Download presentation
Presentation is loading. Please wait.
Published byKristin Rice Modified over 8 years ago
1
Automated Creation of a Forms- based Database Query Interface Magesh Jayapandian H.V. Jagadish Univ. of Michigan VLDB 2008 1
2
Outline Motivation Database Analysis Queriability Generate Query-Forms Experiments Future Works (My ideas) 2
3
Forms-based Database Query 1. What’s Forms-based database query? 2. Why we need Forms-based database query?
4
Why we need Forms-based database query? for $a in doc()//author, $s in doc()//store let $b in $s/book where $s/contact/@name = “Amazon” and $b/author = $a/id return { $a/name, count($b) } $a ?? What is let? Do I need a semi-colon? How do I start writing a query?
5
What’s Forms-based Database Query?
6
Automated Forms-based Database Query Why we need automated Forms-based database query?
7
Automated Forms-based Database Query Why we need automated Forms-based database query? Too many tables in database ! Too many attributes in each table !
8
Automated Forms-based Database Query Why we need automated Forms-based database query? Too many tables in database ! Too many attributes in each table ! Design simple forms to cover most queries required by users ?
9
Automated Forms-based Database Query Why we need automated Forms-based database query? Too many tables in database ! Too many attributes in each table ! difficulty to human Design simple forms to cover most queries required by users ?
10
Survey on real-world’s websites Page 2 10
11
Survey on real-world’s websites Even those complex query-forms cannot handle these users’ queries: Page 2 11
12
Outline Motivation Database Analysis Queriability Generate Query-Forms Experiments Future Works (My ideas) 12
13
Schema Analysis E is a set of entities A is a set of attributes, each belonging to a single entity L is a set of links between nodes (entities or attributes) in graph. 13
14
The Graph of Schema 14
15
The Task Find the most queriable entities and attributes !! 15
16
What’s Queriability? The possibility of a node appears in a query. 16
17
The Basic Idea This queriabilty is inspired by the approach taken by several search engines to rank web documents. A document is considered “important” if it is connected (linked) to other “important” documents 17
18
The Basic Idea This queriabilty is inspired by the approach taken by several search engines to rank web documents. A document is considered “important” if it is connected (linked) to other “important” documents 18 PageRank?
19
Outline Motivation Database Analysis Queriability Generate Query-Forms Experiments Future Works (My ideas) 19
20
Two Postulates POSTULATE 1. The query relevance of an entity depends on how well-connected it is to other parts of the schema.(similar to PageRank) POSTULATE 2. The query relevance of an entity depends on how many instances (records) of it occur in the database. In Page 2. 20
21
Absolute Cardinality Absolute Cardinality : C(n) =The number of instances contain this node n in database. 21
22
Relative Cardinality Relative Cardinality : RC(n i ->n) = C(n i ->n) / C(n) Here, C(n i ->n) = The number of instances contain both n i and n in database. 22
23
Queriability of Entities 23 In Page 3
24
Queriability of Entities 24 In Page 3 p is a user-defined parameter between 0 and 1
25
Queriability of Entities 25 In Page 3 The initial importance of n : I n 0 = C(n)
26
Queriability of Entities 26 In Page 3 Measured by the weight of the link from node n i to node n
27
Queriability of Entities 27 In Page 3 The sum of all nodes’ absolute cardinality C(n i ) I e c = I n r, when I n r converges.
28
Queriability of Related Entities 28 a single entity per query-form?
29
Queriability of Related Entities 29 a single entity per query-form? Not appropriate!
30
Queriability of Related Entities 30 POSTULATE 3. The queriability of a collection of related entities depends on the individual queriabilities of entities in it. POSTULATE 4. The queriability of a collection of related entities depends on the data cardinality of all pair-wise relationships between the entities in it. In Page 4
31
Queriability of Related Entities 31 In Page 4 N(e i -> e j ) is the number of instances of entity e i connected to some instance of entity e j
32
Queriability of Related Entities 32 In Page 5 Considering more than 2 entities’ relationship
33
Queriability of Related Entities 33 In Page 5 Considering more than 2 entities’ relationship The number of permutations of m objects, i.e. m!
34
Queriability of Attributes 34 POSTULATE 5. The queriability of an attribute depends on its necessity, i.e., how frequently it appears in the data relative to its parent entity. In Page 5
35
Queriability of Attributes 35 a is an attribute of entity e In Page 5
36
The Queriability of Operator- Specific Attribute 36 Operations: Selection, Projection, Sorting, Aggregation. The queriability of different operation is different.
37
The Queriability of Operator- Specific Attribute 37 In Page 6
38
The Queriability of Operator- Specific Attribute 38 Selection Projection Sorting Aggregation In Page 6,7
39
Outline Motivation Database Analysis Queriability Generate Query-Forms Experiments Future Works (My ideas) 39
40
Choosing Form Fields 40 k a most queriable attributes. k f the number of fields (of any type) per entity in a form. k e the number of entities in a form. k r the number of related-entities in a form In Page 7
41
Outline Motivation Database Analysis Queriability Generate Query-Forms Experiments Future Works (My ideas) 41
42
Experimental Methodology Evaluate the usefulness of generated query- forms. – See how many real users’ queries in can be satisfied by generated query-forms. 42
43
Testing Datasets MiMI. Geoquery Jobsquery. 43
44
Form Usefulness Testing 44
45
Form Usefulness Testing 45
46
Form Usefulness Testing 46
47
Form Usefulness Testing 47
48
Form Usefulness Testing 48
49
Form Usefulness Testing 49
50
Effect of Postulates Testing 50
51
Effect of Postulates Testing 51
52
Effect of Postulates Testing 52
53
Effect of Postulates Testing 53
54
Outline Motivation Database Analysis Queriability Generate Query-Forms Experiments Future Works (My ideas) 54
55
Future Works (My ideas) 1. How to use history log? 2. Can we use association mining for Related Entities? 3. The application in BCiN Project 55
56
(1) How to use history data? Besides PageRank, the search log also helps search engines to make better ranking of pages. Therefore, can we use database query log to make better query-form? Personalized query-form (different roles in application)? 56
57
The Graph of Schema 57 09/12/09: SELECT * FROM profile WHERE profile.income > 1W
58
(2) Can we use association mining for Related Entities? 58 This paper tries all the possible combinations of entities. We can use Apriori algorithm.
59
(3) The application in BCiN Project 59 Generate different query-forms for different device ( Type keywords in Mobile Phone is hard).
60
(3) The application in BCiN Project 60 Generate different query-forms in different period (hurricane coming, hurricane leaving, disaster recovery)
61
End 61 Thank you! Any question?
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.