Human-computer Interaction Model Based on Knowledge Graph Ripple Network
-
摘要: 针对当前人机交互系统机器人存在背景知识缺乏、回复连贯性不高的问题，该文提出一种基于知识图谱波纹网络的人机交互模型。为实现更加自然且智能化的人机交互系统，该文模型模拟了真实的人与人交流过程。首先，通过计算人机交互情感评估值和人机交互情感确信度得到人机交互情感友好度。然后，引入外部知识图谱作为机器人的背景知识，将对话实体嵌入知识图谱波纹网络获取参与人潜在的感兴趣的实体内容。最后，综合考量情感友好度和内容友好度给出机器人回复。实验结果表明，与对比模型相比，拥有背景知识且加入情感度量的机器人在进行人机交互时，其情感友好度与连贯性能够得到有效提升。Abstract: To solve the problems of lack of background knowledge and poor consistency of robot response in the existing human-computer interaction, a human-computer interaction model based on the ripple network of knowledge graph is proposed. In order to achieve a more natural and intelligent human-computer interaction system, this model simulates the real human-human interaction process. Firstly, the human-computer interaction affective friendliness is obtained by calculating the human-computer interaction emotional evaluation value and the human-computer interaction emotional certainty degree. Then, the external knowledge graph is introduced as the background knowledge of robots, and the dialogue entity is embedded into the ripple network of knowledge graph to obtain the potential interested entity content of the participants. Finally, considering the emotional friendliness and content friendliness, the robot response is given. The experimental results show that, compared with other models, robots that have background knowledge and consider emotional friendliness improve emotionality and coherence when interacting with human.
-
表 1 基于知识图谱波纹网络的人机对话模型构建
输入：第k次参与人输入内容$C_{\rm{H}}^k$，已知知识图谱G； 输出：第k+l (l>=1)次机器人的回复内容$C_{\rm{R}}^{k + l}$； (1) Repeat: (2) 根据参与人第k次交互输入返回语义置信度最高的n个回复，并将回复句子向量化得到其特征表示；通过实体连接对对话内容进行实体
提取与消歧；(3) 依据式(2)—式(7)计算第k次参与人的交互输入情感友好度R(k)； (4) 依据式(8)—式(9)得到关联实体的波纹集； (5) 依据式(10)—式(14)得到候选回复的内容响应概率； (6) 依据式(15)得到内容友好度和情感友好度归一化数值，即候选回复的满意度值； (7) 选取满意度最大值yv作为回复内容； (8) $k=k + 1$； (9) Until 参与人停止交互输入； (10) 人机交互会话终止。 
下载: 导出CSV
表 2 Fluency和Sentiment评价标准
Fluency 评价标准 +2 内容相关，语法通顺，符合人类交流方式 +1 内容逻辑勉强相关，语法表达尚可 +0 内容逻辑不相关，答非所问，表达混乱 Sentiment 评价标准 +2 回复的情感上恰当，表达内容生动有趣 +1 回复的情感上恰当 +0 表达模糊，无意义回复如”嗯”，
“呵呵”，“好的”等等
下载: 导出CSV
表 3 不同认知模型的客观评测结果表
模型 MRR MAP Seq2Seq 0.3729 0.4123 ChatterBot 0.4514 0.4871 MECs 0.5820 0.6114 ConceptNet 0.6317 0.6513 本文 0.6492 0.6696
下载: 导出CSV
表 4 志愿者与模型交互轮数与时间统计
模型 平均交互轮数(人) 平均交互时间(s) Seq2Seq 6.350 71.32 ChatterBot 6.850 72.57 MECs 7.925 83.26 ConceptNet 9.850 109.63 本文 11.250 122.45
下载: 导出CSV
表 5 性别组志愿者对各模型Fluency和Sentiment打分统计
模型 Fluency Sentiment 组1 组2 组1 组2 Seq2Seq 0.82 0.75 0.63 0.68 ChatterBot 1.26 1.18 0.84 0.97 MECs 1.47 1.41 1.42 1.46 ConceptNet 1.48 1.46 1.36 1.43 本文 1.55 1.48 1.50 1.53
下载: 导出CSV
表 6 年龄组志愿者对各模型Fluency和Sentiment打分统计
模型 Fluency Sentiment 组3 组4 组3 组4 Seq2Seq 0.68 0.62 0.67 0.74 ChatterBot 1.32 1.27 0.82 0.85 MECs 1.39 1.32 1.50 1.50 ConceptNet 1.50 1.41 1.43 1.49 本文 1.54 1.45 1.54 1.53
下载: 导出CSV
-
[1] RICHMOND J and NELSON C A. Relational memory during infancy: Evidence from eye tracking[J]. Developmental Science, 2009, 12(4): 549–556. doi: 10.1111/j.1467-7687.2009.00795.x [2] 黄宏程, 刘宁, 胡敏, 等. 基于博弈的机器人认知情感交互模型[J]. 电子与信息学报, 2019, 41(10): 2471–2478. doi: 10.11999/JEIT180867HUANG Hongcheng, LIU Ning, HU Min, et al. Cognitive emotion interaction model of robot based on game theory[J]. Journal of Electronics &Information Technology, 2019, 41(10): 2471–2478. doi: 10.11999/JEIT180867 [3] LIU Xin, XIE Lun, and WANG Zhiliang. Empathizing with emotional robot based on cognition reappraisal[J]. China Communications, 2017, 14(9): 100–113. doi: 10.1109/CC.2017.8068769 [4] RODRÍGUEZ L F, GUTIERREZ-GARCIA J O, and RAMOS F. Modeling the interaction of emotion and cognition in Autonomous Agents[J]. Biologically Inspired Cognitive Architectures, 2016, 17: 57–70. doi: 10.1016/j.bica.2016.07.008 [5] NANTY A and GELIN R. Fuzzy controlled PAD emotional state of a NAO robot[C]. 