxM bins. Each bin contains the percentage of pixels in the image that have corresponding H and S colors for that bin. They use the intersection ....

Ortega, M., Y.Rui, K.Chakrabarti, K. Porekaew, S.Mehrotra, T.S.Huang, "Supporting ranked Boolean similarity queries in MARS", IEEE Trans. on Knowledge and Data Engineering, Vol. 10, No. 6, pp. 905-925, 1998.

.... contains only the equivalence relation on one or more attributes, the join is called an equi join. If in addition, the equi join is defined on key attributes only, we call the resulting join unique. Unique ordered joins have been considered previously by Fagin et al. 1, 2] Ortega et al. [6], and Guntzer et al. 3] However, the general class of joins based on arbitrary join predicates was previously unsupported efficiently, and is the main focus of this work. 1.3 Proposed Approach and Contributions In this paper, we address the ordered join problem defined in Section 1.2. In ....

....valid combinations of such objects. In contrast, the above algorithms all apply to the scenario of joining multiple sets of the same objects that are ordered differently in each stream. Another treatment of the same problem of equi joins over key attributes was presented by Ortega et. al in [6]. The authors defined a query tree whose nodes represented intermediate matches derived from the matches at the children nodes, and evaluated it bottom up to get the final similarity score at the root. They also proposed algorithm variations for specific score aggregation functions so that the ....

M. Ortega, Y. Rui, K. Chakrabarti, K. Porkaew, S. Mehrotra, and T. S. Huang. Supporting ranked Boolean similarity queries in MARS. IEEE Trans. on Knowledge and Data Engineering, 10, Nov.--Dec. 1998.

....al. 9] presented variations of Fagin s original FA algorithm [6] for processing queries over multimedia databases. In particular, Guntzer et al. 9] reduce the number of random accesses through the introduction of more stop condition tests and by exploiting the data distribution. The MARS system [15] also uses variations of the FA algorithm and views queries as binary trees where the leaves are single attribute queries and the internal nodes correspond to fuzzy query operators. Chaudhuri and Gravano also built on Fagin s original FA algorithm and proposed a cost based approach for ....

M. Ortega, Y. Rui, K. Chakrabarti, K. Porkaew, S. Mehrotra, and T. S. Huang. Supporting ranked boolean similarity queries in MARS. TKDE, 10(6):905--925, 1998.

....# contains only the equivalence relation on one or more attributes, the # join is called an equi join. If in addition, the equi join is defined on key attributes only, we call the resulting join unique. Unique ordered joins have been considered previously by Fagin et al. 1, 2] Ortega et al. [6], and Guntzer et al. 3] However, the general class of joins based on arbitrary join predicates was previously unsupported efficiently, and is the main focus of this work. 1.3 Proposed Approach and Contributions In this paper, we address the ordered join problem defined in Section 1.2. In ....

....valid combinations of such objects. In contrast, the above algorithms all apply to the scenario of joining multiple sets of the same objects that are ordered differently in each stream. Another treatment of the same problem of equi joins over key attributes was presented by Ortega et. al in [6]. The authors defined a query tree whose nodes represented intermediate matches derived from the matches at the children nodes, and evaluated it bottom up to get the final similarity score at the root. They also proposed algorithm variations for specific score aggregation functions so that the ....

M. Ortega, Y. Rui, K. Chakrabarti, K. Porkaew, S. Mehrotra, and T. S. Huang. Supporting ranked Boolean similarity queries in MARS. IEEE Trans. on Knowledge and Data Engineering, 10, Nov.--Dec. 1998.

....user and instantiates the model with these values. This instantiation is done at query time since the users perception di ers from user to user and query to query. Once the model is instantiated, we retrieve the top results for each predicate and then merge them to get the overall answers [14, 13, 6, 5]. We use attribute and feature interchangeably in this paper. In this paper, we assume that all the feature spaces are metric and an index (called the Feature index or F index) exists on each Due to the subjective nature of top k queries, the answers returned by the system to a user ....

Michael Ortega, Yong Rui, Kaushik Chakrabarti, Kriengkrai Porkaew, Sharad Mehrotra, , and Thomas S. Huang. Supporting ranked boolean similarity queries in mars. IEEE Trans. on Data Engineering, 10(6), December 1998.

