Skip to main content
 
Stage of development
Tested in real-world scenarios.

Intellectual property
Notary registration of the software..

Intended collaboration
Licensing and/or co-development

Contact
Virginia Cousté
Vice-presidency for Innovation and Transfer
Virginia.Couste@uab.cat
comercializacion@csic.es

Reference
CSIC/VC/029
Additional information
#ICT #Artificial Intelligence #Machine Learning

Collaborative assessment software

A software support tool for trust assessment in online communities. The system recommends the assessments by weighting the expert’s assessments, along with other user’s assessments over the object of evaluation.

Market need
Sometimes the task of evaluating a large number of objects is simply not possible and it is necessary to trust in the opinion from others. This component implements an algorithm that uses the assessments made by members of an online community to approximate the evaluation given by another specific member of that community, the “leader” (teacher, coordinator, buyer, etc.), taking into account the confidence relationship between him and the members.
Comparisons made with a commercial software for content recommendations (collaborative filtering) shows that the recommendations counting towards the opinion of a specific member is biased and does not take benefit of the relationships between the leader and third parties (indirect trust).

Proposed solution
This component implements an algorithm that uses the assessments made by members of an online community to approximate the evaluation given by another specific member of that community, the “leader” (teacher, coordinator, buyer, etc.), taking into account the confidence relationship between him and the members.

Competitive advantages
  • Increased number of automatic evaluations with less margin of error.
  • The component is a JAVA library with an API to be easily integrated in other applications.
  • Two implementation models
  • Implementation 1: the confidence calculation is based on a graph of confidence.
  • Implementation 2: the confidence is represented as probability distributions and an artificial intelligence algorithm calculates it.