Guess who? Multilingual approach for the automated generation of author-stylized poetry

TL;DR

This paper presents a multilingual LSTM-based model with phonetic and semantic embeddings for generating stylized poetry, evaluated with BLEU, surveys, and a new metric, outperforming baselines and aligning with target authors.

Contribution

It introduces a novel multilingual LSTM model with extended embeddings and a new evaluation metric for stylized poetry generation, improving over existing methods.

Findings

The proposed model outperforms random and vanilla LSTM baselines.

Human evaluators associate generated poems with target authors.

A new cross-entropy based metric effectively evaluates stylized text quality.

Abstract

This paper addresses the problem of stylized text generation in a multilingual setup. A version of a language model based on a long short-term memory (LSTM) artificial neural network with extended phonetic and semantic embeddings is used for stylized poetry generation. The quality of the resulting poems generated by the network is estimated through bilingual evaluation understudy (BLEU), a survey and a new cross-entropy based metric that is suggested for the problems of such type. The experiments show that the proposed model consistently outperforms random sample and vanilla-LSTM baselines, humans also tend to associate machine generated texts with the target author.