NCCR Evolving Language (National Centre of Competence in Research)

A brain-inspired computer model that understands speech like humans

Researchers developed a computer model based on human brain mechanisms to improve speech comprehension. The model extracts multilevel information from ongoing speech and uses non-linguistic knowledge for disambiguating word meanings. This approach is more human-like than existing language models like ChatGPT.

SourceNCCR Evolving Language (National Centre of Competence in Research)·JournalPLOS Biology·TypeComputational simulation/modeling·DateMar 22, 2023

Our social interactions begin at a young age

Researchers studied 31 children's social interactions with peers aged 2-4, finding they have more frequent and shorter interactions than great apes. Children learn quickly about coordinating behavior through exposure to multiple partners.

SourceNCCR Evolving Language (National Centre of Competence in Research)·JournalPhilosophical Transactions of the Royal Society B Biological Sciences·DateJul 27, 2022

When simpler is harder

Researchers measured brain activity in Hindi speakers describing different images to study sentence planning before speaking. They found that having fewer grammatical distinctions makes planning particularly demanding for the brain and requires more neural activity.

SourceNCCR Evolving Language (National Centre of Competence in Research)·JournalPLOS Biology·DateJan 27, 2021