PalaceHotel Blog

The enabling IEML technology for the generation of semantic tag clouds is the USL. An USL is a collection of sets of IEML *tags at each of the 6 different layers. Each *tag implies a set of semantic relationships with other *tags; at every layer a USL thus expresses a combination of the semantics of many *tags. At the most basic level, therefore, USLs are already small tag clouds really.

Let us try to understand what meaning does the addition of multiple *tags at each layer bear for us.

What do multiple level 0 elements mean?

This is a trivial case, because level 0 elements, by definition, have no semantic relationships to other elements. In addition, level 0 elements in the context of a USL represent “networks of Collective Intelligence”, and thus, by tagging a resource with multiple such elements, a user exactly means that the resource participates in multiple such networks.

What do multiple level 1 elements mean?

Things get a little bit more interesting here. Every level 1 element can be either nounish or verbish. A verbish expression is going to be either en energy or an act. A nounish expression can be a mutation or an entity. There is also a obvious symmetry between acts (O:M:.) and mutations (M:O:.). Finally, the categories O:M:., M:O:., M:M:. and O:O:. all have etymologic relationships with both their first and second role players. Let’s see how we could visualize these relationships for just one IEML * tag of layer 1:

Now consider another semantic relationships graph:

Imagine that some resource was tagged with these two *tags (this one might be a good candidate), i.e. want and perceive; what can we already conclude by examining their semantic graphs? These are very short graphs, of course, but counting the number of arrows between concepts, the Verbs one is certainly the most common destination. It means that, from just two *tags, we can already conclude that this resource will add a mostly verbish coloration to the tag cloud to which it would participate.

A tag cloud is a list of words in different sizes and colors, with or without a sense of depth (3D), meant to represent the statistical importance of keywords mentioned in a particular document base (a blog, a website, twitter,…). It serves as an indicator of the relative importance of the use of certain ideas in the document base at hand. It is a bottom-up, very fuzzy method for the synthesis of knowledge from an arbitrarily big aggregate of (text) data. Because it rests entirely on statistics, very often there is absolutely no relationships between the keywords of a tag cloud. Worse even, if they existed (by pure chance), there is absolutely no way of finding out about the meaning of those relationships.

On the other hand, ontologies propose a top-down approach to the same problem. They require first that some expert, or committee of experts, agree on a common description of a particular domain. Given an ontology, it is possible to map a document base to the concepts it describes: every concept would act as a filter to retrieve documents that match it. The clear advantage of this method is that it is possible to use the semantics expressed in the ontology to search for a specific combination of concepts. A tag cloud rendition of an ontology would, however, carry two major drawbacks with it:

1. you need an ontology per domain you wish to search for, hence ontology-based navigation is always going to be extremely specific;
2. no semantic value added is generated (note that there is indeed a minor semantic value added in the statistical tag cloud, albeit of a purely quantitative nature)

An IEML tag cloud would combine the advantages of both approaches, while overcoming their limitations: it would generate significant semantic value added through the discovery of the underlying conceptual network of a given document base; and it would allow for navigating in semantics. As for the statistical tag cloud, the conceptual network thus discovered would remain quantitative in nature to a certain degree. Indeed the salience of the semantic relationships is going to vary depending on the precise contents of the document base, as is going to be necessary to reflect in the rendition of the resulting tag cloud.

Another pending issue as far as rendition of an IEML-based tag cloud goes is deciding upon the relative positioning of concepts on the user’s display, in 2D and/or 3D.