Hi,

well, at least HTML5 is a W3C recommendation https://www.w3.org/2014/10/html5-rec.html.en. HTML5 allows to describe also the structure marginally for differentiating title, content, aside, etc. However, converting HTML into RDF will be in most cases probably not possible (since it would require to grasp the semantics of the content).

The purpose of the idea of Berners-Lee's proposal of a Semantic Web was to make the content machine-readable (not to say machine-understandable). So, instead of having the WWW and the Semantic Web in parallel, which would have the consequence of a large maintenance overhead, it is -probably not only- from my point of view better to augment HTML pages with RDF(-like) markup.

Actually, RDFa https://www.w3.org/TR/rdfa-primer/ (also a proposal of W3C) was designed for the purpose of embedding RDF attributes into HTML, XHTML and XML. GRDDL https://www.w3.org/TR/grddl/ (also a W3C recommendation) was developed for this purpose, too. RDFa belongs together with Microdata (also designed for embedding machine-readable information into HTML) to so called Microformats.

Some years ago I have actually seen RDFa within HTML pages of some newspapers. Looking at such pages some weeks ago, I found that actually HTML5 already allows to embed annotations (with itemprop, itemscope and itemtype) and that some sites use JSON-LD in combination with vocabulary from schema.org to include machine-readable content in HTML. Since this content now becomes machine-readable, it can be transformed more easily to RDF.

Let's make the long story short. In some cases you can derive some RDF data from the Microformats embedded in HTML or from HTML5 directly. But a general transformation of arbitrary HTML content would require sophisticated text understanding procedures which can determine what the contents means.

However, if you can restrict the application domain, than the chances are high, that you can design some special purpose RDF extraction procedure for it. But, this requires to augment the extraction procedures already with domain knowledge, in order to detect the relevant parts for the RDF extraction.

So I would say "the current web is a semantic web, but only partially at some sites".

Hope this answers your question at least partially.

Well, since I am preparing a lecture about semantic web and tried to find examples of RDFa within HTML, I came across the following site: http://dws.informatik.uni-mannheim.de/en/news/singleview/detail/News/387-billion-quads-microdata-embedded-json-ld-rdfa-and-microformat-data-published/

Especially, the follwing statistics: http://webdatacommons.org/structureddata/2017-12/stats/stats.html is interesting. I think, combined with an estimation of the "size of the WWW", we could derive the "degree of semantification of the WWW". Could be interesting.

Hi,

Thomas already gave a complete answer in the perspective of Semantic Web Technologies.

The main challenge here is to transform unstructured data from plain HTML files to structured data ( e.g RDF or OWL ).

This transformation can be performed by Text Mining techniques, here are some examples of free APIs: https://www.predictiveanalyticstoday.com/top-free-software-for-text-analysis-text-mining-text-analytics/

Parsing HTML files from legacy websites is not very simple, but if you have a clear understanding of the application domain that you want to map, from unstructured-data to structure-data, it is feasible.

Cheers

Hi

There are some methods that extract facts from texts of the Web in RDF format. YAGO is one of them. This is a knowledge base with ontology structure that contains facts of Web. So this can be a method that convert html files (texts) into an ontology with the RDF format.

Also, I believe that current Web is not steel a semantic web with Lee difinition.

Is current web a semantic web

the direct answer is no and it will not be. the HTML is for human. To build "smart" applications we need to build the semantic web (XML-based). many big players are doing that and mapping the HTML files with under-laying semantic. the tools are available but there is no well-defined ontology that we can agree about the meaning and (if any) we can not  mapping  with all concepts in the HTML pages. In some domain (health ) this might be easy but in other domains it is not that easy ( social media).     

Hi,

I think that the Semantic Web is not just a matter of translation from HTML to RDF/OWL languages. It is a matter of understanding the content of the Web. This content is composed by resources (Web pages) that contain, for example, text, images and audio. Thus, there is the need of extracting entities from these resources, for example, "Donald Trump" and "the president of USA" and "Melania Trump". Then, these entities need to be related with relationships: co-refer("Donald Trump", "the president of USA") and married("Donald Trump", "Melania Trump").

Up to my knowledge, the only example of SW is the one proposed by Google on a right column shown beside the results of a query for an entity, such as Donald Trump.

@Ivan: There is more on this earth (or better said web) than just Google ;-)

Of course, Google makes visual use of its knowledge graph, but I wouldn't call it "semantic web", since it does not follow the principles of the semantic web, i.e. to use (defacto) standards and URIs. You can easily inspect the box you mentioned and what you see there appears to be proprietary. Thus, in order to make this content machine-processable you have to understand it and hack your own interpreter.

However, sites which make use of the semantic web, often do it under the hood without any visual hint for human users. Inspect for example the source code of the pages:

• http://www.springer.com/de/book/9783662554326 or https://www.rindchen.de/ and search for "schema.org" . You'll find microdata encoded in HTML5 itemprop, itemscope, itemtype, which allows to interpret the content in terms of the schema vocabulary (a defacto standard) and transform it to RDF
• https://www.monster.de/jobs/suche/?q=Data-Scientist&cy=de&rad=20&intcid=swoop_HeroSearch or https://www.flickr.com/search/?text=semantic%20web&license=2%2C3%2C4%2C5%2C6%2C9, which use schema with JSON-LD microdata
• http://clashmusic.com/ which uses the standards of the semantic web stack

Hence, the semantic web is actually imperceptible and that's probably the problem, why it has difficulties to be adopted more broadly: its use is not directly visible. Therefore, I found the statistics about the use of RDF and microdata, I mentioned above, very helpful.

Of course you are right, transforming HTML into RDF/microformats isn't the problem, the problem is - as you pointed it out - understanding the content. But, I am pretty sure that sites which use the semantic web under the hood today, do not interpret their content. With high probability they have integrated the semantic web markup already into their CMS or HTML generation algorithms.

I would be happy to receive any confirmation about my last assumption.

Maybe the SW is a matter of understanding the content (e.g., through knowledge graphs) and encoded it with standards. This combination could make SW more visible than the current one. I think that, in many cases, the interpretation of the data (along with the standards) is fundamental for the SW effectiveness.

If data in web arranged in any logical way, that data can be processed by machine. So there comes the Sementic web. What logic is used is matter of choice and so many logics or complex logic in Sementic web may confused its normal users if normal users acts as developers of the same web.