--------------------------------------------- The existing w3c semantic web standards are well thought out, and provide coverage for well beyond the 80% common use cases. Improvements needed in rules (RIF: better tools and advocacy, and RIF/SPARQL integration), sparql (recursive queries), and declarative graph transformation language (xslt for graphs). --------------------------------------------- as for tools, i think rather friendly UIs and data cleansing and publishing tools are missing than rdfs and owl reasoners or sparql engines. --------------------------------------------- Semantic Web principles (global identifiers, linked data, ontologies) are unavoidable in any large information system. For me, the only question is whether these principles are used under the heading of Semantic Web or under any other heading. Having one heading may improve convergence of different technologies. Current tools rarely implement Semantic Web standards to 100%. I would see an advantage in throwing little used features (RDF containers, collections,....) out of the standard, adding ones that make developers use other technologies (e.g. property graphs also exploited in Wikidata), and pursuing 'hijacking strategies' (JSON-LD, hijacking on JSON, but adding a core Semantic Web principle, is an excellent one). Steffen Staab --------------------------------------------- RDF/triple store is too poor a data model for (eg engineering) data. Models built with RDF in this context, like ISO 15926, get idiosyncratic and ugly very quickly. Problem goes back to what is the semweb intended for. If it is for modeling really simple data types then OK. But anything of a scientific or engineering bent, forget it. This needs to be made clear up front. --------------------------------------------- The SW needs more advertising, better education in logic and databases and should rather make use of machine learning instead of complaining it would make the SW obsolete. For me personally, the Semantic Web is much more than standards alone and needs to combine with other technology and approaches. But maybe, we also need new standards (for OBDA, etc). --------------------------------------------- The SW community should eat its own dog food when it comes to finding a detail in the W3C documentation. --------------------------------------------- After years and years of discussing these issues (since Tim Berners-Lee started discussing Semantic Web) I personally came to the conclusion that if these visions made sense then we would have seen them implemented already. The reason we don't is that they must be lacking in some sector; meaning that there's some major weakness that outweighs the strength so much that Facebook/Google/Amazon will win every time we repeat the story. Perhaps we should be discussing these obstacles more if we want to understand the problem better. And I really like this questionnaire because it finally tackles them. I am not sure what the most critical problems are. They may be technological: users want sub-zero requests times. They may be economical: if the goal is to get richer, then big centralised solutions will always try their hardest to build these fences. Or it may simply be an innate trait of our species: we always seek for leaders, hierarchies, control, people with keys to blame and ask for solutions. Anyway I am glad you are keeping the discussion alive and am watching closely with popcorn in my hands how the situation evolves. But we may have reached a point where we are trying too hard to make it work. --------------------------------------------- Thanks. Nice survey, even if I would say some of it missed its mark (of course) :-) Also, I suspect that if you were able to capture it, you would find there are serious differences of what they think the Semantic Web *is* between your respondents. One thing that the survey does rather ignore is the social aspect, which I think is possibly the most important (as is so often the case). It arguable that the biggest negative driver to the Semantic Web is the people who now own and control the technology definitions and the policies etc.. And that may also be the biggest positive driver. --------------------------------------------- It would have been good to allow for don't know option as well. --------------------------------------------- Common sense in a very general meaning is needed. People are so easily divided by trivialities, e.g. nationalism. --------------------------------------------- The concentration of the (semantic) web is not due to decentralisation being too costly. It is because centralisation attracts capital investment because the pathway to return is more obvious. Building the centralised infrastructure that we have today was enormously costly. Perhaps even more so than a decentralised one would have been. However it remains that investment has largely been in centralisation and that does not appear likely to change in the near term. This is not unique to the semantic web. It is reflected more generally, and the phenomenon that it is reflecting is the massive underlying inequalities in society. (1) OWL was designed by logicians, and is difficult for non-logicians to use; although it has been taken up in some specialised areas, e.g. bioinformatics. (2) Data provenance and data accuracy will always be a problem. (3) It seems likely that decentralised systems will always be