Natural-Language-Processing

Natural Language Processing

 

How deeply is the evolution of natural language rooted in animal communication, in birdsong?  Birds' notes are like characters, their syllables like words, their song motifs like sentences, their song bouts like paragraphs, and their vocal repertoire like a dictionary. Are birds' song learning strategies also potentially relevant for natural language processing (NLP)?

 

One important aspect of our birdsong research is to probe the links between birdsong and NLP. Modern NLP methods can fuel birdsong research thanks to behavioral similarities between human and animal communication.

 

With our collaborator Prof. Dina Lipkind, we found that juvenile zebra finches learn their songs by matching each of their song syllables to the most acoustically similar syllable in the template song, regardless of its temporal position. Thereby, zebra finches prioritize efficient learning of syllable vocabulary at the cost of inefficient syntax learning. Computationally, birds reduce the complex nonlinear problem of song learning to a tractable linear problem. Interestingly, nearly identical linearization strategies have been applied to the problem of identifying related documents within large text corpora; these strategies achieve excellent document classification performance.  

 

Reference

 

Lipkind, D., Zai, A. T., Hanuschkin, A., Marcus, G. F., Tchernichovski, O., & Hahnloser, R. H. (2017). Songbirds work around computational complexity by learning song vocabulary independently of sequence. Nature communications8(1), 1247.

https://www.nature.com/articles/s41467-017-01436-0