Hi, I am quite new at the field of semantic web and i need some help in order to conduct my research. I came up with the idea of a tangible interface which would take as an input the physiological sense of the user and recognize the emotion of it and the system can recommended simultaneously a playlist of songs and a collection of pictures. My basic idea is when for example a user is sad, the system can recommended a playlist of happy songs and a collection of pictures that provides positive feelings to the user. As for the methodology of the system, I thought about align 3 ontologies together (emotion, music, pictures). Is it right the way that I am thinking or there is another way like tag acquisition? If it is possible, what sources shall i look into? And about the recommendation method, should i look for the semantic-based recommendation?

Also, in the recommendation machines why knowledge base(construction of ontologies) is an important component?

Thank a lot for your valuable help, Debbie

asked 19 Dec '12, 17:11

debbieSemantic's gravatar image

accept rate: 0%

Hi in my opinion what you are actually thinking on it's a machine-learning system, so ontologies here are not needed and could be somewhat countert intuitive to use. For example stereomood: http://www.stereomood.com/ uses an approache based on machine learning, if i'm not wrong. The task it's interesting and there are many research on that since some years, the basic approach could be:

  1. analyze features in the signal / lower abstraction point of view
  2. analyze features from simbolic point of view in the music domain (here one could think about music ontology and MusicXML serialization, and some musical analysis regarding "functional" harmonies etc)
  3. analyze the relations between different metadata schema. In that context i think you could use the three taxonomy you think (i think you intended them as taxonomies, but maybe i'm wrong)
  4. design a tagging component in your system, so user should provide their own feature to your learning and/or inference phase.

Starting from here there are a lot of options to be considered to the actual implementation: i think you basically need a machine learning engine (suche as weka, for example), in order to collect emerging evidence from the data and the metadata, that you could use in some kind of cluster drive approach (with information retrieval) or to map them into a commonsense ontology (which i think it's really hard to construct on topic so fuzzy and personal by means of definition)

I hope this help

permanent link

answered 24 Dec '12, 14:46

seralf's gravatar image

accept rate: 13%

Seralf, thanks so much for your valuable information.

(25 Dec '12, 11:29) debbieSemantic debbieSemantic's gravatar image
Your answer
toggle preview

Follow this question

By Email:

Once you sign in you will be able to subscribe for any updates here



Answers and Comments

Markdown Basics

  • *italic* or _italic_
  • **bold** or __bold__
  • link:[text](http://url.com/ "title")
  • image?![alt text](/path/img.jpg "title")
  • numbered list: 1. Foo 2. Bar
  • to add a line break simply add two spaces to where you would like the new line to be.
  • basic HTML tags are also supported

Question tags:


question asked: 19 Dec '12, 17:11

question was seen: 1,103 times

last updated: 25 Dec '12, 11:29