Research On Retrieval Method Based On Diversity And Proportionality Based On Semantic And Spatial Distribution

The emergence of spatial information retrieval technology has brought great convenience to our lives,likes knowing the location information of the world with no leaving the home.The technology is mainly based on spatial data sets with spatial position and semantic attributes.Under the premise of location given by users,how to get the l(nature number)information(tuples)that makes the user satisfied.Due to the fact that the user's intention is not clear in most cases when searching,this brings a great challenge to the retrieval technology.Therefore,the main research of this paper in spatial information retrieval technology is as follows:(1)A new offline sorting strategy named Value Rank is proposed.This method is mainly used to calculate the initial weight of each node in the dataset,which avoids the singleness of most of the current technologies that simply select the results based on user ratings.Value Rank is an extension of Object Rank,which considers the concept of dynamic “value”,that is,when calculating the VR values of certain attribute nodes,it not only considers the quantitative relationship and the given static value flow rate(Value flow rate is the degree of mutual contribution between nodes in relational data schema diagrams.See Chapter 1 for details.),but also considers its value and forms a dynamic value flow rate.For the Northwind dataset,the evaluation for a consumer is not just the number of orders,but mainly based on the total value of all orders to calculate its weight.Calculating the initial weights using Value Rank not only avoids the use of user scores as a single unit of initial weights,but also provides a reliable data theory support to prepare for subsequent searches.(2)Proposes the method for calculating the semantic diversity and equal proportions of search results.Since the existing retrieval techniques are based on tuples' weight,the top k is returned as the result.The drawback is that it may cause the result to be aggregated in a certain type of semantics.At this time,when the user's intention is unknown,the result is difficult to meet the needs of users.Therefore,a method for calculating the semantic diversity and proportional of results is proposed.This method means that the user can get l(natural number)tuples as semantically as possible in a given location(in spatial dataset)or in a keyword(in plain text dataset).The result of semantic diversity is to consider that when a tuple of a certain type(tuples have semantic similarity will be identified as the same type,see Chapter 4 for details)appears multiple times in the result set,then the coefficient of its weight(even if its weight is smaller)will be dynamically reduced in the next selection of tuples in the same type from the candidate set,in order to achieve the semantic diversity of the results;the results of the semantic proportional of characteristics when considering the tuples of a certain type with higher frequency but lower weight,then it can also indicate that there are some certain links between such tuples of this type and keywords or positions of the search.Therefore,this paper will dynamically increase the coefficients of the weights of such tuples from the candidate set,in order to achieve the semantic proportional of the results.(3)Proposes the method for calculating the spatial distribution diversity and equal proportions of search results.This method is mainly aimed at the spatial dataset.Spatial retrieval is mostly sorted according to the distance,and gets the first l points closest to the retrieval point to form the result set.It can be seen that whether getting results in this way or by the weight may cause the results to be aggregated in a certain spatial distribution.Therefore,the method for calculating the spatial distribution diversity and equal proportions of search results is proposed,In terms of diversity,the candidate tuple is selected according to the characteristics of the Euclidean distance formula.In the aspect of proportional,the spatial distribution is divided into four directions with the retrieval point as the center,and results are generated using the method of equal proportions of the result semantics.Finally,Dsize-l OS is composed by l tuples considering combining the diversity of semantic and spatial distribution,Psize-l OS is composed by l tuples considering combining the proportional of semantic and spatial distribution.Experimental results prove that the retrieval method proposed in this paper is effective.