Just put the new database live on Able Grape. It’s got close to 1000 new sites, and is some 10% larger than its predecessor. It’s also much fresher. Along the way, I managed to sneak in a few improvements.
For a closer look at what’s changed, plus some fun examples, read on!
Relevance should in general be improved. There’s lots more I plan to do here, but based on some initial feedback, I made some adjustments.
I’ve extended Able Grape’s concept database a fair bit. The concept database helps Able Grape learn which things are related to which, which things are specific cases of which, which things have nothing to do with each other, and which things are synonymous. There were hundreds of additions and extensions in this version of the database. Here’s a smattering of examples:
- It’s got a deeper understanding of the various ways people refer to mousy taint or cork taint.
- It now knows that there are a bunch of different kinds of phomopsis. The interesting one, Phomopsis viticola, is the fungus that causes excoriose. Others aren’t very relevant to viticulture, like phomopsis of junipers, phomopsis juniperovora, or phomopsis soft rot of strawberries, phomopsis obscurans. It’s not perfect, but it does a much better job of weeding out (no pun intended) these extraneous critters.
- It’s learned that a Stelvin is a specific kind of screwcap, that synthetic corks are made of plastic, and that all of these are examples of closures, which can be useful in looking for things like trade press about closures.
- It’s learned that non-vintage champagne and vintage champagne are two different things. (This might seem quite obvious to us flesh-and-blood types, but to a piece of silicon looking to match the words in your query to the words in a document, the phrase “non-vintage” has the word “vintage” in it, so it’s fair game for the query “vintage.” The first result for the query “vintage champagne” on Google highlights this).
- It knows that Ch. Latour and Louis Latour are two different producers. If you type Latour you’ll get results for both. It should do a pretty good job of sorting out the gros Gros Family as well: Anne Gros, Gros F&S, Michel Gros, and AF Gros. And similarly, if you type de Pez (or Ch de Pez, or Château de Pez), you won’t get results for Les Ormes de Pez or Tour de Pez. And then you’ve got your Conternos, Scavinos, Mascarellos, etc. etc.
I should mention that the concept database, like everything else, is a work in progress. If you see poor results, drop me an email; there’s probably something I can do to make things better.
On a related note, I’ve tried to make looking for vintage information more interesting. Here are some examples:
I see a lot of queries coming in looking for things like maps and statistics, and I can see people struggling to find good results. The data’s in there, but it can be hard to get to. These are really hard to do well in any search engine, because you’re not really looking for the words maps or statistics, or acreage, you’re trying to tell the search engine what you’re looking for. But the silly search engine doesn’t know any better, it just goes off blithely looking for documents with those words. The problem is compounded by the myriad ways in which documents refer to statistics in any given language, and even further by the fact that the documents may well be in another language. I’ve got some ideas on how to make this better, and I’ve taken a small first step in this database, really just an experiment, and I’d like to hear what people think. Here are some examples:
- Beaune maps.
- Bordeaux map.
- Chardonnay statistics (filtered by official sites, which is often useful for this kind of thing). Should you wish to narrow this to a particular region, remember that you have two ways to accomplish that: either include the name of the region in your query, or use region filters. They’ll give different results, so both approaches may be worth a try.
There’s a lot more to do here, but it’s a start. Please let me know what you think!
Finally, queries for German users should be considerably improved. The new database accepts the ue/oe/ae spelling for ü/ö/ä (and also, for us anglophones not used to that, it continues to support the unaccented form u/o/a), so that people searching for JJ Prüm can spell that Prüm, Pruem, or Prum. The results will differ slightly because Prüm is accent-sensitive, but things should work much better than before. Comments and suggestions on improved accent-handling are most welcome.
As always, let me know if you have questions, suggestions, or ideas-