{"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/improving-neural-response-diversity-with","title":"Improving Neural Response Diversity with Frequency-Aware Cross-Entropy Loss","arxiv_id":"1902.09191","date":"2019-02-25","proceeding":null,"authors":["Shaojie Jiang","Pengjie Ren","Christof Monz","Maarten de Rijke"],"abstract":"Sequence-to-Sequence (Seq2Seq) models have achieved encouraging performance\non the dialogue response generation task. However, existing Seq2Seq-based\nresponse generation methods suffer from a low-diversity problem: they\nfrequently generate generic responses, which make the conversation less\ninteresting. In this paper, we address the low-diversity problem by\ninvestigating its connection with model over-confidence reflected in predicted\ndistributions. Specifically, we first analyze the influence of the commonly\nused Cross-Entropy (CE) loss function, and find that the CE loss function\nprefers high-frequency tokens, which results in low-diversity responses. We\nthen propose a Frequency-Aware Cross-Entropy (FACE) loss function that improves\nover the CE loss function by incorporating a weighting mechanism conditioned on\ntoken frequency. Extensive experiments on benchmark datasets show that the FACE\nloss function is able to substantially improve the diversity of existing\nstate-of-the-art Seq2Seq response generation methods, in terms of both\nautomatic and human evaluations.","url_abs":"http://arxiv.org/abs/1902.09191v1","url_pdf":"http://arxiv.org/pdf/1902.09191v1.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":"improving-neural-response-diversity-with","repo_url":"https://github.com/ShaojieJiang/FACE","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"improving-neural-response-diversity-with","repo_url":"https://github.com/fadedcosine/pos-guided-neural-text-generation","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"diversity","task_name":"Diversity"},{"task_slug":"response-generation","task_name":"Response Generation"}],"methods":[{"method_slug":"lstm","method_name":"LSTM"},{"method_slug":"seq2seq","method_name":"Seq2Seq"},{"method_slug":"sigmoid-activation","method_name":"Sigmoid Activation"},{"method_slug":"tanh-activation","method_name":"Tanh Activation"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1902.09191","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}