....i = w ij similarity ij where (1) w ij = 1 and similarity ij is the measure of similarity of the object with the jth child of node N i . Many other retrieval models to generate overall similarity between an object and a query have been explored in the literature. For example, in [82], a Boolean model suitably extended with fuzzy and probabilistic interpretations is used to combine ranked lists. A Boolean operator AND ( OR ( NOT ( is associated with each node of the query tree, and the similarity is interpreted as a fuzzy value or a probability and combined with ....

....(or aggregates) the individual similarities (with respect to color and texture features) of an image to the query image to obtain its overall similarity to the query image. Examples of aggregation functions are weighted summation, probabilistic and fuzzy conjunctive and disjunctive models etc. [82]. The functions F color ; F texture and F agg and their associated weights together constitute the retrieval model (cf. Section 2.3) In order to support top k queries inside the database engine, the engine must allow users to plug in their desired retrieval models for the queries and tune the ....

[Article contains additional citation context not shown here]

Michael Ortega, Yong Rui, Kaushik Chakrabarti, Kriengkrai Porkaew, Sharad Mehrotra, and Thomas S. Huang. Supporting Ranked Boolean Similarity Queries in MARS. IEEE Trans. Knowledge and Data Engineering, 10(6):905--925, December 1998.

....focus on properties such as human fingerprints, faces, or gestures. Many techniques have been developed for both general and specific features. For example, there are many different ways to represent the color content of a multimedia object including color histogram, color moments, color sets [17], etc. This variety corresponds to the subjectivity with which humans perceive the content of multimedia objects, and each of these feature representations capture the feature from a different perspective. Extensive descriptions of feature representations appear in other chapters of this book. ....

....approaches. In the min max approach, the retrieval system keeps track of the minimum and maximum distances observed for each distance function and feature representation and normalizes the final distance by the distance range (maximum minimum distance) The Gaussian approach advocated in [17], among others, proposes to treat all distances as a Gaussian sequence. To normalize distance values, they are divided by a multiple of the standard deviation to ensure most distances (99 ) lie in the range [0,1] The drawbacks of this technique are a high up front normalization cost, and that it ....

Michael Ortega, Yong Rui, Kaushik Chakrabarti, Kriengkrai Porkaew, Sharad Mehrotra, and Thomas S. Huang, Supporting Ranked Boolean Similarity Queries in MARS, Appeared in IEEE Transaction on Knowledge and Data Engineering, Vol. 10, No. 6, Pages 905-925, December 1998.

No context found.

Ortega,M., Rui,Y., Chakrabarti,K., et al., (1998) Supporting Ranked Boolean Similarity Queries in MARS, IEEE Tran on Knowledge and Data Engineering, Vol. 10(6), pp. 905-925.

....objects makes much more sense. Accordingly, users expect to see the results of their query in ranked order, starting with the results most similar to the query rst. Multimedia similarity retrieval has been studied extensively by several researchers and projects including QBIC [10] MARS [16], WebSeek[7] DELOS [13] and HERMES [3] among others. Another eld where similarity retrieval has been explored extensively is in Text Information Retrieval [4] This work was supported by NSF CAREER award IIS 9734300, and in part by the Army Research Laboratory under Cooperative Agreement No. ....

.... related to similarity predicates in the table SIM PREDICATES( predicate name, applicable data type, is joinable) In queries with multiple similarity predicates, a scoring rule combines the individual scores of predicate matches, weighted by their relative importance, into a single score [16, 9]. We store the meta data related to scoring rules in the table SCORING RULES(rule name) De nition 4 (Scoring rule) Given a list of scores s 1 ; s 2 ; s n ; s i 2 [0; 1] and a weight for each score w 1 ; w 2 ; wn with w i 2 [0; 1] and i w i = 1, the scoring rule function returns ....

[Article contains additional citation context not shown here]

Michael Ortega, Yong Rui, Kaushik Chakrabarti, Kriengkrai Porkaew, Sharad Mehrotra, , and Thomas S. Huang. Supporting ranked boolean similarity queries in mars. IEEE Trans. on Data Engineering, 10(6), December 1998.

