KEYWORD BASED OPTIMAL SEARCH FOR EXTRACTING NEAREST NEIGHBOR

Transcription

1 KEYWORD BASED OPTIMAL SEARCH FOR EXTRACTING NEAREST NEIGHBOR 1 M. SHILPA, 2 M. VASAVI 1 PG Scholar, Department of CSE, TKR College of Engineering & Technology, Hyderabad, India. 2 Associate Professor, Department of CSE, TKR College of Engineering & Technology, Hyderabad, India. ABSTRACT Conservative spatial queries, such as range search and most proximate neighbor reclamation, involve only conditions on object s numerical properties. Today, many modern applications call for novel forms of queries that aim to find objects gratifying both a spatial predicate, and a predicate on their associated texts. For example, in lieu of considering all the restaurants, a most proximate neighbor query would instead ask for the restaurant that is the most proximate among those whose menus contain steak, spaghetti, brandy all at the same time. Currently the best solution to such queries is predicated on the InformationRetrieval2-tree, which, has a few deficiencies that solemnly impact its efficiency. Motivated by this, there is a development of an incipient access method called the spatial inverted index that elongates the conventional inverted index to cope with multidimensional data, and comes with algorithms that can answer most proximate neighbor queries with keywords in authentic time. As verified by experiments, the proposed techniques outperform the Information Retrieval2-tree in query time significantly, often by a factor of orders of magnitude. Index Terms Information retrieval, spatial index, keyword search INTRODUCTION Keyword search in document performed with various approaches ranked retrieval results, clustering search results & identifying the nearest neighbor Keyword search on xml document categorized as two different approaches one is Keyword search on xml document which can be performed by ranking the searched results based on match or the answer to keyword & finding the nearest neighbor of keyword by using GST or by Xpath Query. In paper [2] the problem of returning clustered results for keyword search on documents the core of the semantics is the conceptually related relationship between keyword matches, which is based on the conceptual relationship between nodes in trees. Then, we propose a new clustering methodology for search results, which clusters results according to the way they match the given query. A spatial database supervises multidimensional objects (such as points, rectangles, etc.), and give fast access to those objects based on diverse selection criteria. The significance of spatial databases is reflected by the handiness of modeling entities of certainty in a geometric manner. For example, locations of restaurants, hotels, hospitals and so on are often characterize as points in a map, while bigger extents such as parks, lakes, and landscapes repeatedly as a combination of rectangles. Many functionalities of a spatial database are helpful in different ways in precise contexts. For instance, in a geography information system, range hunt can be organize to locate all restaurants in a definite area, while adjacent neighbor reclamation can learn the restaurant closest to a specified address. Nowadays, the extensive use of search engines has made it sensible to write spatial queries in a brand novel way. Traditionally, queries focus on objects geometric possessions only, such as whether a point is in a rectangle, or how close two points are from each other. We have seen some prevailing applications that entitle for the capability to choose objects based on both of their geometric coordinates and their related texts. For example, it would be somewhat useful if a search engine can be make use of to find the adjacent restaurant that proffer steak, spaghetti, and brandy all at the same time. Note that this is not the globally nearest restaurant, but the neighboring restaurant among only those providing all the ordered foods and drinks. There are easy ways to support queries that combine spatial and text features. For example, for the above query, we could first fetch all the restaurants whose menus contain the set of keywords {steak, spaghetti, brandy}, and then from the retrieved restaurants, find the nearest one. Similarly, one could also do it reversely by targeting first the spatial conditions browse all the restaurants in

2 ascending order of their distances to the query point until encountering one whose menu has all the keywords. The major drawback of these straightforward approaches is that they will fail to provide real time answers on difficult inputs. A typical example is that the real nearest neighbor lies quite far away from the query point, while all the closer neighbors are missing at least one of the query keywords. Spatial queries with keywords have not been extensively explored. In the past years, the community has sparked enthusiasm in studying keyword search in relational databases. It is until recently that attention was diverted to multidimensional data. The boom of Internet has given rise to an ever increasing amount of text data associated with multiple dimensions (attributes), for example, customer reviews in shopping websites (e.g., Amazon) are always associated with attributes