1. 负责人工智能团队研发质量的持续提升,指导软件架构,参与底层算法技术的研发;
2. 负责人工智能在声学,自然语音处理,计算机视觉方向的规划和推进。带领多个研发团队,推进研发成果在公司相关事业部产品的落地;
3. 客服系统已在线使用,主要分为三个模块,意图识别,标准问题匹配,和相似问题匹配。通过精准的匹配来寻找准确回复;
4. 建立了基于影视域的知识图谱系统,将不同的数据源通过爬虫汇集,对数据进行清理,汇总,建立关系,并倒入图数据库中。做出基于知识图谱的问答;
5. 开发可配置的自然语言训练和推理系统,系统包括模型训练网络,基于配置文件的模型训练推理组合,该系统可拓展到不同领域,为快速建模提供系统支持;
6. 研发基于Convmixer和ECAPA-TDNN的语音唤醒和声纹识别的混和模型,在语音增强上,主要研究基于实时,轻量化的降噪模型。

1. 做人工智能和数据的开发,支持公司的数字化转型工作;
2. 参与团队将项目研发成果落地到南美洲电信公司,印尼电信等公司

1. 为约翰迪尔公司建立大数据平台和升级其ERP系统。设计和开发了数据仓库、数据立方体和数据可视化门户,并升级其ERP系统到基于.NET的平台上;
2. 研发一款IBM智能客服的系统, 将知识分为四个预定义的方面:硬件、软件、服务器和服务。每次的人机交流都使机器更接近找到目标知识;
3. 研发辉瑞公司的多语言产品营销系统。团队分布在印度,中国,欧洲和美国四个区域,构建的基于内容的管理系统,使用ES,机器学习等;
4. 为花旗银行开发基于苹果手机端的财富管理平台,项目由UI,SOA和IoS团队组成。该系统为花旗银行的超级客户提供股票、期权交易、银行理财产品购买等服务。

1. 开发支持数百万次浏览的社交网络软件系统。该系统能够平稳地处理高流量。
2. 在保险项目中作为解决方案的架构师,实施和交付了基于.NET的现代化保险平台。
主要代表作
专利:
1. 升降梯安全预警和维护系统及其预警和维护方法梁兴伟 专利号. 2017 I0293407. I
2. 机器设备数据采集方法以及采集系统,发明人:梁兴伟 专利号:201711315280.5
3. 语音情感识别方法和装置, 梁兴伟申请号:PZ2022A0446SD,实审阶段
4. 语音关键词的识别方法、装置和电子设备,梁兴伟申请号:PZ2022A0424SD,实审阶段
5. 基于深度神经网络的说话人验证方法、终端及存储介质,梁兴伟申请号:PZ2022A0329SD,实审阶段
6. 基于深度神经网络的说话人验证方法、终端及存储介质,梁兴伟申请号: PZ2022A0305SD,实审阶段
7. 一种远场场景下说话人身份的验证方法和装置、电子设备,梁兴伟申请号: PZ2022A0902SD,已受理
8. 一种基于RASA的任务型智能多轮对话方法及相关设备,梁兴伟申请号: PZ2022A0744SD,实审阶段
9. 对话处理方法、装置和电子设备及存储介质,梁兴伟申请号: PZ2022A0626SD,初审合格
论文:
1. Liang, Xingwei, Jiacheng Du, Taiyu Niu, Lanjun Zhou, and Ruifeng Xu. 2023. "Knowledge Interpolated Conditional Variational Auto-Encoder for Knowledge Grounded Dialogues" Applied Sciences 13, no. 15: 8707. https://doi.org/10.3390/app13158707
2. Liang, Xingwei., Zhang, Z. & Xu, R. Multi-task deep cross-attention networks for far-field speaker verification and keyword spotting. J AUDIO SPEECH MUSIC PROC. 2023, 28 (2023). https://doi.org/10.1186/s13636-023-00293-8
3. Liang, Xingwei, You Zou, and Ruifeng Xu. "SI-LSTM: Speaker Hybrid Long-short Term Memory and Cross Modal Attention for Emotion Recognition in Conversation." arXiv preprint arXiv:2305.03506 (2023). Paper Source Code
4. Liang, Xingwei, You Zou, Xinnan Zhuang, Jie Yang, Taiyu Niu, and Ruifeng Xu. "MMATERIC: Multi-Task Learning and Multi-Fusion for AudioText Emotion Recognition in Conversation." Electronics 12, no. 7 (2023): 1534.2. (SCI,IF 2.69,JCR Q3) Paper Source Code
5. Li, Ang, Xingwei Liang, Miao Zhang, Bingbing Wang, Guanrong Chen, Jun Gao, Qihui Lin, and Ruifeng Xu. "A Coarse-to-Fine Text Matching Framework for Customer Service Question Answering." In Cognitive Computing–ICCC 2022: 6th International Conference, Held as Part of the Services Conference Federation, SCF 2022, Honolulu, HI, USA, December 10-14, 2022, Proceedings, pp. 39-53. Cham: Springer Nature Switzerland, 2023.(EI, Joint First Author) Paper Source Code
6. Wang, Jun, Qianlong Wang, Zhiyuan Wen, Xingwei Liang, and Ruifeng Xu. "Interactive Fusion Network with Recurrent Attention for Multimodal Aspect-based Sentiment Analysis." In Artificial Intelligence: Second CAAI International Conference, CICAI 2022, Beijing, China, August 27–28, 2022, Revised Selected Papers, Part III, pp. 298-309. Cham: Springer Nature Switzerland, 2022.(EI) Paper Source Code
7. Liang, Xingwei, You Zou, Tian Xie, and Qi Zhou. "DCRNNX: Dual-Channel Recurrent Neural Network with Xgboost for Emotion Identification Using Nonspeech Vocalizations." In Artificial Intelligence and Mobile Services–AIMS 2022: 11th International Conference, Held as Part of the Services Conference Federation, SCF 2022, Honolulu, HI, USA, December 10–14, 2022, Proceedings, pp. 18-30. Cham: Springer International Publishing, 2022. (EI) Paper Source Code
8. Li, Zhenyu, Geng Tu, Xingwei Liang, and Ruifeng Xu. "Developing Relationships: A Heterogeneous Graph Network with Learnable Edge Representation for Emotion Identification in Conversations." In Artificial Intelligence: Second CAAI International Conference, CICAI 2022, Beijing, China, August 27–28, 2022, Revised Selected Papers, Part III, pp. 310-322. Cham: Springer Nature Switzerland, 2022.(EI,Corresponding Author) Paper
9. Zhou, Xia-Bing, Zhong-Qing Wang, Xing-Wei Liang, Min Zhang, and Guo-Dong Zhou. "Neural Emotion Detection via Personal Attributes." Journal of Computer Science and Technology 37, no. 5 (2022): 1146-1160.