Lower Perplexity is Not Always Human-Like.

Tatsuki Kuribayashi Yohei Oseki Takumi Ito Ryo Yoshida Masayuki Asahara Kentaro Inui

Published in: ACL/IJCNLP (1) (2021)

Keyphrases

language model
topic models
significantly lower
databases
speech recognition
significantly higher
data sets
image sequences
pattern recognition
artificial neural networks
email
hidden markov models
knowledge base
virtual environment
machine learning
intelligent agents
real world
human behavior