#Duplicate Sentences

I think it would be more intuitive to display the error message below the input. Just like this:
http://imgur.com/a/W6IoO

I recommend you to take a look at this page: https://material.google.com/com...ine-text-field

Seconded. Brilliant work on the part of Trang, and anyone else involved.

She rocks! :)

Greatly done, Trang, and many thanks to you. ☺

+1,000,000,000,000

Thank you all =)

The cause for the recent slowdown was not obvious, but it was not too difficult to identify. And it was very easy to solve. The tough part was to figure out where to focus the investigations.

Before I found out what was the cause, my main fear was that the slowdown was caused by the growing number of sentences in the corpus, and if that was the case, it would have been much more difficult to fix. But it turns out that even with 5+ million sentences, we're still fine.

The problem was related to the lists, and it was a problem that existed ever since we introduced the lists feature, years and years ago. It was just clearly noticeable before because we had less traffic on the website, and because we also had less sentences added to lists.

If nobody used the lists feature and we had just a couple thousand sentences added to lists, the website would have kept good response time. But we have over 1 million sentences added to lists today.

If you were visiting the website while only a few other people are browsing at the same time as you, you would also not notice any significant slowdown, because with fewer visitors the server can manage to process your request right away.

If however you visit the website during rush hour, that's pretty much the same problem as when you go to the supermarket during rush hour. You have to wait in line. The server cannot work fast enough, it cannot process your request right away and you have to wait for it to process the requests of everyone before you.

Anyway, the crisis is over, and it's a relief. But this episode is clearly a warning that we should invest more efforts in optimization, to keep the server healthy. This is not really my cup of tea but thankfully, we have a rising active developer, Zach, who joined us back in April this year, and who will be helping on this task :)

Nice!

This won't solve everyone's problem, but for those who want to browse English sentences with audio, you could temporarily try this.

http://study.aitech.ac.jp/browse/

These are 240,199 English sentences with audio by CK.

This is something I put online a couple of months ago.




There is also another option for browsing English sentences with audio, showing translations owned by native speakers.

http://www.manythings.org/sentences/audio/

These pages were generated in early 2016, so aren't up-to-date, but would still likely be useful for those who are looking for English sentences to translate into their own languages.

Depending on the time of day, this website may have a lot of traffic and be a little slow for you.

Wow. Just so I'm clear, there's no longer a way for me to find sentence examples of specific words (in my case, mandarin ones) that I'm learning? Or is there some work-around I'm not seeing?

What kind of use are we expected to do of the website if it is not possible to find sentences? It might sound like a rhetoric question, but it is not, I am really curious how are we supposed to interact with tatoeba now.

Honestly, I prefer a slow website. What's the point of a fast website if missing the main feature of all Tatoeba?

Since the google search site (http://tatoeba.ueuo.com/google.html) is so fast, a temporary fix could be having Tatoeba execute a search with the google search URL behind the scenes, then "screen-scrape" the returned page with awk, sed, regexp whatever to present the user with what he is normally used to seeing. Of course, we'd have to give up the ability to submit fancy search requests, but at least Tatoeba gets its search capability back to some extent.

The search is currently re-enabled.

I have possibly found the main source of the slowdown: https://github.com/Tatoeba/tatoeba2/issues/1324

If things go well, we can keep the search enabled while taking care of the rest of the optimization.

The "Browse by language" page will remain disabled for now because there is an issue to fix about it: https://github.com/Tatoeba/tatoeba2/issues/1323

** Search the 400,000+ Proofread English Sentences on List 907 **

http://aitech.ac.jp/~itesls/c/qs/907.html

I'm not sure if my web sever will support this if it gets a lot of use, but at least temporarily, I'll leave it online. So far, it seems to be working well.

Dankeeeee

Thanks, CK! That page is quite useful.

Note that if you wanted to search for, say, the word מתעטש in Hebrew, you could type this in:

allintitle: מתעטש "Hebrew example sentence"

That would show only English versions of the sentence page, such as:

Hebrew example sentence: אני מתעטש הרבה. - Tatoeba

and would eliminate search results with such titles as:

Phrase d'exemple en hébreu : אני מתעטש הרבה. — Tatoeba
ejemplo de oración en Hebreo: אני מתעטש הרבה.- Tatoeba

all of which would point to the same page.

In practice, you can leave out the "alltitle:" and your results will be mostly the same.

Wow, I didn't know to come in here before to learn of a temporary alternative to doing searches. The http://tatoeba.ueuo.com/google.html URL works pretty well. Maybe this info should be part of the error message on the tatoeba website when a search is attempted and fails.

I agree. Tatoeba becomes worse every day. This new interface, desired vocabulary set that doesn't work as expected and disabled search and random sentences block now.

But if you had experienced when tatoeba was really fast (when I first joined), you'd understand why the slowness issue has to be addressed. I used to use tatoeba.org as a goto page for testing if my home's wifi connection was up. It was pretty damn fast.