brittle. The success of the Web is in part due to the acceptance of Error 404; the semantic web may be less tolerant of errors. --------------------------------------------- Folks, eat your own dogfood. Use it or lose it. #ControlYourself --------------------------------------------- Actually a fair share of semantic web functionality has gone mainstream; consider schema.org / rich snippets, the RDF support in mainstream databases that were originally relation databases, etc. It's rather just that the terms "semantic web" and "ontology" sound too academic and therefore are not being used to characterise such mainstream deployments, which may have the effect that people don't realise that there is actually semantics inside. --------------------------------------------- It is an incentive issue, not a technical one. The motivating problem described by Tim Berners-Lee is mainly solved today by single ontology centralized applications. For instance, several apps offer the service of finding a hotel, where host owners fill the attributes asked by each of these apps. The result is, thus, several separate single ontology apps (e.g., Airbnb, Booking, etc.) for finding hotels where also hosters submit their data, instead of several apps taking its data from the Web. The main problem is economic incentives. In the current model, the three actors (apps, hosters, and people finding hotels) have their selfish motivations in the game. On the contrary, in the Semantic Web approach, we require other motives: 1. Apps prefer to own the data to compete with other applications. Thus, the only possibility to use data from the Web is that such data is already available and have enough good quality to be used. 2. To publish the data to be available for apps, we require better publication tools. They have to imply lower costs than publishing the data in current applications. This may require simple ontologies for each case, and good tools to integrate all produced data. 3. Also, better publication tools are not enough. We also require the already published data to be so valuable to be ignored by apps. We face thus an egg-chicken problem. 4. To publish data, we need a fourth actor - the one that provides friendly tools to publish data in the Web to build the online dataset required to solve the egg-chicken problem. The main complication is that the incentive for this fourth actor seems to be unselfish. I believe that the main problem is then the motivation of the fourth actor. If we provide a sustainable motivation for it, then we will solve the egg-chicken problem. However, solving the fourth actor problem is not the end of the story. One app will eventually rule other apps. Then, the winner app will have to much power to build its closed publishing tool for receiving data from hosts, generating a closed dataset that will compete with the open one. It may eventually kill our efforts to build an open web of data. --------------------------------------------- I'm missing questions concerning the state of shared vocabularies, non-simplistic information modelling and agility of standardization. --------------------------------------------- Reading the questionnaire there's a lot of emphasis on what I would call the semantic web technology stack. But the semantic web is an idea and many technologies can be used to implement it. I think the semweb tech stack is useful but also has inspired other technology stacks. As a community we should embrace the diversity and movement of technology and ideas that embrace the web and the need for semantics. Technology is not fixed. The merger of the idea and the tech stack I think is unfortunate. --------------------------------------------- If you doing some good result, please make it public, its not about getting a publication, because after 3-4 we don't get any information about the work except 8-10 page pdf. --------------------------------------------- Great questionnaire! --------------------------------------------- It would have been nice if you had included questions about the practicallity, learnability, and implementation about syntaxes like RDFa, microdata and JSON-LD as well as the concept of semantic modelling. --------------------------------------------- Experience from working with semantic web applications for more than a decade leads me to believe that the initial proposition that data first, applications second was a mistake. We focussed on publishing data rather than using data, use cases should have been developed from practical, pragmatic needs. Practical, scaleable linked-data applications are entirely possible, but the knowledge needed to develop these does not seem to be present in the majority of practitioners. Assumptions about efficiency and performance have been aimed at poor performance of chosen technologies rather than the how these have been used. The practical aspect of using semantic web technologies is still emergent. The use of semantic technology as a panacea was also heavily present early on — like every other technology, sometimes, semantic technology is a good fit, sometimes not. Some practitioners still have a tendency to push a given technology, e.g. reasoning, as a good fit for everything, while other approaches would be a better fit. This technological fundamentalism has not helped us gain traction and has actively contributed to the semantic web being deemed a failure as