**5. Conclusion**

of maximal segment (*regions*) are computed using DNF expression whose union is a cluster shown in **Figure 6(c)** at higher dimension. Subsequently, move to 3rd, 4th, 5th …. dth dimension in search of relevant clusters. This consummates the exploration of each data subspace around the pertinent objects of anterior query. All the computed clusters are equivalent to the morph of initial/previous query. Morphs contains addition relevant results as well subset of originally retrieved results. The data items present in the morphs are dignified as relevant by standard measure to previous query and future probable search interest. Now based on implicit and explicit relevance top N morphs and set of relevant terms are suggested to the user. In our example, after computing relevance score using standard relevance measures we can say that query morphs containing movie genre 'Drama' and directed year >1995 scored higher number then morph with movie genre 'Thriller'. Hence, morphs with movie genre 'Drama' and year >1995 considered as high relevance. The system would also suggest top N terms from computed morphs like 'Coppola' based on relevance to the initial query as well result set. These terms help the user in formulating his next exploratory/variant query. As a next step, user may encounter shift towards different query results after reviewing result variants. The newly formulated query now surrounds both past and new variant of the user

Many design issues are identified during the development of the propose solution which are

**1. Neighborhood selection and query morph generation**: The key challenge is identifying and defining the borderline for neighborhood of relevant data objects. Various researchers have addressed in their existing work. In Proposed algorithm, subspace clustering is used to define a non-overlapping boundary based on relevance of neighborhood objects. Each neighborhood region will be explored and exploited for extraction of keywords and phrases query reformulation. If the density of d-dimensional spatial cell is less than the threshold (τ), then cluster forming becomes a challenging task. Exploring cluster at higher dimension may also face issues like cluster overlapping, cluster size, number of clusters. **2. Evaluation of relevant data objects and Top N morph suggestion**: Relevance is estimated to measure how closely data objects of different clusters are connected and also to define importance of the result items [39]. Identification of various information to define relevance criteria is one of the key challenges, as it influences overall system performance. In our approach, each cluster will be exploited as a region of user's interest and data objects will be extracted based on explicit and implicit relevance measure. Two key issues are identified during designing that are criteria selection for relevance and techniques for the

**3. Demonstration of additional information extracted from retrieved data objects through various visualization:** A visualization of entire result set with frequent terms is not a

request.

as follow:

**4. Design issues and analysis**

158 From Natural to Artificial Intelligence - Algorithms and Applications

computation of relevance score.

We proposed an algorithm for query reformulation using object's proximity, 'Query morphing' that mainly design to recommend additional relevant data objects from neighborhood of the user's query results. Each relevant data object of user query act as an exemplar query for generation of optimal intermediate reformulations. Multiple challenges are inferred during solution designing, includes: (i) neighborhood selection and Query morph generation (ii) Evaluation of relevant data objects and Top-K morph (iv) Evaluation of data object's relevance, (III). Demonstration of additional information extracted from retrieved data objects through various visualization. The discussed approach primarily based on proximity-based data exploration, and generalized approach of query creation with small edit distance. It could be realized with major adjustments to the query optimizer. The ultimate goal would be that morphing the query pulls towards the area where information is accessible at low cost.
