Information from Yandex.Subbotnik

I’ll try to tell without any emotions what I learned at the Yandex seminar on April 18.

First, conclusions, and then more about each report.

Conclusions:

I did not regret that I went to the seminar. Asked interesting questions and got answers. Because of the reports, it’s not worth the ride, but it’s worth the ride to be able to ask important questions during breaks. People were all adequate.

There will be no photos. In the first shot in the room, the flash planted batteries, and I forgot the spare ones. Murphy's Law in action. ;-) A

horse is here for a reason, I write about it at the end of the article.

Further: a description and conclusions for each report, wishes to the organizers.

1. Search the site using Yandex

Bogdan Garkushin An

introductory report for those who don’t even know what Yandex means of searching the site. The report is, in principle, not bad, but out of place. The vast majority of those who came were webmasters, experienced webmasters, and the like qualified people who know very well how Yandex can search the site. Therefore they have come. This was shown by the questions: all the questions asked were specific for Yandex.XML, Yandex.Server not mentioned in the report.

I learned from the report:

• That now Yandex estimates the Internet at 60-70 Tb. Three years ago, Google said they had 20 TB. Good agreement, exponential growth.
• Yandex.Server, how poorly searched in English, will not improve in the foreseeable future.

I was driving because of this report. The report turned out to be just a loss of 45 minutes, but I asked interesting questions during the break and received answers, thanks Bogdan.

Information: 1
Usefulness: 2
(I would get fives at some review and advertising conference)

2. Spell Checker Service

Alexey Baitin, Alexey Osipov

Presented a beta version of spellchecker. In every possible way praised her and scolded competitors in the best traditions of the bourgeois presentations. (Everything is just like in people ;-) From advertising, as a by-product, it was possible to extract information about the main functions and characteristics of the service.

I found out that:

• There is a check of Russian and English at the same time.
• A large dictionary, Yandex bought / licensed the Orfo dictionary (which Microsoft uses in Word) and develops it, although so far only slightly.
• Now (while beta) there is a limit of the volume of the checked text in 1500 characters (with variations).

We showed a prototype on TinyMCE (such an editor embedded in the page on Java Script).
I’m thinking about whether to apply this spell check and TinyMCE in my projects.

Information: 3 (although the report is fairly extended)
Usefulness: 4 (the most useful report)

3. Smart hints when searching a site

Artyom Vorozhtsov

Announcement of Yandex technology for downloading and displaying search results at the same time as the user types the query; how search query options are being loaded.

The technology is not just raw, but very raw; in my opinion: pre-alpha. Everyone understood that.
I did not introduce the parameter “presentation quality” in the report evaluations, but this was the most disastrous report. The speaker was pretty worried (the only one of all), although during the break he spoke quite freely with the people. It's okay, experience gained. But the speaker’s excitement was not the worst. The technology itself was not enthusiastic. The most drowsy report. The people woke up a bit in the pictures and the prototype, asked out of politeness. But besides, the report was full of rudesemantic errors. (Example: a proposal to remove the “Find” button.) I

practically received no information (I remembered the announcement), only a few ideas arose about Java Script.

Information: 2
Usefulness: 1

4. Algorithm for searching phrases in documents

Dmitry Agafonov

The only non-advertising and most technical report. There is a task: to select keywords (and, accordingly, ads) for Yandex.Direct, Yandex.Market for a web page. Dmitry told the general essence of the main algorithm (sparse search interval). Not just told, but rather well, in pictures, chewed for a long time. Only I got the impression that five people understood the report from the whole hall.

I found out that:

• In the direct + market 3-4 million phrases.
• 120 million web pages where they are installed.
• The restriction on the operation of the algorithm is 20 ms.
• (well, the algorithm itself, of course)

It was the most interesting and at the same time absolutely useless report. The above algorithm may be of interest to programmers or students, and it is hardly possible to use it only if a similar problem arises. If you distinguish ++ k from k ++, then download the video when they post it, an entertaining report, the most worthwhile, to download and watch.

Information: 5
Usefulness: 1 (most useless)

5. How Blog Search Works

Anton Volnukhin, Roman Ivanov

Advertising and educational report about blog search. Pretty high quality and most consistent with the audience. The program read “how blog search works ,” but it just didn’t talk about the internal search device, the report is actually about how blog search works : at what intervals it indexes, who gets into the top, what blog services have some agreements.

I found out that:

• Yandex has an agreement with LJ (and others): as soon as you publish something, LJ notifies Yandex, and a blog search indexes a new publication.
• If a publication has been made publicly available once, it will remain in the blog search no matter how much you make it private. (If you really need to make it private, then write in support of searching for blogs)
• There are 1 billion publications in blogs, this is 1/5 of the entire Runet, but the publications themselves are very small.
• From one blog, only one entry can be in the top.
• Top is updated every 20 minutes.
• Publications not older than 8 hours have a double rating.
• As soon as the publication reaches the main page, it begins to lose (spend) a rating every hour.
• Habra comments blog search indexes, but the articles themselves do not.

Information: 3
(there seemed to be a lot of information, but in reality it is all kind of small and insignificant)
Usefulness: 2

Wishes to the organizers

1. I would very much like to divide the seminar into a couple of more narrowly targeted (targeting):
   • For webmasters. Was this seminar supposed for a webmaster? Well, here it is. Would make the first report specifically for the webmaster. What is the best way to apply (best practices), scaling, non-trivial examples (which were just asked at the break). And the rest of the reports are the same.
   • Technical. Analogue of Google tech talks. The fourth report is wonderful, suitable (expand and deepen a little). Well, and look at Google tech talks in my opinion they have a good format. 1-2 major performances, not a whole day.
   • For the decision maker. (Technical Director, Director, Department Head, etc.) From this point of view, I am interested in what problems the technology solves, when and under what conditions I can use it. I consider the IBM seminar for directors to be an ideal example. 15-20 minutes reports, without water, everything about the case (but with a share of advertising, of course), 3 hours.
2. All bourgeois presentations have a mandatory slide about plans. So in the next quarter we will do it, six months later, this year. (Roadmap) You will laugh, but you need him. To make a decision on technology, you need to know about when and what will happen.

Off-topic

1. I was pleasantly surprised by the small delay in the start of the seminar, only 4 minutes.
   

   Report No.	Start Delay (min)
   1	4
   2	4
   3	5
   4	16
   5	6

   But there were delays everywhere, and this is nothing good.
   
   
   
2. Maybe what a particularly dark cloud I found in the sun, but when I was driving that way, that way back, I did not see a pointer to Yandex. The whole area is built up with old-type red brick buildings. What to choose is unknown. The arrows on the gate are visible only from a couple of meters.
   If I did not know that horses should be at the entrance, I would be late. Arrived 5 minutes before the start.
   
   (Legends say that the Yandex director bought two horses at a charity auction and since then they have been guarding the gate)
   
   How do you feel about the idea of ​​placing Yandex in huge letters on the roof in Yandex letters? Really simplifies navigation, well, and advertising.
   (And Rambler has huge blue letters)
   
   
   
3. When a malfunction occurred before the fourth report, hosted by Timur, as well as Anton Volnukhin and Roman Ivanov, had to entertain the audience for a quarter of an hour from the stage. Another employee joined them, muffled exclamations were heard in the hall: “Bobuk. Bobuk! ”
   I did not know what a beech was, but I thought it was some kind of famous blogger. And so it turned out. With audio bias. Bobuk = pseudonym of Grigory Bakunov. In communication after the reports left an impression of an adequate person.
   
   I found out who Bobuk is.
   

Conclusion

Thanks to the guys for the reports and to the organizers.