Adding a Tag to a New Document. A researcher has a new publication on the interactions with the software from a project called mSpace, and is adding it to her local publication repository. The user wants to tag the document. Previous papers about mSpace have been tagged with mSpace as a Semantic Web browser, but this new paper has nothing to do with the technical implementation, only the interactions and so does not want to be directly associated with it. As a result the user also adds the tags "Direct Manipulation" and "Multiple Selection" to the document. However, people will only now find this document if they enter these specific keywords in the tag search. In order to be picked up by people who search for "HCI" or "Usability", the user has to also add these tags to the document, as well as every other word that might be used to define this. In Semantic Tagging, however, as users begin to type a tag, optional tags are presented to the user for selection; this is shown in Figure 1.
Once the user has reached the tag page, as the user starts typing (level 3 of the diagram) the system looks up the tags and returns them to the EPrints interface to display. The user can see each of these tags and what they refer to; the user may have a choice in tagging the paper as being about mSpace interaction or mSpace and its use of Semantic Technologies. The user could choose both, or create a new tag; if a user continues to type a keyword not found from the central server then a new one is created, but its associations are not defined. As the user finishes, the bottom layer of Figure 1 shows that the tag use is recorded by the Semantic Tagging Service using trackback (another Web2.0 technology). This allows the document to be tagged with just one tag - "mSpace" - but the document can now be found through any search related to the theme of interaction, because of the defined associations. Appendix C considers some of the potential interfaces for carrying out the actions listed here and in the remainder of this section. Searching for a document. Consider first, a user who wants to find all publications about a known project called mSpace. Current tagging systems will grab any document tagged with the word "mSpace". mSpace, however, has many research interests associated with it, and if a user wants to read more on the interaction aspects of the mSpace project, the user has to first discover this aspect and then hope that all the appropriate documents are tagged with the keyword "interaction". In Semantic Tagging, however, the tag "mSpace" may have multiple uses, based upon the associations applied to the tag. One "mSpace" tag may have associations about interaction and another tag "mSpace" may have associations about the Semantic Web. The user can now see these options at the time of query and select either or both of them to provide context to their search. Similarly, a user may want to see all publications relating to "HCI", and although not tagged with the exact keyword, all papers concerned with mSpace interaction would be included in the search results.
Cross-Repository Exploration. The first two user actions in Figure 2 show the action of finding Semantic Tags to search by, followed by the selection of tags used to constrain search results. Search can be done at various levels and the third user action allows users to select the repositories they wish to search over. Tagged items can easily be found from multiple publication repositories through the central taxonomy of semantic tags being used. Even greater is the opportunity to use Semantic Tags for services like Flickr, so that larger scale eResearch can be carried out; pictures of the mSpace interface and reviews could also appear in searches about "interaction". Creating a Tag. If a user is presented with two tag options for mSpace, interaction and technological implementation, but is actually concerned with the mSpace team as a group of researchers, a new semantic tag can be created. Potential tags for creating associations can be found in a similar way as when tagging a document, but are then simply included to represent the tag that is being constructed. Figure 4 in Appendix C shows this creation, where the user has already clicked on the "new tag..." option seen in Figure 5, Appendix C. Automatic Tagging. Some tagging systems, like Connotea, recommend tags based upon the use of tags by others; this states that documents tagged with "mSpace" are often also tagged with "Semantic Web" and "RDF". Semantic Tagging has greater opportunity for providing appropriate recommendations, using more than just similar tags used by other users. As tags are related, graph-matching techniques can also highlight tags that may have rarely been used together, but have similar associations in the taxonomy.