Pipilika

» see pipilika.com as it lived, on the wayback machine

~ the eulogy ~

Pipilika was built for a language I cannot read. That is the honest starting position, and it also happens to be the whole reason the thing existed, so it belongs in the first line rather than buried at the end.

It launched on 13 april 2013 at a hotel in Dhaka, opened by the country's ICT minister, which tells you how seriously Bangladesh was taking the moment. The code came out of the computer science and engineering department at Shahjalal University of Science and Technology in Sylhet. Students wrote it. Muhammad Zafar Iqbal, who was a professor there and one of the best known writers in the country, supervised the project as its director.

The engine took queries in Bangla and in English. What it went after was the material the global crawlers handled badly: the national newspapers publishing in Bangla and the blog world that had grown up around them. Bengali Wikipedia went into the index too. Any engine that wants to do this properly has to understand how Bangla words change shape, because a search for a root form has to find every inflected version of it or the index is decoration. That work is unglamorous and it does not transfer from English.

Turkey ran the same experiment earlier and lost it earlier. There were Turkish engines in the late nineties, Arabul and Netbul among them, and academic papers measuring how well each one coped with Turkish morphology. By the time I was at university none of that was a live question, because the box everyone typed into was Google and its Turkish had become good enough. Reading about Pipilika from here is reading a question I already know the ending to, asked in an alphabet I do not have.

~ the dossier ~

~ what the ants indexed ~

The name is the Bangla word for ant, and the logo was an ant, which is a better metaphor for a crawler than most companies manage. The premise was coverage rather than cleverness. Google indexed Bangla pages, obviously, but it indexed them the way it indexed everything, and in 2013 the results for a Bangla query were thin in a way that was invisible to anyone searching in English.

A national engine is really an argument about who decides what is worth crawling. Pipilika's answer was that the Bangladeshi press and the Bangladeshi blogs were worth crawling exhaustively, even though the ad market attached to them was almost nothing. That is a defensible position for a university and an impossible one for a company.

~ the funding that never came ~

It stayed what it started as. The project was a thesis, run inside a department, and no serious money ever arrived to convert it into something that paid for its own servers. Search is not a product you ship once. The index rots weekly and the crawler needs babysitting, and the cost of both goes up with the size of the web whether or not anyone is using you.

Against that, the revenue side was empty. No advertising business was ever started, and no licensing or government deal that anyone documented came in behind it. Eight years of running on institutional goodwill is longer than most people would guess, and it ended the way that always ends.

~ why nobody heard of it ~

The launch got press because a minister was standing next to it. After that there was close to nothing, no campaign and no reason for an ordinary person in Dhaka to change the habit they already had. A search engine is a habit before it is anything else, and habits do not move because a better index exists somewhere.

So the thing that killed Pipilika was not a competitor arriving. It was already there, it was already the default, and it kept getting better at Bangla every year the students were not being paid. By 2021 the gap the engine was built to fill had mostly closed on its own, and the site went dark without an announcement, which is what happens when the audience you would have to notify was never assembled.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »