{"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/what-to-talk-about-and-how-selective","title":"What to talk about and how? Selective Generation using LSTMs with Coarse-to-Fine Alignment","arxiv_id":"1509.00838","date":"2015-09-02","proceeding":"NAACL 2016 6","authors":["Hongyuan Mei","Mohit Bansal","Matthew R. Walter"],"abstract":"We propose an end-to-end, domain-independent neural encoder-aligner-decoder\nmodel for selective generation, i.e., the joint task of content selection and\nsurface realization. Our model first encodes a full set of over-determined\ndatabase event records via an LSTM-based recurrent neural network, then\nutilizes a novel coarse-to-fine aligner to identify the small subset of salient\nrecords to talk about, and finally employs a decoder to generate free-form\ndescriptions of the aligned, selected records. Our model achieves the best\nselection and generation results reported to-date (with 59% relative\nimprovement in generation) on the benchmark WeatherGov dataset, despite using\nno specialized features or linguistic resources. Using an improved k-nearest\nneighbor beam filter helps further. We also perform a series of ablations and\nvisualizations to elucidate the contributions of our key model components.\nLastly, we evaluate the generalizability of our model on the RoboCup dataset,\nand get results that are competitive with or better than the state-of-the-art,\ndespite being severely data-starved.","url_abs":"http://arxiv.org/abs/1509.00838v2","url_pdf":"http://arxiv.org/pdf/1509.00838v2.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":"what-to-talk-about-and-how-selective","repo_url":"https://github.com/HMEIatJHU/SelectiveGeneration","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"none","reach":null}],"tasks":[{"task_slug":"data-to-text-generation","task_name":"Data-to-Text Generation"},{"task_slug":"decoder","task_name":"Decoder"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1509.00838","atlas_url":"https://app.syntology.ai/?focus=1509.00838","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}