like price, model, and rate. A traditional OLAP data cube can be naturally extended to summarize and navigate structured data together with unstructured text data. Such a cube model is called text cube [1]. A cell in the text cube aggregates a set of documents/items with matching attribute values in a subset of dimensions. Keyword query, one of the most popular and easy-to-use ways to retrieve useful information from a collection of plain documents, is being extended to RDBMSs to retrieve information from text-rich attributes [2], [3]. Given a set of keywords, existing methods aim to find relevant items or joins of items (e.g., linked by foreign keys) that contain all or some of the keywords. Traditional IR techniques can be used to rank documents according to the relevance. In a large text database, however, the number of relevant documents to a query could be large, and a user has to spend much time reading them. If a document is associated with attribute information, in a data cube model (a multidimensional space induced by the attributes), e.g., the text cube, a cell aggregates the documents with matching values in a subset of attributes. Such a collection of documents is associated with each cell, corresponding to an object that can be directly recommended to the user for the given query. This paper studies the problem of keyword-based top-k search in text cube, i.e., given a keyword query, find the top-k most relevant cells in a text cube. When users want to retrieve information from a text cube using keyword question, believe that relevant cells, rather than relevant documents, are preferred as the answers, because: 1) relevant cells are easy for users to browse; and 2) relevant cells provide users insights about the relationship data. between the values of relational attributes and the text Problem Definitions Let P be a set of multidimensional points. As our goal is to amalgamate keyword search with the subsisting location-finding accommodations on facilities such as hospitals, restaurants, hotels, etc., will fixate on dimensionality 2, but our technique can be elongated to arbitrary dimensionalities with no technical obstruction. will postulate that the points in P have integer coordinates, such that each coordinate ranges in [0, t], where t is a sizably voluminous integer. This is not as restrictive as it may seem, because even if one would relish to insist on authentic-valued coordinates, the set of different coordinates represent able under a space constraint is still finite and enumerable; therefore, could as well convert everything to integers with felicitous scaling. As with [12], each point p P is associated with a set of words, which is denoted as Wp and termed the document of p. For example, if p stands for a restaurant, Wp can be its menu, or if p is a hotel, Wp can be the description of its accommodations and facilities, or if p is a hospital, Wp can be the list of its out-patient specialties. It is pellucid that Wp may potentially contain numerous words. Traditional nearest neighbor search returns the data point closest to a query point. Following [12], we extend the problem to include predicates on objects texts. Formally, in our context, a nearest neighbor (NN) query specifies a point q and a set Wq of keywords (refer to Wq as the document of the query). It returns the point in Pq that is the nearest to q, where Pq is defined as Pq = {p P Wq Wp } (1) In other words, Pq is the set of objects in P whose documents contain all the keywords in Wq. In the case where Pq is empty, the query returns nothing. The problem

3 definition can be generalized to k nearest neighbor (knn) search, which finds the k points in Pq closest to q; if Pq has less than k points, the entire Pq should be returned. Related Work Many applications need finding objects that are most proximate to a given location that contains a group of keywords. An incrementing variety of applications need the economical execution of most proximate neighbour (NN) queries affected by the properties of the spatial objects. Owing to the apperception of keyword search, eminently on the net, several of those applications enable the utilizer to engender a listing of keywords that the spatial objects ought to contain, in their description or alternative attribute. For example, authentic estate websites enable users to go probing for properties with categorical keywords in their description and rank them in line with their distance from a given location. We incline to decision such queries spatial keyword queries [12]. A spatial keyword query consists of a query space and a group of keywords. The solution could be a list of objects hierarchical in line with a commix of their distance to the query space and additionally the connection of their text description to the query keywords. A simple nevertheless widespread variant that is employed is the distance first spatial keyword query, wherever objects square measure hierarchical by distance and keywords square measure applied as a conjunctive filter to eliminate objects that don t contain them. Unfortunately there is no economical support for top-k abstraction keyword queries, wherever a prefix of the results list is needed. Instead, current systems use ad -hoc combos of