Generating empathetic responses through emotion tracking and constraint guidance

doi:10.1007/s11704-023-2792-7

Front. Comput. Sci.

2024, Vol. 18

Issue (2) : 182330 https://doi.org/10.1007/s11704-023-2792-7

Artificial Intelligence

Generating empathetic responses through emotion tracking and constraint guidance

Jing LI¹, Donghong HAN^1,²(

), Zhishuai GUO³, Baiyou QIAO¹, Gang WU¹

¹. School of Computer Science and Engineering, Northeastern University, Shenyang 110000, China
². Key Laboratory of Intelligent Computing in Medical Image (Ministry of Education), Northeastern University, Shenyang 110000, China
³. Neusoft Corporation, Shenyang 110000, China

Download: PDF(1180 KB) HTML
Export: BibTeX | EndNote | Reference Manager | ProCite | RefWorks

Corresponding Author(s): Donghong HAN

Just Accepted Date: 12 July 2023 Issue Date: 14 September 2023

Cite this article:

Jing LI,Donghong HAN,Zhishuai GUO, et al. Generating empathetic responses through emotion tracking and constraint guidance[J]. Front. Comput. Sci., 2024, 18(2): 182330.

URL:

https://academic.hep.com.cn/fcs/EN/10.1007/s11704-023-2792-7
https://academic.hep.com.cn/fcs/EN/Y2024/V18/I2/182330

Fig.1 The related diagrams of ETHREED, including (a) the overall architecture, (b) the policy network and (c) the pointer generation network

Method	Params.	Automatic evaluation					Human rating			A/B test (ETHREED vs)
Method	Params.	Acc	PPL	Dis-1	Dis-2	$F B E R T$	Emp.	Rel.	Flu.	Win/%	Loss/%	Tie/%	$κ$
MIME	17.80M	0.3239	37.28	0.39	1.56	0.189	3.14	3.19	4.85	46.67	27.33	26.00	0.45
EmpDG	29.29M	0.3278	36.26	0.45	2.13	0.195	3.21	3.19	4.72	41.33	35.33	23.33	0.44
KEMP	32.54M	0.3852	34.56	0.61	2.69	0.223	3.20	3.16	4.79	43.33	31.67	25.00	0.46
EmpHi	10.92M	0.4046	33.49	0.98	4.64	0.187	3.07	3.05	4.67	40.67	37.67	21.67	0.36
ETHREED	11.98M	0.4294	27.74	1.14	4.76	0.226	3.25	3.23	4.85	?	?	?	?

Tab.1 All results of automatic and human evaluation. The best performing results are highlighted. We reproduce the baselines based on the hyperparameters in their papers, repeat 5 runs with different seeds and average the results for each automatic metrics. Emp., Rel. and Flu. are short for Empathy, Relevance and Fluency. The Flessia-Kappa

κ

for three rating aspects are 0.47,0.51 and 0.41, which indicate three annotators reach a moderate agreement.

κ

of a/b test all achieve the fair or moderate agreement

Tab.2 Examples of comparison model responses

1	I V, Serban A, Sordoni Y, Bengio A, Courville J Pineau . Building end-to-end dialogue systems using generative hierarchical neural network models. In: Proceedings of the 30th AAAI Conference on Artificial Intelligence. 2016, 3776−3784
2	S Levine . Motor skill learning with local trajectory methods. Stanford University, Dissertation, 2014
3	Majumder N, Hong P, Peng S, Lu J, Ghosal D, Gelbukh A, Mihalcea R, Poria S. MIME: mimicking emotions for empathetic response generation. In: Proceedings of 2020 Conference on Empirical Methods in Natural Language Processing. 2020, 8968−8979
4	Q, Li H, Chen Z, Ren P, Ren Z, Tu Z Chen . EmpDG: multi-resolution interactive empathetic dialogue generation. In: Proceedings of the 28th International Conference on Computational Linguistics. 2020, 4454−4466
5	Q, Li P, Li Z, Ren P, Ren Z Chen . Knowledge bridging for empathetic dialogue generation. In: Proceedings of the 36th AAAI Conference on Artificial Intelligence. 2022, 10993−11001
6	Chen M Y, Li S, Yang Y. EmpHi: Generating empathetic responses with human-like intents. In: Proceedings of 2022 Conference of the North American Chapter of Association for Computational Linguistics: Human Language Technologies. 2022, 1063−1074
7	Li J, Galley M, Brockett C, Gao J, Dolan B. A diversity-promoting objective function for neural conversation models. In: Proceedings of 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2016, 110−119
8	T, Zhang V, Kishore F, Wu K Q, Weinberger Y Artzi . BERTScore: evaluating text generation with BERT. In: Proceedings of the 8th International Conference on Learning Representations. 2020
9	H, Rashkin E M, Smith M, Li Y L Boureau . Towards empathetic open-domain conversation models: a new benchmark and dataset. In: Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics. 2019, 5370−5381

[1]	FCS-22792-OF-JL_suppl_1	Download
[2]	FCS-22792-OF-JL_suppl_2	Download

Viewed

Full text

Abstract

Cited

Shared

Discussed