As online and mobile technology takes off, how will people search Myanmar-language content? The answer is uncertain: The country has one of the lowest internet penetration rates in the world and, so far, no very good search tools. That’s why Bindez, a local start-up, is working on it.
Nurtured through Project Hub’s 2013 fellowship for entrepreneurs, the project belongs to computer science grads Ko Htet Will, 23, and Ko Ye Wint Ko, 24, who co-founded Bindez with their fellowship mentor and now CEO, Rahul Batra, 29. With a recently confirmed seed investment and a uniquely qualified team, the project is one to watch as the country makes the leap from virtually no tech to the future.
“We’re trying to imagine what a search looks like if you block out the picture of having Google and Yahoo. What else could it be? Would it be just a blank page, or would it have a more intuitive approach?” Mr. Batra said, sitting at a table covered with laptops, wires and tins of snacks in Bindez’s current home office in Kyeemyindaing.
Google launched its Myanmar site in March 2013 and tech watchers expect it will invest more resources here, too. Bindez said it hopes to offer users something different from the search giant. “We want to redefine the way Myanmar people consume information to progress in their daily lives,” Ko Ye Wint Ko said.
Any research team faces a difficult challenge. “When you launch a search engine, you have to have crawled everything in Burmese on the net and it’s spread a lot of places … so you need a lot of time and resources spend culling the data,” said Ravi Chhabra, an adviser to the Myanmar Computer Federation. That’s partly why few have invested in similar projects so far, he said.
Building search tools also requires work on Myanmar natural language processing (NLP) – the field of computer science concerned with the interaction between human (natural) language and computers – which is in its infancy relative to English, experts said. A government-run research lab opened in 2006, only to be shut down around the time Ko Htet Will finished his university courses in spring 2013. That inspired him to open their own lab, and they quickly enlisted the help of two researchers from the now-defunct government project.
At about the same time, Bindez was accepted into the Project Hub fellowship, where the team was matched with a mentor, Mr. Batra, an ex-Google specialist from India who initially came to Myanmar to help an NGO work with rural youth on small-business development.
By the time the fellowship ended in November, Bindez had interested a few investors and the team had dedicated themselves to the project full time. Over the next months, the team debated what Myanmar users might want. Although the English-speaking world almost takes it for granted that people want a “pull” mechanism like Google, which responds to user queries for information, in other languages, like Russian and Vietnamese, people have demonstrated a preference for tools more like Yahoo!’s “push” style of navigable subject directories.
Ko Htet Will and Mr. Batra point to the success of localised search engines in other countries, in particular Vietnam, as models of what they’d like to achieve. CocCoc and Wada have both made successful inroads by offering directory-style “click” searches. Wada’s underlying technology is also able to recognise Vietnamese names quickly and easily compared to other products.
So far, no Myanmar search tool is very good at turning out relevant search results. Even Google’s country site is rough and doesn’t, for example, weed out search results with obscenities, Mr. Chhabra said. “It’s actually quite ugly right now.”
If an engine is going to turn out relevant results, researchers must improve the tool’s ability to recognise things like parts of speech, individual words and phrases and word meaning, which means also building resources like dictionaries and lexicons. Toward this end, so far the Bindez team has built a text corpus of thousands of Myanmar terms –research they intend to make available for open-source development of Myanmar NLP in the long term.
“It’s quite difficult because the language is complex, as is digitisation of the glyphs and encoding,” said U Thar Htet, a tech expert and founder of Zwenex.com web and software development. “Without that research … the larger population cannot use [computer] systems without a second language.”
Myanmar people currently input type in one of two encoding standards: Unicode, the international standard, and Zawgyi, a non-standard encoding. Currently, most of the Myanmar content on the internet currently exists in Zawgyi, making more work for the engineers of search algorithms, U Thar Htet said.
Bindez said their goal is to create tools that accommodate both standards and are also customised to support local language and knowledge.
“If we can really identify the content that people would like to hear and know, if we connect to our local people and provide it very much with our own ‘secret sauce,’” then Bindez could take off, Ko Htet Will said. They plan to release a private beta next month and a public beta in the near future.
Source: Myanmar Times