most proximate neighbor (NN) and keyword search techniques to tackle the matter. For an example, associate points R-Tree is employed to seek out the most proximate neighbors associate points for every neighbor an inverted index is employed to envision if the query keywords area unit contained. The economical methodology to answer top-k spatial keyword queries is predicated on the amalgamation of erudition structures and algorithms utilized in spatial in - formation search and Information Retrieval (IR). Particularly, the strategy consists of building an Information Retrieval R- Tree (IR2- Tree) that could be a structure fortified the R- Tree. At query time an incremental algorithm is employed that utilizes the IR2-Tree to efficiently engender the top results of the query. The IR2-Tree is a R-Tree wherever a signature is supplementary to every node v of the IR2- Tree to denote the matter content of all spatial objects within the sub tree nonmoving at v. The top-k spatial keyword search formula that is impressed by the work of Hjaltason and Samet [14] exploits this data to find the highest query results by accessing a bottom portion of the IR2-Tree. R-TREE: R-Tree makes utilization of solely Associate in Nursing R- Tree organization [16]. Given a distance-first top-k spatial keyword query, the algorithmic rule initial finds the top-1 most proximate neighbour object to the query purport Q.p. Then it retrieves that object (since the R-tree solely contains object point- ers) and compares that object s matter description with the keywords of the query. If the comparison fails then that object is discarded, and therefore the next most proximate object is retrieved. The progressive NN algorithmic rule is employed. This method perpetuates till Associate in nursing object is found whose matter description contains the query keywords. Once a gratifying object is found it s came back and therefore the method reiterates till k obj ects are came back. The drawback of this algorithmic rule is that it s to retrieve each object came back by the NN algorithmic rule till the top-k result objects are found. This doubtless will result in the retrieval of the many useless objects. Within the worst case (when none of the objects satiates the query keywords) the whole tree must be traversed and each object must be inspected. Proposed System A spatial info manages dimensional objects (such as points, rectangles, etc.), and provides expeditious access to those objects fortified exhaustively different cull criteria. The paramount of spatial databases is mirrored by the accommodation of modeling entities of authenticity in an exceedingly geometric manner. for instance, locations of restaurants, hotels, hospitals so on square measure typically described as points in an exceedingly map, whereas more astronomically immense extents like parks, lakes, and landscapes typically as a commix of rectangles. several functionalities of a spatial info square measure subsidiary in varied ways in which in categorical contexts. as an example, in an exceedingly geographic system, vary search will be deployed to probe out all restaurants in an exceedingly sure space, whereas most proximate neighbor retrieval will discover the eating place most proximate to a given address. Furthermore, because the SI-index relies on the traditional

4 technology of inverted index, it's without delay incorporable in an exceedingly business computer programmed that applies astronomically immense kindred attribute, implicatively insinuating its immediate industrial deserves. Since verification is the performance bottleneck, should endeavor to eschew it. There is a simple way to do so in an I- index: one only needs to store the coordinates of each point together with each of its appearances in the inverted lists. The presence of coordinates in the inverted lists naturally motivates the engenderment of an R-tree on each list indexing the points. Next, discuss how to perform keyword-predicated most proximate neighbor search with such a cumulated structure. The R-trees sanction us to remedy an inelegance in the way NN queries are processed with an I-index. Recall that, to answer a query, currently first get all the points carrying all the query words in Wq by merging several lists (one for each word in Wq ). This appears to be intransigent if the point, verbally express p, of the final result lies fairly proximate to the query point q. It would be great if could discover p very anon in all the germane lists so that the algorithm can terminate right away. This would become an authenticity if could browse the lists synchronously by distances as opposed to by ids. In particular, as long as could access the points of all lists in ascending order of their distances to q (breaking ties by ids), such a p would be facilely discovered as its copies in all the lists would definitely emerge consecutively in our access order. So all have to do is to keep counting how many replicas of the same point have popped up perpetually, and terminate by reporting the point once the count reaches Wq. At any moment, it is enough to recollect only one count, because whenever an incipient point emerges, it is safe to forget about the antecedent one. As an example, surmise that optate to