[3Rin4-53] Continuous Transformation of Natural Language Syntactic Structure using Grammar Variational Autoencoder
Keywords:Grammar Variational Autoencoder, Syntactic structure transformation, Context-free Grammar
Deep generative models are generative models configured by a deep neural network, such as Variational Autoencoder or Generative Adversarial Network.
In deep generative model, input data is generally continuous data such as images and sounds.
On the other hand, it is not easy to handle and generate discrete data such as chemical formulas and mathematical formulas.
Ingenuity is required to handle it.
Kusner et al. developed a Grammar Variational Autoencoder (GVAE), a model that enables data with discrete structures such as chemical and mathematical formulas to be treated as continuous values in the framework of deep learning.
In this research, we try to continuously convert the syntactic structure of natural language represented by context-free grammar in the framework of GVAE.
In deep generative model, input data is generally continuous data such as images and sounds.
On the other hand, it is not easy to handle and generate discrete data such as chemical formulas and mathematical formulas.
Ingenuity is required to handle it.
Kusner et al. developed a Grammar Variational Autoencoder (GVAE), a model that enables data with discrete structures such as chemical and mathematical formulas to be treated as continuous values in the framework of deep learning.
In this research, we try to continuously convert the syntactic structure of natural language represented by context-free grammar in the framework of GVAE.
Authentication for paper PDF access
A password is required to view paper PDFs. If you are a registered participant, please log on the site from Participant Log In.
You could view the PDF with entering the PDF viewing password bellow.