it leaves behind it a trail of failed projects, and poor performance and difficult development in those projects that do reach completion. --------------------------------------------- I have interpreted the term ""semantic web technologies"" very broadly - it may be that future SW technologies used are not immediately recognizable as one of the core standards mentioned in this questionnaire (e.g. JSON-LD). The vision could be carried forward by new standards that are not directly compatible with today's RDF; e.g. I can conceive a future in which some development of property graphs using URIs becomes part of the SW family. One of the criticisms not mentioned in the questionnaire is the lack of higher-level abstractions for creating and working with SW data; e.g. expressive capabilities provided by property graphs that are hard(er) to construct using RDF. --------------------------------------------- Many of the critiques on the Semantic Web are based on costs vs. benefits. “Today it’s not worth the effort” or “people will never do this because they don’t see a direct return”. While I believe this is generally true today, in a couple of years we will need to solve computer problems that can only be solved with technology such as RDF. While AI and machine learning might help with understanding unstructured data, and thus automatically mapping it to structured data models, I don’t think machine learning and late reconciliation will be the strongest enabler for the Semantic Web. I still strongly believe that the big revolution will be in better data maintenance, something we have been investing time and effort in for quite a few decades now. This slow world-wide better maintenance of key datasets is the slow revolution we are waiting for. Pieter Colpaert --------------------------------------------- That the semantic web allows different semantic models to coexist is a major strength. AI can enable processing using multiple semantic models. This is the direction in which the semantic web should evolve. --------------------------------------------- For the the SW has mostly been about structure/meaning, the current standards may not necessarily be the best option to achieve/support the SW. They should be considered a starting point/learning curve to produce something more robust and leaner. Some SN and publishing platforms will adopt good practices, some won''t for different reasons. The web will evolve and so will its users, but good practices should be preserved, even if they are not popular. The world is complex, so are its artefacts. I am very happy with the web even if bits of it are made to fall to pieces. Lets try to keep the vision for more structured content and better search outcomes, using a variety of approaches. --------------------------------------------- The problems with tools are not so much about supporting RDF, OWL or SPARQL, but supporting features that we do not even have yet, such as easy standard ways to express and apply custom inference rules, and easy standard ways to select, transform and align RDF data for integration with other data. --------------------------------------------- Mediocre software engineers and collateral personnel fear the Semantic Web because they deal with mediocre problems. Why would one invest in rigorous ontology design and development, when one's domain of discourse is limited to a few entity types? Why would one care about standards compliant, machine readable data interchange when one's oblivious to their audience? Why would one be passionate about globally unique identifiers in a stable namespace when they're deluded into thinking they're building short lived products? From several years' recent experience building the UK Parliament's latest iteration of open linked data platform, knowledge graph and semantic publishing pipeline, I conclude that products, services, tools and technology for Semantic Web development are mature, cost effective and scalable. Organisations however have too short a memory to persist in solving their own complex problems. The understanding that the world is complex enough to require complex software escapes most people. They'll get it eventually. --------------------------------------------- The need for data interchange is only increasing, also for previously not related communities. Standards for this are necessary. One important driver in my opinion is research data management, involving heterogenous communities, and the FAIR principles, which translate complex ideas (for example the need for globally persistent identifiers for data and metadata) into comprehensible formula. --------------------------------------------- I think one of the barriers to adoption is that the general public does not consume RDF data. I can try to put information into a webpage, but overwhelmingly, Web browsers don't make use of it. Web browsers are increasingly moving towards models where server operators have to write entire applications that integrate with specific other applications, instead of just seeing data and automatically being able to do things with it (for example, offering to import a date into your calendar). --------------------------------------------- It's very sad, but I don't think it very likely we will get any more webby in the future. The web is now a single browser engine and walled gardens everywhere. ---------------------------------------------