perform NN search whose query point q is as shown in Figure 1, and whose Wq equals {c, d}. Hence, will be utilizing the lists of words c and d in Figure 4. Instead of expanding these lists by ids, the incipient access order is by distance to q, namely, p2, p3, p6, p6, p5, p8, p8. The processing culminates as anon as the second p6 emerges, without reading the remaining points. Apparently, if k most proximate neighbors are wanted, termination transpires after having reported k points in the same fashion. Distance browsing is facile with R-trees. In fact, the bestfirst algorithm is precisely designed to output data points in ascending order of their distances to q. However, must coordinate the execution of best-first on Wq R-trees to obtain a global access order. This can be facilely achieved by, for example, at each step taking a peek at the next point to be returned from each tree, and output the one that should come next globally. This algorithm is expected to work well if the query keyword set Wq is diminutive. For sizable Wq, the astronomically immense number of arbitrary accesses it performs may inundate all the gains over the sequential algorithm with merging. A earnest drawback of the R-tree approach is its space cost. Notice that a point needs to be duplicated once for every word in its text description, resulting in very expensive space consumption. In the next section, will surmount the quandary by designing a variant of the inverted index that fortifies compressed coordinate embedding. Conclusion Visually perceived plenty of applications calling for a search engine that is able to efficiently support novel forms of spatial queries that are integrated with keyword search. The subsisting solutions to such queries either incur prohibitive space consumption or are unable to give authentic time answers. In this paper, have remedied the situation by developing an access method called the spatial inverted index (SI-index). Not only that the SI-index is fairly space economical, but additionally it has the ability to perform keyword-augmented most proximate neighbor search in time that is at the order of dozens of milliseconds. Furthermore, as

5 REFERENCES [1] S. Agrawal S. Chaudhuri, and G. Das. Dbxplorer: A system for keyword-based search over relational databases. In Proc. Of International Conference on Data Engineering (ICDE), pages 5 16, [2] N. Beckmann, H. Kriegel, R. Schneider, and B. Seeger. The R*-tree: An efficient and robust accessmethod for points and rectangles. In Proc. of ACM Management of Data (SIGMOD), pages , [3] G. Bhalotia, A. Hulgeri, C. Nakhe, S. Chakrabarti, and S. Sudarshan. Keyword searching and browsing in databases using banks. In Proc. of International Conference on Data Engineering (ICDE), pages , [4] Y.-Y. Chen, T. Suel, and A. Markowetz. Efficient query processing in geographic web search engines. In Proc. of ACM Management of Data (SIGMOD), pages , [5] E. Chu, A. Baid, X. Chai, A. Doan, and J. Naughton. Combining keyw ord search and forms for ad hoc querying of databases. In Proc. of ACM Management of Data (SIGMOD), [6] G. Cong, C. S. Jensen, and D. Wu. Efficient retrieval of the top -k most relevant spatial web objects. PVLDB, 2(1): , [7] C. Faloutsos and S. Christodoulakis. Signature files: An access method for documents and its analytical performance evaluation. ACM Trans- actions on Information Systems (TOIS), 2(4): , [8] I. D. Felipe, V. Hristidis, and N. Rishe. Keyword search on spatial databases. In Proc. of International Conference on Data Engineering (ICDE), pages , [9] R. Hariharan, B. Hore, C. Li, and S. Mehrotra. Processing spatial keyword (SK) queries in geographic information retrieval (GIR) systems. In Proc. of Scientific and Statistical Database Management (SSDBM), [10] G. R. Hjaltason and H. Samet. Distance browsing in spatial databases. ACM Transactions on Database Systems (TODS), 24(2): , [11] V. Hristidis and Y. Papakonstantinou. Discover: Keyword search in relational databases. In Proc. of Very Large Data Bases (VLDB), pages , [12] I. Kamel and C. Faloutsos. Hilbert R-tree: An improved r-tree using fractals. In Proc. of Very Large Data Bases (VLDB), pages , [13] C. Faloutsos and S. Christodoulakis. Signature files: An access method for documents and its analytical performance evaluation.acm Transactions on Information Systems (TOIS), 2(4): , [14] I. D. Felipe, V. Hristidis, and N. Rishe. Keyword search on spatialdatabases. In Proc. of International Conference on Data Engineering(ICDE), pages , [15] R. Hariharan, B. Hore, C. Li, and S. Mehrotra. Processing spatialkeyword (SK) queries in geographic information retrieval (GIR) systems. In Proc. of Scientific and Statistical Database Management(SSDBM), [16] G. R. Hjaltason and H. Samet. Distance browsing in spatial databases. ACM Transactions on Database Systems (TODS), 24(2): , [17] V. Hristidis and Y. Papakonstantinou. Discover: Keyword search in relational databases. In Proc. of Very Large Data Bases (VLDB), pages , [18] I. Kamel and C. Faloutsos. Hilbert R-tree: An improved r-tree using fractals. In Proc. of Very Large Data Bases (VLDB), pages , [19] J. Lu, Y. Lu, and G. Cong. Reverse spatial and textual k nearest neighbor search. In Proc. of ACM Management of Data (SIGMOD), pages , 2011.