Yesterday I was at the launch of Arte.it, the first semantic search engine dedicated to a very important theme (to me at least): Italian art. The product, launched by Rome-based internet editor Nexta, was presented in a morning’s conference session that they sponsored in the context of Florens 2010 (my review of that coming soon).
Arte.it is a powerful vertical search engine that resolves the problem of returning irrelevant search results when you’re looking for something like “Leonardo da Vinci”, which might be an airport or be referred to in a number of non-art-related documents. It is also much more than just search. Nexta has realized that the future of web content is to harness existing good material, not just to produce new content. While their other vertical channels (turismo.it, stile.it) contain custom-written articles related to each topic within the blog format most favoured by emotion-based marketing companies these days, arte.it is the next step.
When you search for an artist or art-historical related content, you often turn up either unrelated material or plain crap. Arte.it gets around this by using semantic search, which means that, based on a geometric model (Latent Semantic Analysis or LSA) it looks at relations between words and concepts inside a document; these are analyzed as a vector and given a score that determines the document’s rank in the search engine. Technically this is great; in practise it needs refining by human editors. The Italian language version is already more refined than the English side of the website, and the folks at Nexta are apparently working on it – it’s still in Beta. Furthermore, arte.it depends on crowdsourcing so we can all make it better.
The problem of low quality search results is one that we art history bloggers have been discussing, and H.Niyazi has recently built a search engine that enables us to search for articles in a list of approved blogs that he is gathering: the art history database. While this list is limited, it’s growing, and there’s the benefit of knowing pretty much where your answers are coming from.
While AHDB and we “#arthistory” bloggers are interested in privileging content from each others’ quality blogs, Arte.it CEO Piero Muscarà repeatedly emphasized in his eloquent presentation that their search engine recognizes and prioritizes “reliable editorial content”. This means that wikipedia is not the first result for everything as it is on google, and that’s admirable. On the other hand, arte.it seems to ignore blogs entirely in favour of certain much more staid editorial products (especially on the Italian side). It also does not seem to include material for which you need a subscription – none of my test searches pulled up answers from JSTOR. While it can be frustrating to click through to material that you then have to pay for, any serious research in the humanities online should include results from journals.
Arte.it features pages about artists, semantic diagrams, a map that locates works by an artist in Italy, and even itineraries.
- The Artist summaries (they call it artist info – schede in Italian) are manually created by the editors and include a blurb of text that is pulled from another external website that is then linked (“read all”) at the bottom of a few lines. This is without doubt a good driver of traffic to the chosen website! Users can also create new artist pages after first making sure they’re not duplicating work already done.
- On the artist pages there’s a map that shows the location of that artist’s works in Italy.
For all search terms there’s a semantic diagram that shows suggested related terms. (Google has a further step: image swirl, here tested with the word “titian”).
- Registered users can rate articles or notify admins of “wrong” (they mean irrelevant) links, thus helping generate more relevant results.
- Perhaps the coolest feature is the way that arte.it connects art to place through itineraries – this is relevant for anyone interested in traveling to see art (my favourite topic, obviously). These are a little hard to find – in fact I haven’t found one yet on the site, but it was demonstrated at the conference and I’m sure they’ll show up more prominently soon. These are suggested itineraries based on the location, in one city, of various works of art related to the one you’ve searched.
- Another nice way the website connects to reality is through the changing images on the home page. These feature famous works of Italian art; if you click the + symbol on the bottom bar, a window comes up telling you where that work is located, with opening hours of the institution and its website.
Opportunities for bloggers
It seems to me that the folks at arte.it are open to suggestions – and their platform most certainly is too. Being crowdsourced means we can help them – and their users – find our writing. Material produced in English seems especially lacking and getting in there early on would probably give us a competitive edge. I would be careful not to spam this search engine but, as a registered user, only suggest good and specific articles related to a topic. For example, if you’ve written about a specific work by Giorgione, insert your article under the terms Giorgione + name of work. If you have an article about Matisse, insert it under that artist’s name. You can also suggest an entire website to be spidered.
Room for improvement
A little help is needed with the English language section – the term “notify web page” should be changed to “suggest web page”. The “news” section on the English side is still in Italian, and various pages are missing or not yet translated. As mentioned, it’s in Beta testing so they’ll be finishing and fixing these things soon.
One bug I noticed is that the English side rigorously takes English words in the search, even when an Italian word (like the name of a place) would turn up English language results. For example, a search of “Ospedale degli Innocenti” produced zero results; I had to search “Hospital of the Innocents” – which is not the building’s common name – to find information about Brunelleschi’s loggia and the orphanage written in English. And none of that information was truly relevant.
Second bug: the artist’s names are too rigorously defined, so that “Brunelleschi” does not bring up the “scheda artista” (artist info or summary); I did get the scheda by searching “Filippo Brunelleschi”, but I challenge the average university student, or member of the general public, to know the architect’s first name.
Third, using the English search engine, a lot of the results are from hotel or ticket reseller websites which I do not consider good or reliable sources (I’d go as far as to call these spam results). If I were the programmer I’d exclude any url with the word “hotel” in it, as well as the urls related to posters and horrible oil reproductions.
One feature I’d like to see is a bit more social integration. While you can share pages on facebook, user registration to this site does not use facebook connect or any other universal login system. I don’t have percentages for this but I am quite sure that users are more likely to register if they don’t have to create yet another password – especially as the login on this site doesn’t have a “remember me” check-box and cookie (my browser, firefox, remembers me, but when I leave the site I have to log in each time). As a registered user, there’s really nothing I can do to improve my public profile. You can insert a photo, but right now that feature doesn’t work. I can’t write who I am, what education I have, or provide links to my websites. Registered users are likely to be “influencers” (bloggers or other people serious about content) who like to be recognized. This recognition ought to be tangible: using a system of points and levels, users would feel rewarded and more inclined to participate by rating links, an action that actually causes a page reload so takes time and thus will only be done by people who feel strongly about participating.
Finally, while arte.it can be used to find general articles in the field of art history, it seems to prioritize biography. This is certainly the case in the structure of the pages headed up by an artist’s photo and bio, followed by links to biographies, and then to other websites. This seems to be a solution born out of Italian art history which still focuses on monographical production and connoisseurship type study, while Anglo-Saxon art history has, since the 1970s, been interested in social art history, feminist, pyscho-anlaytical and other branches of the discipline. I may be wrong, but it seems that there is room given here to biographies, works of art, current news, and sources, but little space given to identifying good resources for the analysis of artists and their works. Analysis can be found in academic journals available online and in the ongoing discussion on some art historians’ blogs and websites.
This mild criticism is all written in the true spirit of collaboration for I hope the folks at arte.it will take these points into consideration and that all together we can make this an even better search engine. They’ve identified a problem on the web and developed an innovative solution that is clearly expandable to other verticals that might have earning potential; in the meantime their product (with no apparent gain through advertising, just an important corporate sponsor) is a gift to us all.