{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/skipflow-incorporating-neural-coherence","title":"SkipFlow: Incorporating Neural Coherence Features for End-to-End Automatic Text Scoring","arxiv_id":"1711.04981","date":"2017-11-14","proceeding":null,"authors":["Yi Tay","Minh C. Phan","Luu Anh Tuan","Siu Cheung Hui"],"abstract":"Deep learning has demonstrated tremendous potential for Automatic Text\nScoring (ATS) tasks. In this paper, we describe a new neural architecture that\nenhances vanilla neural network models with auxiliary neural coherence\nfeatures. Our new method proposes a new \\textsc{SkipFlow} mechanism that models\nrelationships between snapshots of the hidden representations of a long\nshort-term memory (LSTM) network as it reads. Subsequently, the semantic\nrelationships between multiple snapshots are used as auxiliary features for\nprediction. This has two main benefits. Firstly, essays are typically long\nsequences and therefore the memorization capability of the LSTM network may be\ninsufficient. Implicit access to multiple snapshots can alleviate this problem\nby acting as a protection against vanishing gradients. The parameters of the\n\\textsc{SkipFlow} mechanism also acts as an auxiliary memory. Secondly,\nmodeling relationships between multiple positions allows our model to learn\nfeatures that represent and approximate textual coherence. In our model, we\ncall this \\textit{neural coherence} features. Overall, we present a unified\ndeep learning architecture that generates neural coherence features as it reads\nin an end-to-end fashion. Our approach demonstrates state-of-the-art\nperformance on the benchmark ASAP dataset, outperforming not only feature\nengineering baselines but also other deep learning models.","url_abs":"http://arxiv.org/abs/1711.04981v1","url_pdf":"http://arxiv.org/pdf/1711.04981v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"skipflow-incorporating-neural-coherence","repo_url":"https://github.com/carsonyang518/aaai24-aes-afg","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"automated-essay-scoring","task_name":"Automated Essay Scoring"},{"task_slug":"deep-learning","task_name":"Deep Learning"},{"task_slug":"feature-engineering","task_name":"Feature Engineering"},{"task_slug":"memorization","task_name":"Memorization"}],"methods":[{"method_slug":"lstm","method_name":"LSTM"},{"method_slug":"sigmoid-activation","method_name":"Sigmoid Activation"},{"method_slug":"tanh-activation","method_name":"Tanh Activation"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/automated-essay-scoring-on-asap","task":"Automated Essay Scoring","dataset":"ASAP-AES","model":"SkipFlow","rank_in_archive_order":5,"of":8,"metrics":{"Quadratic Weighted Kappa":"0.764"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1711.04981","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}