Generating Topical Poetry

Ghazvininejad, Marjan; Shi, Xing; Choi, Yejin; Knight, Kevin

doi:10.18653/v1/d16-1126

Cited by 118 publications

(139 citation statements)

References 16 publications

Supporting

Mentioning

137

Contrasting

Unclassified

Order By: Relevance

“…One main reason is that the generic Seq2Seq baseline model tends to generate short sentences, often less than five words, which decreases the meaningfulness and increases repeatability of generated sentences. While with our generation model, two encoders of structure and content will mutually promote the effect of decoding referred to (Ghazvininejad 2016). Moreover, the KG model owns higher scores than SG, because the encoded keyword represents global context of next generation, resulting in the increase of meaningfulness and diversity scores.…”

Section: Human Evaluation Resultsmentioning

confidence: 99%

A Syllable-Structured, Contextually-Based Conditionally Generation of Chinese Lyrics

Wang

Zhuang

et al. 2019

PRICAI 2019: Trends in Artificial Intelligence

View full text Add to dashboard Cite

This paper presents a novel, syllable-structured Chinese lyrics generation model given a piece of original melody. Most previously reported lyrics generation models fail to include the relationship between lyrics and melody. In this work, we propose to interpret lyrics-melody alignments as syllable structural information and use a multi-channel sequence-to-sequence model with considering both phrasal structures and semantics. Two different RNN encoders are applied, one of which is for encoding syllable structures while the other for semantic encoding with contextual sentences or input keywords. Moreover, a large Chinese lyrics corpus for model training is leveraged. With automatic and human evaluations, results demonstrate the effectiveness of our proposed lyrics generation model. To the best of our knowledge, there is few previous reports on lyrics generation considering both music and linguistic perspectives.

show abstract

Section: Human Evaluation Resultsmentioning

confidence: 99%

A Syllable-Structured, Contextually-Based Conditionally Generation of Chinese Lyrics

Wang

Zhuang

et al. 2019

PRICAI 2019: Trends in Artificial Intelligence

View full text Add to dashboard Cite

show abstract

“…We focus here on an encoder-decoder architecture. Originally applied to machine translation , encoder-decoder models have been extended to other sequence modeling tasks like dialogue generation (Serban et al, 2016;Shang et al, 2015) and poetry generation (Ghazvininejad et al, 2016;. We propose that this technique could be similarly useful for our task in establishing a mapping between cause-effect sequence pairs.…”

Section: Neural Network Approachmentioning

confidence: 99%

Proceedings of the First Workshop on Storytelling

2018

View full text Add to dashboard Cite

Human storytelling has existed for as far back as we can trace, predating writing. Humans have used stories for entertainment, education, cultural preservation; to convey experiences, history, lessons, morals; and to share the human experience.Part of grounding artificial intelligence work in human experience can involve the generation, understanding, and sharing of stories. This workshop highlights the diverse work being done in storytelling and AI across different fields.Papers at this workshop are multi-disciplinary, including work on neural, pipeline, and linguistic approaches to understanding and creating stories.We are also pleased to host a Visual Storytelling challenge, highlighting different methods for automatically generating stories given a set of images; and an invited talk from Nasrin Mostafazadeh on communicating about events through storytelling.Enjoy the workshop! Workshop website: http://www.visionandlanguage.net/workshop2018iii Storytelling ChallengeThe Storytelling challenge provides teams with the VIST dataset to generate stories from sequences of five images, and is hosted on EvalAI. Submissions for the challenge at NAACL 2018 were evaluated using both automatic metrics and human evaluation. The winner was chosen based on the best performance across the human evaluations for focus, structure and coherence, detail, how visually grounded the stories were, how shareable the stories were, and whether they sounded like they were written by a human. More details can be found on the workshop website.Submissions came from the following teams all over the world. Building AI systems that can process natural language input, comprehend it, and generate an engaging and contextually relevant output in response, has been one of the longest-running goals in AI. In human-human communications, a major trigger to our meaningful communications are "events" and how they cause/enable future events. We often communicate through telling stories in the form of related series of events. DG-DLMXIn this talk, I present my work on language processing in terms of events and how they interact with each other in time. Mainly through the lens of storytelling, I will focus on story comprehension and collaborative story generation, with a major emphasis on commonsense reasoning and narrative knowledge as showcased in the Story Cloze Test framework. Through different use cases, I will highlight the importance of establishing a contentful context and modeling multimodal contexts (such as visual and textual) in various AI tasks.vi AbstractIn this opinion piece, we argue that there is a need for alternative design directions to complement existing AI efforts in narrative and character generation and algorithm development. To make our argument, we a) outline the predominant roles and goals of AI research in storytelling; b) present existing discourse on the benefits and harms of narratives; and c) highlight the pain points in character creation revealed by semi-structured interviews we conducted with 14 individuals deeply involv...

show abstract

“…The work of Zhang and Lapata (2014) is one such example, where they were able to outperform all other classical Chinese poetry generation systems with both manual and automatic evaluation. Ghazvininejad et al (2016) and Goyal et al (2016) apply neural language models with regularising finite state machines. However, in the former case the rhythm of the output cannot be defined at sample time, and in the latter case the finite state machine is not trained on rhythm at all, as it is trained on dialogue acts.…”

Section: Related Workmentioning

confidence: 99%

Automatically Generating Rhythmic Verse with Neural Networks

Hopkins¹,

Kiela

2017

Proceedings of the 55th Annual Meeting of the Association For Computational Linguistics (Volume 1: Long Papers)

View full text Add to dashboard Cite

We propose two novel methodologies for the automatic generation of rhythmic poetry in a variety of forms. The first approach uses a neural language model trained on a phonetic encoding to learn an implicit representation of both the form and content of English poetry. This model can effectively learn common poetic devices such as rhyme, rhythm and alliteration. The second approach considers poetry generation as a constraint satisfaction problem where a generative neural language model is tasked with learning a representation of content, and a discriminative weighted finite state machine constrains it on the basis of form. By manipulating the constraints of the latter model, we can generate coherent poetry with arbitrary forms and themes. A large-scale extrinsic evaluation demonstrated that participants consider machine-generated poems to be written by humans 54% of the time. In addition, participants rated a machinegenerated poem to be the most human-like amongst all evaluated.

show abstract

Generating Topical Poetry

Cited by 118 publications

References 16 publications

A Syllable-Structured, Contextually-Based Conditionally Generation of Chinese Lyrics

A Syllable-Structured, Contextually-Based Conditionally Generation of Chinese Lyrics

Proceedings of the First Workshop on Storytelling

Automatically Generating Rhythmic Verse with Neural Networks

Contact Info

Product

Resources

About