Why Work on So-Called "Low-Resource" Languages?
From time to time, students ask me to supervise a master's or PhD project related to natural language processing for so-called low-resource languages.
Let me be clear from the outset: this is not a research topic. At best, it is a broad research area.
It makes little sense to insist on building an ASR system or a parser for a particular language simply for the sake of it. What really matters is why doing so is worthwhile. Yet most students have no clear idea of the actual value of such work. I often hear arguments like, "It will allow people to communicate in their own language…"—as if the mere existence of a parser or an automatic transcription system (using which writing system, exactly?) were somehow going to change the world.
The real question is the research question: how will my work advance the state of the art? What concrete impact will my research have?
This lack of scientific reflection is often even more apparent among young computer scientists. No, not everything boils down to vectors.
The solution? Read. And not just linguistics. Read widely—and read seriously. Then we can have this conversation again.