Allow me to begin this with a disclaimer: I fell in love with Yamli the first time I saw it back in late ’07. Yamli is such a simple but awesome technology. What makes it interesting today is its data collection process more than the core function of enabling non-Arabic keyboards to type Arabic words.
I can’t help but draw parallels between Yamli today and the yesteryear of Arabeyes. A couple months ago, I was having coffee with Nadim Shaikli (former Core member of Arabeyes) and we were reminiscing the good ol’ days when we tirelessly worked on enabling Arabic on Linux. We were also sad to see that the project today is limping without much in terms of leadership.
The truth of the matter is, the goals of the project were realized. Today, you can use Arabic on Linux out of the box. I would like to think that our work had something to do with that.. from translation projects, to filing bug reports with FireFox to submitting patches to popular applications like VIM. Oh the labor of love!
This brings me back to Yamli today. Habib Haddad (@habibh) saw a problem and he sought out to fix it. I think the fix is a patch to the symptoms rather than a permanent solution. While I understand that commercially, solving the actual problem is not feasible, I predict that Yamli’s usefulness today will slowly disappear as Arabic support on the keyboards become the norm.
I will concede that there will always be people who would rather use Yamli. I just don’t think it will continue to have the critical mass it is starting to get today. I have little reason to use Yamli on my mobile device when I can switch the virtual keyboard to Arabic. That removes the problem and makes Yamli rather useless.
I predict that unless Habib has a strategy to ensure that Yamli remains relevant, it will not last for another 5 years.
Please feel free to chime in!
Thanks for the post and thoughts. You raise a very good point, even though Arabic keyboards are becoming the norm, it’s about the users being familiar with them so I do strongly believe the need for transliteration is here to stay and kind of shows in our usage growth. We have been having a record usage traffic every week.However we are not planning on keeping our focus on Transliteration, in fact we have started pivoting away a while back into a data play where we work with enterprises, banks etc … and help them clean/normalize/navigate through their Arabic data, and that’s thanks to our technologies. (http://bit.ly/bf61SQ). In general I’d like to give an advice to all entrepreneurs to always take a step back from time to time and to be willing and ready to pivot. PayPal was a cryptography company before they became an online payment method, Flickr was a gaming company before they became an image sharing, and Yamli started with transliteration (and will keep that) and pivoting to also serve the enterprise data and search space which we think has a lot of potential growth in the region.
Thanks for your comment. I would be interested to know more about the underlying technology. A competent developer can probably whip out a transliteration tool similar to that on the website within a couple days, using basic heuristics, a database of words and a bit of Arabic morphology rules. There might be more to it, cause I’m surprised no one has attempted to replicate yamli (from individuals that is). Of course putting together the extensions, apps on the iphone, etc. is a whole other thing and makes the difference.I also don’t quite understand the enterprise use here. Like I had discussed with @tomgara, I don’t quite understand the whole disambiguation issue. I feel that I am the only one who doesn’t get it, so would appreciate the elaboration on that. Just because a certain word in Arabic can have multiple iterations when transliterated doesn’t mean anything, unless you are suggesting that the enterprise ONLY use technology like Yamli for data entry/retrieval.. and that’s a huge thing to ask.. though would be very interesting.
Sure my pleasure. Regarding the complexity of the algorithm, yes it’s pretty easy to whip out a transliteration tool, however the challenge was to make it work in a heuristic way that adapts and undertands all patterns and dialects … Orascom took a stab at it with Onkosh but was far in terms of accuracy and in fact they had to shut the company down after few years of operation … In terms of enterprise use case, there is a huge space in data normalization and deduplication. Take people, place names. Companies with multiple databases want to be able to reach match normalize their data accross, in AML (Anti Money Laundry) banks have to check a customer’s name against multiple black list sources … This is not by any means a new space, IBM and other major companies like Informatica have been working in this space for a while but no one has mastered the intricacies of Arabic simply because you would need to hire people from different dialects to sit down and fill those entries manually. With our transliteration engine and our collected data we can accurately and very easily do that … You should look at Informatica, IBM, AddressDoctor and see the kind of work they do in that space, we would be the Arabic expert
And now I have installed Yamli for chrome. Just because of this conversation.
I just found this interesting thread a year later
Mo Elzubeir, you are right, it is easy to whip up something like yamli over a few weekends, in fact I did just that -although a bit simpler- in the summer of 2005 which later became yoolki.com , but that’s not the hard part of the business.
I think, as habib said, the strength of yamli is not just the transliteration but the whole set of services they can provide since they now have a huge data set to solve actual problems companies are having.
Kudos to the guys at yamli and hopefully they’ll still be here for much longer than 5 years.