WIRE:WIsdom-awaRE computing

Wisdom-Aware Computing: On the Interactive Recommendation of Composition Knowledge

Soudip Roy Chowdhury, Carlos Rodríguez, Florian Daniel and Fabio Casati

WESOA 2010, December 7, 2010, San Francisco, USA

2

What are we talking about?

• Today: simplifying technology and reusing code• MDD, BPM, SOA, mashups

• But there are still two major issues• Tools typically don’t speak the language of the user• Users typically don’t speak the language of the tools

End users

Software development

3

What are we talking about?

• Today: simplifying technology and reusing code• MDD, BPM, SOA, mashups

• But there are still two major issues• Tools typically don’t speak the language of the user• Users typically don’t speak the language of the tools

End users

Software developmentWe aim to “teach”users how to develop by showing them how others solved similar problems in the past = By harvesting and recommending community composition knowledge

• We want to develop a data processing logic that• Fetches news from a news site• Adds geo-coordinates to each retrieved item (where

possible)• Filters the feed according to some keywords• Plots the resulting items onto a map

Let’s see an example: Yahoo! Pipes




Too complex for end users!




Too complex for end users!

7

So what?

• Wisdom-aware development = learn from existing mashups/compositions + advise composition knowledge• Wisdom = the knowledge of the crowd/community

• Challenges• Identifying the types of advices that can be given and the

right times when they can be given• Discovering computational knowledge• Representing and storing knowledge • Searching and retrieving knowledge• Reusing knowledge

8

• O. Greenshpan, T. Milo, N. Polyzotis. Autocompletion for mashups. VLDB’09, pp.538-549.top-k next components from a network of compatible components

• A.V. Riabov, E. Bouillet, M.D. Feblowitz, Z. Liu, A. Ranganathan. Wishful Search: Interactive Composition of Data Mashups. WWW’08, pp. 775-784.

AI planning for goal-driven composition• A.H.H. Ngu, M. P. Carlson, Q.Z. Sheng. Semantic-Based Mashup of Composite Applications. IEEE

Transactions on Services Computing, vol. 3, no. 1, Jan-Mar 2010.

Suggestion of semantically compatible components• H. Elmeleegy, A. Ivan, R. Akkiraju, R. Goodwin. MashupAdvisor: A Recommendation Tool for

Mashup Development. ICWS’08, pp. 337-344.

Semantics + prediction of user goals + AI planning

• T. Hornung, A. Koschmider, G. Lausen. Recommendation Based Process Modeling Support: Method and User Experience. ER’08, pp. 265-278.

Copy/paste of business process parts based on text label similarity

State of the art

9










State of the art

On crowd knowledge (CK)

Syntactic similarity

10










State of the art



Semantic similarity

11










State of the art



Semantic similarity

Crowd knowledge

12

But how does CK look like?

• Well, it depends. Especially on the complexity of the target composition environment.

Meta-model of Yahoo! Pipes

• Other complexities: mashArt (13), BPMN (20), BPEL (60)

13

CK as composition patterns

14

Parameter Value Pattern

15

Component Association Pattern

16

Connector Pattern

17

Data Mapping Pattern

18

Complex Pattern

19

Discovering CK

• Mine patterns from existing mashup specifications• Techniques:

• Frequent item set and association rule mining: e.g., for Component Association Patterns and Parameter-Value Patterns

• Sequential pattern mining: e.g., for Complex Patterns, Component Association Patterns, and Connector Patterns.

• Graph mining: e.g., for Complex Patterns and Connector Patterns.

• Link mining: for the discovery of any of the proposed advices, e.g., Data Mapping Patterns

• Key to success: limited complexity of mashups

20

When to give an advice?

21

Wisdom-aware dev. environment

• Extension of mashArt or OMELETTE editor

22

Status and future work

• Currently, work in it’s conceptionphase

• Understandability/acceptabilitystudy of advice paradigm with end users ongoing (mockups!)

• Next:• Knowledge extraction algorithms• Advice repository and query interface• Extension of mashup editor

End users

23

Conclusion

• We propose the idea of wisdom-aware computing, i.e., the reuse of community composition knowledge to empower end users

• If successful:• Extend developer base toward non-experts• Enable progressive learning and knowledge transfer

• No explicit semantics provided by anybody• People don’t like to tag or annotate• Semantics should derive from domain

(need for domain-specific mashup platforms!)

WIRE:WIsdom-awaRE computing

Technology