Humans Who Read Grammars

Linguistic description and documentation involves a lot of data annotation and management. I am interested in how we can use computational tools to assist in this process, naturally only as an aid and not a replacement for human work. It is necessary to have humans in the loop and guards agains biases in tools that cause smaller languages to be misrepresented. At a basic level of data annotation, I have made posts in the past about my ELAN workflow for  segmenting speakers into tiers  , creating  tiers from search results in ELAN  and  another way of doing segmentation in ELAN via PRAAT (kudos to Kashima and Ellis). I think there's a common misconception that computational tools in language documentation/description is something very new, that before the dawn of LLMs there wasn't much done. That's not the case, researchers have long been interested in how to effective part of their labour - especially the language technology branch of the  ARC Centre of Exc...

Lots of languages in the world will fall out of use within the next decades and  century. There are between 6 - 9,000 languages in the world, and the language reference catalogue Glottolog (v5.3) estimates that over 4,000 languages are in a threatened or worse status*.   There are many reasons why this is a bad situation. One aspect that is often  mentioned is that when languages disappear , we lose an important facet of what it means to be human. All languages are different from each other, they represent different records of our history and capacity . When one falls out of use, especially when there is no documentation  of it, we as humankind loose information about what we are capable of, what it means to be human - how far we can reach, what unites us and what varies. This was for example argued very well by Dr Levinson in a talk I had the pleasure of attending in Berlin last week (poster ⇨)   It also forms part of the discussion between Cooperrider and Dr B...