captsolo weblog

This is a followup to "Yahoo! opens [a Web of] Pipes".

From a comment to the Yahoo! Pipes article by Tim O'Reilly:

" It finally seems, Yahoo! secretly started building a Semantic Web powerhouse, by making clunky RDF chunks (re-)deployable for the average web developer without having to care about the groundwork… ;-) "

Yahoo! Pipes is a wonderful service which makes creating new mashups very easy, but its dependence on RSS, while part of its success, could also be limiting its potential. Please correct me if I am wrong, but the structure of data that flows through these pipes is RSS or something very similar.

This may create problem if you want to use data that are not RSS and are not easily representable in RSS. E.g., how would you describe in RSS who are your friends?

Pipeline architecture itself is not unique to Pipes. Applying it to the Semantic Web (or to RSS) is not unique either - many have been collecting / integrating information from a number of sources and asking questions like "show me the publications by people who know [user input here, e.g. "Stefan Decker"] and filter the answers according to some criteria" and passing the results to other applications in JSON or RDF/XML. That makes a simple pipeline, even if it is built manually and without a nice UI.

What is unique, though, is how easy it is to create new mashups with Yahoo! Pipes and the potential they provide to make Semantic Web used by a large group of people.

All that needs to be done is to allow Pipes to feed on arbitrary RDF data sources. RDF can represent any information, including RSS, Atom, friends lists, publications and more, therefore no information will be lost.

Information can then be processed using generic, user-built queries, e.g., "For each item that is in a foaf:knows relation to me". All existing Pipes operators should work with some minor modifications and having a flexible content model will open new possibilities. SPARQL would then be a natural choice for representing queries (actual choice is up to developers / architects).

At the end of a pipeline will be output format generators capable of generating RSS (possibly loosing some data), JSON and RDF (preserving all the information). Using RDF does not mean inventing something new either - just use RSS 1.0 (which is also RDF) or SIOC to represent information about portions of content (posts).

Summary - take Yahoo! Pipes, let RDF flow into them and through its pipelines and you will create Semantic Pipes! They should be even more flexible and present interesting possiblities for reuse of information. Remember: data out there are not limited to just RSS or what can be represented in RSS.

There will be some technical questions to be solved, but that should be well worth it.

Related:

• See "[RT] Rethinking the need for inferencing" where Steffano Mazzocchi discusses applying pipeline architecture to RDF. I wonder how far has this effort progressed since then.
• Danny Ayers goes into a detailed discussion re. Semantic Pipes in "The pipes are calling". He also suggests that RSS can be limiting if used as an internal data format and asks the question: "What's the data model?". I hope someone out there can provide an answer to this.

Technorati Tags: pipes | semantic web | sioc | yahoo