Even if the Philippines is regularly proclaimed as the social media capital of the world, what good is all that chatter for companies if they can’t understand what people are saying? Most social media tools, after all, can only decipher and analyze English. This leaves the greater majority of Filipino netizens who prefer to speak in the Tagalog vernacular crouched in the dark, away from company eyes.
Manila startup Senti wants to help Philippine companies reach those people through its proprietary platform, which promises to understand “keywords and terms in English, Filpino, Taglish, and any other variations of the language.” Those “other variations” impressively include text-speak, which often intersperses numbers with letters as in “nkk2wa” for “nakakatuwa” (something that makes you glad), and beki-speak (gay lingo), which uses a bevy of co-opted words (“anaconda”, which means “traitor”) and coined words (“deadmatologist”, which means “snob”).
According to Senti co-founder Ralph Vincent Regalado, the volume of this data is considerable because “Filipino acquires borrow words from other languages faster than English.”
After consolidating this information “from different social media streams,” Senti uses it to provide “analytics and data insights about products, events, and personalities.” This ability effectively makes it one of the first social media tools locally produced in the Philippines, making Regalado a sort of techno-linguistic pioneer.
Our output is better than yours
The intersection of tech and linguistics had long been an interest of co-founder Ralph Vincent Regalado. He says, “During the time I was taking my BS and MS computer science degree at De La Salle University, I was introduced to Filipino natural language processing.”
Regalado had wanted to pitch the idea for what would become Senti as early as the fourth Startup Weekend Manila, but because of scheduling conflicts, he had to wait until the fifth Startup Weekend Manila to pursue it. There, he partnered with some other attendees that complimented his skills – Regalado was tech, his co-founders were biz-dev – to develop Senti, or at least the idea for it. Together they pitched the concept of Senti, and in heroic fashion, they won the fifth Startup Weekend Manila competition.
Those co-founders are not the same people with Regalado today. “We had conflicting ideas on how to handle Senti,” he says. “Though I know that their intentions were good, there are things that you want to risk so that you’ll learn fast. This made me decide to press the reset button and start again with a new team.”
Regalado’s team may have changed – it now consists of Charibeth Cheng (Regalado’s thesis adviser during his BS and MS degree), Marc Obaldo (development lead), Arnold and Archie Choa (“our developer twins”), and Courtney Ngo (researcher) – but the dream remains the same: he wishes to unearth the riches of data buried in the Filipino languages.
To begin to understand the complexity of this task, we can look at a single word. “Take “siya” (him / her), it is also spelled as “sya” and “xa” (in text-speak),” says Senti co-founder Charibeth Cheng. “Spelling is a major challenge for Filipino because there are many variants. Filipinos also love to speak with hidden meanings, thus making language analysis more difficult.”
When pitching to potential client companies, Regalado relishes in discussing the big picture as much as he does the linguistic details:
Senti is suited to serve the Philippine market as most people use the local language when they tweet or post. It gives a better view of what consumers really feel about a specific brand. Understanding the local language and contextualizing data to fit the Philippine market makes our output more useful for local companies.
Regalado also does not shy away from directly appealing to nationalism. He says, “I share that it’s us who can best understand Filipinos because we ourselves are also Filipinos – we know the language, we know the culture.”
The angle, for the most part, has worked – Senti is currently in talks with “eight possible clients and strategic partners” that can help them grow. In the meantime, however, the small team of six has to avoid spreading themselves too thin.
“Given that we are a small team it is very important that we balance our skills to the needs of our startup – management, product development, client acquisition, and so on,” Regalado says. “We are trying to overcome this challenge by prioritizing tasks and resolving issues as quickly as possible.”
Senti as a telepath
The outlook for Filipino social media products is bright:
There’s a huge market for local social media tools and e-commerce tools in the country. It is continuously growing given the fact that there’s a shift now from traditional media to digital and the high penetration and usage rate of social media in the country.
This is not a good thing for Senti, of course. As the first to market in the world of Filipino social media products, Senti has an early advantage, but that can be easily squandered. As such, they are refusing to rest on their laurels. One of their immediate goals is to “make the engine smarter and its analysis more understandable to its users,” so that the company can catch on and eventually expand into southeast Asia, where it can reign as the “leading innovative social media provider.”
To get there, Regalado thinks Senti should be able to see into the future. “One of the features we are planning to add is the ability to forecast trends in social media, more importantly in the local context,” he says. “This will help in determining what consumers will want or not want in the future.”
As an example, Regalado cites “loom bands”. He explains, “Detecting the loom band trend early may help Company X to release related products or utilize this trend in their marketing and sales.”
The benefits also extend to television. “Another example is detecting for what type of shows Pinoys [Filipinos] will watch. This feature will help provide information on where to position your TV advertisements that targets your desired audience,” Regalado explains.
In the end, the feature, however advanced it ends up being, will still only cast predictions. It’s an imperfect science and the real world has a convenient way of surprising pundits. As such, Senti’s success, at least with this feature, will be determined as much by as its accuracy as with how well they manage client expectations.
On the technical side, their web crawler would have to mine a “voluminous amount of social media data” as the basis for forecasting trends. This kind of scale can drain a company’s resources and manpower, but Regalado remains optimistic, if a bit romantic:
Editing by Paul Bischoff
Understanding local social media data is a challenging and daunting task. Why? Because each of us is unique, the way we speak, the way we do things and the way they interact is different from one another. But no matter how challenging it is, making sense of this data is more valuable because it help us understand better who we are. Giving everyone a chance for their voices to be heard loud and clear.