eprintid: 10059972 rev_number: 21 eprint_status: archive userid: 608 dir: disk0/10/05/99/72 datestamp: 2018-10-30 11:50:49 lastmod: 2021-11-10 01:57:50 status_changed: 2018-10-30 11:50:49 type: proceedings_section metadata_visibility: show creators_name: Shah, H creators_name: Zheng, B creators_name: Barber, D title: Generating Sentences Using a Dynamic Canvas divisions: UCL divisions: B04 divisions: C05 divisions: F48 keywords: NLP; bayesian; variational; generative; sentences note: Copyright © 2018, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved. abstract: We introduce the Attentive Unsupervised Text (W)riter (AUTR), which is a word level generative model for natural language. It uses a recurrent neural network with a dynamic attention and canvas memory mechanism to iteratively construct sentences. By viewing the state of the memory at intermediate stages and where the model is placing its attention, we gain insight into how it constructs sentences. We demonstrate that AUTR learns a meaningful latent representation for each sentence, and achieves competitive log-likelihood lower bounds whilst being computationally efficient. It is effective at generating and reconstructing sentences, as well as imputing missing words. date: 2018 date_type: published publisher: AAAI Association for the Advancement of Artificial Intelligence official_url: https://www.aaai.org/ocs/index.php/AAAI/AAAI18/paper/view/16260 oa_status: green language: eng primo: open primo_central: open_green verified: verified_manual elements_id: 1511508 lyricists_name: Barber, David lyricists_name: Shah, Harshil lyricists_id: DBARB05 lyricists_id: HSHAH40 actors_name: Barber, David actors_name: Turnbull, Sarah actors_id: DBARB05 actors_id: SLTUR91 actors_role: owner actors_role: impersonator full_text_status: public publication: Association for the Advancement of Artificial Intelligence event_title: Thirty-Second AAAI Conference on Artificial Intelligence, 2 – 7 February 2018, New Orleans, Louisiana, USA institution: AAAI Association for the Advancement of Artificial Intelligence book_title: Thirty-Second AAAI Conference on Artificial Intelligence (AAAI-18) citation: Shah, H; Zheng, B; Barber, D; (2018) Generating Sentences Using a Dynamic Canvas. In: Thirty-Second AAAI Conference on Artificial Intelligence (AAAI-18). AAAI Association for the Advancement of Artificial Intelligence Green open access document_url: https://discovery.ucl.ac.uk/id/eprint/10059972/1/Shah-Zheng.pdf