Neural Style Transfer for Texts

Student: Bagiyan Nerses

Faculty: HSE Tikhonov Moscow Institute of Electronics and Mathematics (MIEM HSE)

Educational Programme: Applied Mathematics (Bachelor)

Final Grade: 9

Year of Graduation: 2019

Style transfer is a task that requires to create a mapping between source and target styles without changing the intent of the original sample (usually we choose an image or a sentence as a sample). In this work, we present the methodology for style transfer in NLP set up. We show that models which use techniques of adversarial examples generation outperform models proposed in recent works. Moreover, experiments prove that conditioning decoder with proper latent representation in variational autoencoder for text modeling could change the style of the sentence preserving its meaning. Compared to the previous state-of-the-art works model proposed in this research shows better results in the same experimental setup.

Full text (added May 23, 2019)

Student Theses at HSE must be completed in accordance with the University Rules and regulations specified by each educational programme.

Summaries of all theses must be published and made freely available on the HSE website.

The full text of a thesis can be published in open access on the HSE website only if the authoring student (copyright holder) agrees, or, if the thesis was written by a team of students, if all the co-authors (copyright holders) agree. After a thesis is published on the HSE website, it obtains the status of an online publication.

Student theses are objects of copyright and their use is subject to limitations in accordance with the Russian Federation’s law on intellectual property.

In the event that a thesis is quoted or otherwise used, reference to the author’s name and the source of quotation is required.

Search all student theses