 In this type of text generation a pre-trained TGM model is combined with one or more attribute classifiers which govern text generation.

San Jose State University

It is possible to control the topic and sentiment of text generated ,two methods by which we can control the text generated are

 - Training with control tokens
 - Control using attribute classifier


Controllable Text Generative Models

Automatic Detection of Machine Generated Text: A Critical Survey
by Ganesh Jawahar, Muhammad Abdul-Mageed, Laks V.S. Lakshmanan
https://arxiv.org/pdf/2011.01314.pdf

Automatic Detection of Machine Generated Text: A Critical Survey


 - Meta information about the article such as author, date of creation, source domain will be appended as extra token in addition to the input sequence and sent for training 

 - Meta information allows the TGM to learn the relationship between original content and the meta information.

 - First controllable TGM proposed is the GROVER model which can create trustworthy fake news content using meta-information of the news article (such as headline, author, and date).  that is harder for humans to identify than human written fake news and can thus pose a significant threat.


Learn Before

Related