2013 Conference on Technologies and Applications of Artificial Intelligence, Taipei, China, 2013: 90–96. doi: 10.1109/TAAI.2013.30. [6] LOWE R, POW N, CHARLIN L, et al. Incorporating unstructured textual knowledge sources into neural dialogue systems[C]. Machine Learning for Spoken Language Understanding and Interaction, Montreal, Canada, 2015. [7] YOUNG T, CAMBRIA E, CHATURVEDI I, et al. Augmenting end-to-end dialog systems with commonsense knowledge[C]. The 32nd AAAI Conference on Artificial Intelligence, (AAAI-18), New Orleans, USA, 2018: 4970–4977. [8] YANG Liu, QIU Minghui, QU Chen, et al. Response ranking with deep matching networks and external knowledge in information-seeking conversation systems[C]. The 41st International ACM SIGIR Conference on Research & Development in Information Retrieval, New York, USA, 2018: 245–254. doi: 10.1145/3209978.3210011. [9] LIN Yankai, LIU Zhiyuan, SUN Maosong, et al. Learning entity and relation embeddings for knowledge graph completion[C]. The 29th AAAI Conference on Artificial Intelligence, Austin, USA, 2015: 2181–2187. [10] 于洪涛, 丁悦航, 刘树新, 等. 一种基于超节点理论的本体关系消冗算法[J]. 电子与信息学报, 2019, 41(7): 1633–1640. doi: 10.11999/JEIT180793YU Hongtao, DING Yuehang, LIU Shuxin, et al. Eliminating structural redundancy based on super-node theory[J]. Journal of Electronics &Information Technology, 2019, 41(7): 1633–1640. doi: 10.11999/JEIT180793 [11] 丁永刚, 李石君, 付星, 等. 面向时序感知的多类别商品方面情感分析推荐模型[J]. 电子与信息学报, 2018, 40(6): 1453–1460. doi: 10.11999/JEIT170938DING Yonggang, LI Shijun, FU Xing, et al. Temporal-aware multi-category products recommendation model based on aspect-level sentiment analysis[J]. Journal of Electronics &Information Technology, 2018, 40(6): 1453–1460. doi: 10.11999/JEIT170938 [12] PARK J W, KIM W H, LEE W H, et al. How to completely use the PAD space for socially interactive robots[C]. 2011 IEEE International Conference on Robotics and Biomimetics, Karon Beach, Thailand, 2011: 3005–3010. doi: 10.1109/ROBIO.2011.6181762. [13] SIL A and YATES A. Re-ranking for joint named-entity recognition and linking[C]. The 22nd ACM international conference on Information & Knowledge Management, San Francisco, USA, 2013: 2369–2374. doi: 10.1145/2505515.2505601. [14] MILNE D and WITTEN I H. Learning to link with wikipedia[C]. The 17th ACM conference on Information and Knowledge Management, Napa Valley, USA, 2008: 509–518. doi: 10.1145/1458082.1458150. [15] MIKOLOV T, SUTSKEVER I, CHEN Kai, et al. Distributed representations of words and phrases and their compositionality[C]. The 26th International Conference on Neural Information Processing Systems, Lake Tahoe, USA, 2013: 3111–3119. [16] JI Guoliang, HE Shizhu, XU Liheng, et al. Knowledge graph embedding via dynamic mapping matrix[C]. The 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing, Beijing, China, 2015: 687–696. doi: 10.3115/v1/P15-1067. [17] WANG Hongwei, ZHANG Fuzheng, WANG Jialin, et al. Ripplenet: Propagating user preferences on the knowledge graph for recommender systems[C]. The 27th ACM International Conference on Information and Knowledge Management, Torino, Italy, 2018: 417–426. doi: 10.1145/3269206.3271739. [18] SUTSKEVER I, VINYALS O, and LE Q V. Sequence to sequence learning with neural networks[C]. The 27th International Conference on Neural Information Processing Systems, Montreal, Canada, 2014: 3104–3112. [19] GUNTHER C. Tutorial-ChatterBot 1.0.2 documentation[EB/OL]. https://chatterbot.readthedocs.io/en/stable/tutorial.html, 2019. [20] ZHANG Rui, WANG Zhenyu, and MAI Dongcheng. Building emotional conversation systems using multi-task Seq2Seq learning[C]. The 6th CCF International Conference on Natural Language Processing and Chinese Computing, Dalian, China, 2018: 612–621. doi: 10.1007/978-3-319-73618-1_51. -
图(5) / 表(6)
计量
- 文章访问数: 777
- HTML全文浏览量: 564
- PDF下载量: 71
- 被引次数: 0
下载:
