There are some possibilities to improve the systemn of tags (and meta-information in general), it would be interesting what you think about it. What's your experience and hopes regarding tags and can the proposed match your hopes?
I.1 Tags could be entities with different representations. For example the same tag entity could be represented by grayscale, greyscale or Grey-scale. This could cover not only different spellings but also synonyms (funny, humourous, comical) or in future other languages (like already with existing tag systems e.g. in drupal or on kongregate).
This would enable search to abstract from spelling, synonyms or even language (particularly tag-based search).
I.2 Advanced search options for tags, allowing to connect tags by (operators and, or, not) could significantly improve the usefulness and fun of searches.
I.3 Tags could be organized in a hierarchical structure. E.g. animals are a popular motif for drawings. According tags could be structured into a hierarchy like animal>cat,bird>peacock,parrot and so on. People would only need to choose one tag (peacock) and this choice would automatically imply the parents (bird and animal). And hence reduce the number of explicit tags on a drawing and increase the information carried by a tag.
Other hierarchies could refer to style/technique (realistic,abstract>geometric>rectangular,triangular) or genre (fictional>fantasy,sci-fi*)
a) Also this would enable improved search possibilities and even allow new exciting options to browse drawings. (at least, if things like the epically failed hierarchy at deviantart, mixing dimensions of genre and style, is avoided) Think of browsing drawings by content, genre or style. Or search for a speficic combination of these properties.
b)Additionally, being part of a hierarchy could load a tag with semantics which could be used to distinguish the tag card referring to the drawing being meant as a greeting card from the tag card referring to the drawing motif containing a playing card. However, for semantics I would prefer a different approach described below:
Replies
II: Semantic tags and their benefits
II.1 The concept is a semantic approach to meta-information in general. Particularly tags of a drawing would be semantic relations from a drawing to a tag. Such relations have semantic meaning, as contains-in-motif or drawn-in-style.
Being a general approach on meta-information, the complete meta-information could be modeled semantically, including authors, continuations, challenges, albums, inspirations and anything to come in future.
Semantic relations can also deal with elementary types, such as strings, dates or numbers. Appropriate languages for ontologies or requests have been standardized by the W3C as part of the Semantic Web (or "Web 3.0"), so this field is already well-explored and a range of tools and libraries exist to speed-up development as well as theres already productive use (e.g. XMP, the ISO standard for image metadata or Nepomuk, providing KDE with semantic desktop capabilities).
II.2: Benefits
a) The use of generalized concept and technology allows easier and faster introduction of new features. A single improved functionality can affect multiple places.
b) Drawings could be browsed by dimensions (=relation types, which could theirselves be organized hierarchically). So, conceptually and technically, browsing for drawings by a user would be the same as browsing drawings of a challenge, inspiration, drawings containing kittens or drawings in rectangular style. Browsing by content or style could invite people to explore the huge content already available on this site in a comfortable way.
c) Basing search on a powerful, standardized request language would allow exciting advanced search options. Ever wanted to search for a tigers drawn in geometric style and with a rating above 10, sorted by date and grouped by user?
d) Such search options could be used to improve any browse pages with dynamically applicable filters, sorting or grouping options. Let's dream:
III: Automated retrieval of semantic meta-information
There is already automatically retrieved meta-information such as author, date or rating of a drawing.
- This could be extended, particularly to extract features like dominant colour, brightness, width, height or number of curves. Such features could be modeled as semantic relations to elementary types.
- Also tags could be extracted or derived from extracted features. (e.g. extract has-style→grayscale or derive has-format→landscape from the width and height.
- Now search options could be even more exciting, allowing to search something like a geometric tiger in landscape format and sort by brightness.
- Such must not be limited to drawings: From a users location its timezone could be derived, or from an album the number of entries could be derived or from a challenge the average brightness or number of curves could be extracted.
I know, such would already be very advanced, but it illustrates the possibilities and future options that can arise from a choice regarding technology and modeling.