....makes much more sense. Accordingly, users expect to see the results of their query in ranked order, starting with the results most similar to the query rst. Multimedia similarity retrieval has been studied extensively by several researchers and projects including QBIC [14] del Bimbo [6] MARS [25], WebSeek[9] Photobook [26] DELOS [21] and HERMES [3] among others. Another eld where similarity retrieval has been explored extensively is in Text Information Retrieval [30] The initial result ranking computed in response to a query may not adequately capture the users preconceived notion of ....

.... related to similarity predicates in the table SIM PREDICATES( predicate name, applicable data type, is joinable) In queries with multiple similarity predicates, a scoring rule combines the individual scores of predicate matches, weighted by their relative importance, into a single score [25, 13]. We store the meta data related to scoring rules in the table SCORING RULES(rule name) De nition 2.4 (Scoring rule) Given a list of scores s 1 ; s 2 ; s n ; s i 2 [0; 1] and a weight for each score w 1 ; w 2 ; wn with w i 2 [0; 1] and i w i = 1, the scoring rule function returns ....

[Article contains additional citation context not shown here]

Michael Ortega, Yong Rui, Kaushik Chakrabarti, Kriengkrai Porkaew, Sharad Mehrotra, , and Thomas S. Huang. Supporting ranked boolean similarity queries in mars. IEEE Trans. on Data Engineering, 10(6), December 1998.

No context found.

M. Ortega, Y. Rui, K. Chakrabarti, K. Porkaew, S. Mehrotra, T.S. Huang, Supporting ranked boolean similarity queries in MARS, IEEE Trans. on Knowledge and Data Engineering 10(1998) 905-925.

No context found.

Ortega M., Rui Y., Chakrabarti K., Porkaew K., Mehrotra S., Huang T.S.: Supporting ranked boolean similarity queries in MARS. IEEE Trans. on KDE 10(6) 905-925 (1998)

No context found.

M. Ortega, Y. Rui, K. Chakrabarti, K. Porkaew, S. Mehrotra, and T. S. Huang. Supporting Ranked Boolean Similarity Queries in MARS. IEEE Transactions on Knowledge and Data Engineering, 10(6):905-- 925, 1998.

No context found.

M. Ortega, Y. Rui, K. Chakrabarti, K. Porkaew, S. Mehrotra, T. S. Huang, Supporting ranked boolean similarity queries in MARS, IEEE Transactions on Knowledge and Data Engineering 10 (6) (1998) 905--925.

No context found.

M. Ortega, Y. Rui, K. Chakrabarti, K. Porkaew, S. Mehrotra, and T. S. Huang. Supporting ranked Boolean similarity queries in MARS. TKDE, 10(6):905--925, 1998.

No context found.

M. Ortega, Y. Rui, K. Chakrabarti, A. Warshavsky, S. Mehrotra, and T. Huang. Supporting ranked boolean similarity queries in mars. IEEE Trans. on Knowledge and Data Engineering, 10(6):905--925, December 1999.

No context found.

M. Ortega, Y. Rul, K. Chakrabarti, K. Porkaew, S. Mehrotra, and T. S. Huang. Supporting ranked boolean similarity queries in MARS. IEEE Transactions on Knowledge and Data Engineering, 10(6):905-925, 1998.

No context found.

Ortega M, Rui Y, Chakrabarti K, Porkaew K, Mehrotra S, Huang TS (1998) Supporting ranked Boolean similarity queries in MARS. IEEE Transactions on Knowledge and Data Engineering 10:905-925.

No context found.

M. Ortega, Y. Rui, K. Chakrabarti, A. Warshavsky, S. Mehrotra, and T. S. Huang. Supporting ranked boolean similarity queries in mars. IEEE Transaction on Knowledge and Data Engineering, 10(6):905-925, December 1999.

No context found.

M. Ortega, R. Yong, K. Chakrabarti, K. Porkaew, S. Mehrotra, T.S. Huang, "Supporting Ranked Boolean Similarity Queries in MARS", D@@@#U...h+hp#v'+#'#F'#yrqtr#hq#9h#h#@tvrr...vt, ##%, 905-925, 1998.

No context found.

M. Ortega, Y. Rui, K. Chakrabarti, K. Porkaew, S. Mehrotra, and T. S. Huang, "Supporting ranked Boolean similarity queries in MARS," IEEE Trans. on Knowledge and Data Engineering 10, Nov.--Dec. 1998.

First 50 documents

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC