Advanced

News text generation with adversarial deep learning

Månsson, Filip LU and Månsson, Fredrik LU (2017) In LU-CS-EX 2017-18 EDA920 20171
Department of Computer Science
Abstract
In this work we carry out a thorough analysis of applying a specific field within machine learning called generative adversarial networks, to the art of natural language generation; more specifically we generate news text articles in an automated fashion. To do this, we experimented with a few different architectures and representations of text, evaluated the results and used the information retrieved from the results, to create a model that should give the best result.
For evaluation, we used perplexity and human evaluation. We also looked at the token distribution to see which model captures the texts most successfully.
We show that it is possible to use generative adversarial networks to generate sequences of tokens that resemble... (More)
In this work we carry out a thorough analysis of applying a specific field within machine learning called generative adversarial networks, to the art of natural language generation; more specifically we generate news text articles in an automated fashion. To do this, we experimented with a few different architectures and representations of text, evaluated the results and used the information retrieved from the results, to create a model that should give the best result.
For evaluation, we used perplexity and human evaluation. We also looked at the token distribution to see which model captures the texts most successfully.
We show that it is possible to use generative adversarial networks to generate sequences of tokens that resemble natural language, but this does not yet reach the quality of human-written text. Further hyperparameter tuning and using a narrower-subjected corpus could improve the output. (Less)
Please use this url to cite or link to this publication:
author
Månsson, Filip LU and Månsson, Fredrik LU
supervisor
organization
alternative title
Generering av nyhetsartiklar med adversarial deep learning
course
EDA920 20171
year
type
H3 - Professional qualifications (4 Years - )
subject
keywords
Machine learning, generative adversarial learning, GAN, natural language generation
publication/series
LU-CS-EX 2017-18
report number
LU-CS-EX 2017-18
ISSN
1650-2884
language
English
id
8925528
date added to LUP
2017-09-13 10:56:00
date last changed
2017-09-13 10:56:00
@misc{8925528,
  abstract     = {In this work we carry out a thorough analysis of applying a specific field within machine learning called generative adversarial networks, to the art of natural language generation; more specifically we generate news text articles in an automated fashion. To do this, we experimented with a few different architectures and representations of text, evaluated the results and used the information retrieved from the results, to create a model that should give the best result.
For evaluation, we used perplexity and human evaluation. We also looked at the token distribution to see which model captures the texts most successfully.
We show that it is possible to use generative adversarial networks to generate sequences of tokens that resemble natural language, but this does not yet reach the quality of human-written text. Further hyperparameter tuning and using a narrower-subjected corpus could improve the output.},
  author       = {Månsson, Filip and Månsson, Fredrik},
  issn         = {1650-2884},
  keyword      = {Machine learning,generative adversarial learning,GAN,natural language generation},
  language     = {eng},
  note         = {Student Paper},
  series       = {LU-CS-EX 2017-18},
  title        = {News text generation with adversarial deep learning},
  year         = {2017},
}