Page 162 - 《软件学报》2025年第4期
P. 162

1568                                                       软件学报  2025  年第  36  卷第  4  期


                 会对不同槽实体的位置进行区分. 而且            artist、music_item  均属于领域共享槽类型, 显然模型可以已经从源域中获
                 得这方面的知识. 经过统计, 当数据中存在            artist 时有  38%  的概率存在  music_item, 有  58%  的概率存在  playlist_
                 owner. artist 与两者有很强的依赖关系, 我们的模型会按照潜在的语言习惯选择合适的答案片段, 而不像                         baseline
                 在预测错误后直接影响另一个强相关的槽类型.

                 4   总 结

                    本文提出了基于槽依赖建模的跨领域槽填充方法, 该方法通过生成式的框架同时对多个槽类型进行预测, 并
                 且每个槽类型的提示序列包含槽语义提示和槽共享提示两部分, 从而引入了槽语义知识和不同槽之间的隐式依
                 赖, 不同槽类型之间可以互相辅助生成. 对于多实体生成带来的实体类型匹配问题, 我们设计了话语填充子任务来
                 弥补这一缺陷. 通过填充掩码后的话语, 增强了实体与话语的语义感知, 进而提升了类型匹配的准确率. 在                               SNIPS
                 和  TOP  上的实验结果证实了本文模型在各个方面的性能优势, 并且通过各种比较实验和示例分析验证了模型各
                 部分的有效性. 在后续的研究中, 我们会进一步挖掘任务中存在的其他依赖或潜在信息, 尽可能从多个角度为模型
                 提供信息.

                 References:
                  [1]  Goo CW, Gao G, Hsu YK, Luo CL, Chen TC, Hsu KW, Chen YN. Slot-gated modeling for joint slot filling and intent prediction. In:
                     Proc.  of  the  2018  Conf.  of  the  North  American  Chapter  of  the  Association  for  Computational  Linguistics:  Human  Language
                     Technologies, Vol. 2 (Short Papers). New Orleans: ACL, 2018. 753–757. [doi: 10.18653/v1/N18-2118]
                  [2]  Li X, Wang YY, Tür G. Multi-task learning for spoken language understanding with shared slots. In: Proc. of the 12th Annual Conf. of
                     the Int’l Speech Communication Association. Florence: ISCA, 2011. 701–704.
                  [3]  Yazdani M, Henderson J. A model of zero-shot learning of spoken language understanding. In: Proc. of the 2015 Conf. on Empirical
                     Methods in Natural Language Processing. Lisbon: ACL, 2015. 244–249. [doi: 10.18653/v1/D15-1027]
                  [4]  Zhang CW, Li YL, Du N, Fan W, Yu P. Joint slot filling and intent detection via capsule neural networks. In: Proc. of the 57th Annual
                     Meeting of the Association for Computational Linguistics. Florence: ACL, 2019. 5259–5267. [doi: 10.18653/v1/P19-1519]
                  [5]  Qin LB, Liu TL, Che WX, Kang BB, Zhao SD, Liu T. A co-interactive Transformer for joint slot filling and intent detection. In: Proc. of
                     the  2021  IEEE  Int’l  Conf.  on  Acoustics,  Speech  and  Signal  Processing  (ICASSP).  Toronto:  IEEE,  2021.  8193–8197.  [doi:  10.1109/
                     ICASSP39728.2021.9414110]
                  [6]  Wu D, Ding L, Lu F, Xie J. SlotRefine: A fast non-autoregressive model for joint intent detection and slot filling. In: Proc. of the 2020
                     Conf. on Empirical Methods in Natural Language Processing (EMNLP). ACL, 2020. 1932–1937. [doi: 10.18653/v1/2020.emnlp-main.
                     152]
                  [7]  Ferreira E, Jabaian B, Lefèvre F. Online adaptative zero-shot learning spoken language understanding using word-embedding. In: Proc. of
                     the 2015 IEEE Int’l Conf. on Acoustics, Speech and Signal Processing (ICASSP). South Brisbane: IEEE, 2015. 5321–5325. [doi: 10.1109/
                     ICASSP.2015.7178987]
                  [8]  E HH, Niu PQ, Chen ZF, Song MN. A novel bi-directional interrelated model for joint intent detection and slot filling. In: Proc. of the
                     57th Annual Meeting of the Association for Computational Linguistics. Florence: ACL, 2019. 5467–5471. [doi: 10.18653/v1/P19-1544]
                  [9]  Li XC, Wang YJ, Gan L, Zhan DC. Exploring transferability measures and domain selection in cross-domain slot filling. In: Proc. of the
                     2022 IEEE Int’l Conf. on Acoustics, Speech and Signal Processing (ICASSP 2022). Singapore: IEEE, 2022. 3758–3762. [doi: 10.1109/
                     ICASSP43922.2022.9746890]
                 [10]  Bapna A, Tür G, Hakkani-Tür D, Heck L. Towards zero-shot frame semantic parsing for domain scaling. In: Proc. of the 18th Annual
                     Conf. of the Int’l Speech Communication Association. Stockholm: ISCA, 2017. 2476–2480.
                 [11]  Shah D, Gupta R, Fayazi A, Hakkani-Tur D. Robust zero-shot cross-domain slot filling with example values. In: Proc. of the 57th Annual
                     Meeting of the Association for Computational Linguistics. Florence: ACL, 2019. 5484–5490. [doi: 10.18653/v1/P19-1547]
                 [12]  Liu Z, Winata GI, Xu P, Fung P. Coach: A coarse-to-fine approach for cross-domain slot filling. In: Proc. of the 58th Annual Meeting of
                     the Association for Computational Linguistics. ACL, 2020. 19–25. [doi: 10.18653/v1/2020.acl-main.3]
                 [13]  He  KQ,  Zhang  JC,  Yan  YM,  Yan  YM,  Xu  WR,  Niu  C,  Zhou  J.  Contrastive  zero-shot  learning  for  cross-domain  slot  filling  with
                     adversarial attack. In: Proc. of the 28th Int’l Conf. on Computational Linguistics. Barcelona: ACL, 2020. 1461–1467. [doi: 10.18653/v1/
                     2020.coling-main.126]
   157   158   159   160   161   162   163   164   165   166   167