**2.2. Query approximation**

techniques track user's querying behaviour, identify the interested area from the available data space and recommends set of queries that retrieved relevant information. The query is steering [12, 26] is one process that navigates the user through complex data structures. For query recommendation and steering interactive query session is required to achieve ultimate

Due to the big data occurrences, traditional ways of query transformation repeatedly encountered challenges of relevance. To contrive such inherent challenges of transformation and relevance for exploration in large data a technique 'Query morphing' is designed. Our proposal suggests additional relevant data objects for the formulation of precise query by exploring available data space and leveraging use feedback. We concur that query morphing will also acquire the properties of traditional methodologies by observing that search query and

The main contribution of this paper is an algorithm designed for query reformulation based on exploration technique. Algorithm named 'Query morphing' explores into the proximity of initial user query and extract additional relevant data objects. These retrieved data objects from the n-dimensional neighborhood assist user in his intermediate query reformulation. Proximate data objects are selected based on implicit and explicit relevance. We expect that our proposal, guides on exploration over several ample databases, such as Medical database, DNA database, social database, scientific database, etc. Finally, various existing reformulation techniques are revisited to establish the fact that how 'Query morphing' is different from

Next section listed some related research prospects and approaches. In Section 3 the proposed approach is conveyed, in which conceptual design is represented with algorithm and schematic diagram. Various design issues, analysis of implementations as well as intrinsic implementation complexity in proposed approach are recognized in Section 4. Lastly, the

Many of today's query processing platforms carry a much profound repertoire and resilient querying techniques to regulate huge observational data in a limited resource environment. [12, 11]. Below some aspects are reviewed that helps user in searching relevant information from a large data set. We consider some prominent researches delivered for automatic exploration in data space, formulation of approximate quires and techniques that assist the user in

Analyzing vast amount of real time data can be an extremely complex task and required automation. In such scenario without any assistance, user ended up with ill-formulated query

search goal [12].

respective results analogous to the history log.

150 From Natural to Artificial Intelligence - Algorithms and Applications

**1.1. Contribution and outline**

traditional transformation techniques.

the query formulation process to cover our aspects.

conclusion is presented.

**2. Literature review**

**2.1. Automatic exploration**

In exploratory query aspects where the user is satisfied with 'closed-enough' answer, approximation modules implemented in search system help to achieve shorter response time. This approximation module is built without changing underling database architecture. For example, Aqua approximate query answering system [4] rewrites queries using summery synopsis to provide approximate answers. Automatic Query Processing (AQP) widely uses statistical techniques based on the synopsis [14] to analyze large amount of data. Four main key synopsis are used by researchers for approximation which is random sample synopsis, histogram synopsis, wavelet and sketches synopsis.

Most fundamental and commonly used synopsis is a random sampling in that subset of data objects are fetched based on stochastic mechanism. It is easy to draw samples from a small available data, although to make the sampling process scalable, advance sampling techniques are required e.g. BlinkDB [6] architecture. In this architecture samples are selected based on accuracy of query and response time that device dynamic sampling strategy. A Histogram synopsis method group the data values into subset by summarizing the attribute frequency distribution or combined attribute frequency distribution. By using advance methods such as aggregation over joints are also used to approximate more general class of query. Another synopsis is wavelet synopsis which is identical with the above but the only variation is that it transforms and express most substantial data into the frequency domain. A faster response is one characteristic of approximate query processing. Speedup with accuracy is the key objective of AQP, therefore, returned results must be verified. Interactive approximate query processing performs error estimation [5] and error diagnosis via close forms or bootstrap that guarantees runtime efficiency and resource usage.
