Your SlideShare is downloading. ×
0
Idea Relationship Analysis inOpen Innovation CrowdSourcing            systemsAdam Westerski, Carlos A. Iglesias and Javier...
Outline   Idea Management Systems   The problem            too many similar ideas   Our solution            idea rela...
Idea Management Systems   Evolution of suggestion boxes...   for open innovation purposes            Collect ideas, sug...
Innovation through             Idea Management Systems                    Idea                Improvement                 ...
Goal: improve how ideas are   automatically filtered
Our research so far   Define Gi2MO ontology for formalising and    interlinking idea management systems (IMS)   Connect ...
Our research so far...   Software released as open source, everything    available at http://www.gi2mo.org
The problem   Too many ideas to be analysed   Many ideas are similar
Proposal: similarity relationships
Proposal Similarity Relationships
Evaluation methodologySemantic scrapper, available athttp://www.gsi.dit.upm.es/index.php/en/software/details/1/2/software-...
Manual annotation
Experiment I   Goal: determine if all the relationships are    useful, apart from duplicate relationship   We compare ou...
Results Experiment I   76.7% new relationships compared with the    original IMS
Experiment II   Goal: new relationships help to identify    duplicates and reduce the data set   We compare            ...
Results II   Ubuntu: 1.13% of duplicate ideas of entire    dataset   Data set with relationships: 0.5%   Data set with ...
Experiment III   558 out of 1000 ideas are not related each    other (so, they are not similar)   Analyse similarity bas...
Gi2MO Types
Results Experiment III   Calculated similarity S between two related    ideas (iA, iB) for 50 ideas for Metric Mx
Analysis   Same correlation in similar and disjoint idea    subsets with the analysed metrics
Conclusions   Semantic relationship can    improve the detection of    similar ideas   Relationship hierarchy    and tra...
Questions?
Idea Relationship Analysis in Open Innovation CrowdSourcing systems
Idea Relationship Analysis in Open Innovation CrowdSourcing systems
Idea Relationship Analysis in Open Innovation CrowdSourcing systems
Upcoming SlideShare
Loading in...5
×

Idea Relationship Analysis in Open Innovation CrowdSourcing systems

333

Published on

Abstract—Idea Management Systems are an implementation
of open innovation notion in the Web environment with the use
of crowdsourcing techniques. In this area, one of the popular
methods for coping with large amounts of data is duplicate detection.
With our research, we answer a question if there is room
to introduce more relationship types and in what degree would
this change affect the amount of idea metadata and its diversity.
Furthermore, based on hierarchical dependencies between idea
relationships and relationship transitivity we propose a number of
methods for dataset summarization. To evaluate our hypotheses
we annotate idea datasets with new relationships using the
contemporary methods of Idea Management Systems to detect
idea similarity. Having datasets with relationship annotations at
our disposal, we determine if idea features not related to idea
topic (e.g. innovation size) have any relation to how annotators
perceive types of idea similarity or dissimilarity.

Published in: Technology
0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total Views
333
On Slideshare
0
From Embeds
0
Number of Embeds
0
Actions
Shares
0
Downloads
15
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide

Transcript of "Idea Relationship Analysis in Open Innovation CrowdSourcing systems"

  1. 1. Idea Relationship Analysis inOpen Innovation CrowdSourcing systemsAdam Westerski, Carlos A. Iglesias and Javier EspinosaGrupo de Sistemas Inteligentes Autor:Universidad Politécnica de Madrid Grupo de Sistemas Inteligenteshttp://www.gsi.dit.upm.es Universidad Politécnica de Madrid
  2. 2. Outline Idea Management Systems The problem  too many similar ideas Our solution  idea relationship hierarchy Evaluation  Ubuntu brainstorm Conclusions
  3. 3. Idea Management Systems Evolution of suggestion boxes... for open innovation purposes  Collect ideas, suggestions, from customers, employees, ...
  4. 4. Innovation through Idea Management Systems Idea Improvement Idea Selection IdeaGeneration Idea IdeaDeployment Implementation
  5. 5. Goal: improve how ideas are automatically filtered
  6. 6. Our research so far Define Gi2MO ontology for formalising and interlinking idea management systems (IMS) Connect ideas with Enterprise systems following the Linked Data model Integrating Opinion Mining techniques with IMS Idea Classification, filtering and analysis of idea similarities independent of idea topic (this talk!)
  7. 7. Our research so far... Software released as open source, everything available at http://www.gi2mo.org
  8. 8. The problem Too many ideas to be analysed Many ideas are similar
  9. 9. Proposal: similarity relationships
  10. 10. Proposal Similarity Relationships
  11. 11. Evaluation methodologySemantic scrapper, available athttp://www.gsi.dit.upm.es/index.php/en/software/details/1/2/software-scrappy.html
  12. 12. Manual annotation
  13. 13. Experiment I Goal: determine if all the relationships are useful, apart from duplicate relationship We compare our manual annotation with the annotation of duplicate available in Ubuntu Brainstorm
  14. 14. Results Experiment I 76.7% new relationships compared with the original IMS
  15. 15. Experiment II Goal: new relationships help to identify duplicates and reduce the data set We compare  Detected duplicates in Ubuntu  Detected duplicates in our dataset based on the new relationships  200 ideas with 5 manual annotations suggested by Lucene  Same as before but taking into account the hierarchy
  16. 16. Results II Ubuntu: 1.13% of duplicate ideas of entire dataset Data set with relationships: 0.5% Data set with relationships + hierarchy: 1.95% → increase in 95%!
  17. 17. Experiment III 558 out of 1000 ideas are not related each other (so, they are not similar) Analyse similarity based on other annotations:  Trigger type  Innovation type  Proposal type  Object type
  18. 18. Gi2MO Types
  19. 19. Results Experiment III Calculated similarity S between two related ideas (iA, iB) for 50 ideas for Metric Mx
  20. 20. Analysis Same correlation in similar and disjoint idea subsets with the analysed metrics
  21. 21. Conclusions Semantic relationship can improve the detection of similar ideas Relationship hierarchy and transitivity improve similarity detection Other annotations (Gi2MO types) and associated metrics do not help for detecting similar ideas
  22. 22. Questions?
  1. A particular slide catching your eye?

    Clipping is a handy way to collect important slides you want to go back to later.

×