基于双通道理论的通信认知增强技术研究

doi:10.11959/j.issn.1000-0801.2024007

电信科学 ›› 2024, Vol. 40 ›› Issue (1): 123-135.doi: 10.11959/j.issn.1000-0801.2024007

• 研究与开发 • 上一篇

基于双通道理论的通信认知增强技术研究

廉霄兴, 宋勇, 朱军, 王淑玲, 叶晓舟, 欧阳晔

亚信科技（中国）有限公司，北京 100193

修回日期:2024-01-10 出版日期:2024-01-01 发布日期:2024-01-01
作者简介:廉霄兴（1990- ），男，亚信科技（中国）有限公司数据智能规划高级专家，主要研究方向为人工智能算法、大数据及数字孪生技术
宋勇（1989- ），男，亚信科技（中国）有限公司算法工程师，主要研究方向为行业大模型开发应用、自然语言处理开发
朱军（1972- ），男，博士，亚信科技（中国）有限公司高级产品总监、首席架构师，主要研究方向为数据智能产品规划、大数据平台架构设计、企业数智化转型
王淑玲（1988- ），女，博士，亚信科技（中国）有限公司研发中心规划部规划总监，主要研究方向为网络通信、云网融合
叶晓舟（1980- ），男，博士，亚信科技（中国）有限公司通信人工智能实验室资深总监、首席科学家，主要研究方向为通信网络与人工智能
欧阳晔（1981- ），男，博士，亚信科技（中国）有限公司首席技术官、高级副总裁，主要研究方向为移动通信、人工智能、数据科学、科技研发创新与管理

Research on dual process theory based telecom augmented cognition

Xiaoxing LIAN, Yong SONG, Jun ZHU, Shuling WANG, Xiaozhou YE, Ye OUYANG

AsiaInfo Technologies (China) Co., Ltd., Beijing 100193, China

Revised:2024-01-10 Online:2024-01-01 Published:2024-01-01

摘要/Abstract

摘要：

结合当前大语言模型增强技术研究进展，提出了基于双通道理论的通信认知增强技术框架（telecom augmented cognition architecture，TACA），解决大语言模型在通信领域应用中专业知识不足、推理能力有限等问题，实现面向通信复杂问题的认知推理增强、知识融合增强、模型工具增强和自主决策执行增强等应用能力。针对不同的通信业务场景给出了相应技术应用参考，并通过实验初步验证认知增强方案在解决复杂问题方面的有效性和可行性。因此，所提出的通信认知增强技术框架可作为未来通信行业大语言模型应用的参考框架。

关键词: 大语言模型, 认知增强, 双通道理论, 自主决策执行, 电信网络

Abstract:

Incorporating the current works in augmented large language models (LLM), a dual process theory based telecom augmented cognition architecture (TACA) was proposed.This architecture addressed challenges faced by LLM in the field of telecom, such as lack of domain-specific knowledge and limited reasoning capabilities.It aims to enhance cognitive reasoning, knowledge integration, tooling, and autonomous decision-execution capabilities for complex telco-scenarios.Technical application references for different telco-scenarios were provided, and the effectiveness and feasibility of the augmented cognition solution were preliminarily validated in addressing complex problems through experiments.The proposed TACA can serve as a future reference architecture for the application of LLMs in the telecom industry.

Key words: large language model, augmented cognition, dual process theory, autonomous decision and execution, telecommunication network

中图分类号:

TP181

廉霄兴, 宋勇, 朱军, 王淑玲, 叶晓舟, 欧阳晔. 基于双通道理论的通信认知增强技术研究[J]. 电信科学, 2024, 40(1): 123-135.

Xiaoxing LIAN, Yong SONG, Jun ZHU, Shuling WANG, Xiaozhou YE, Ye OUYANG. Research on dual process theory based telecom augmented cognition[J]. Telecommunications Science, 2024, 40(1): 123-135.

图/表 10

图1

图2

图3

图4

图5

图6

图7

图8

表1

表2

参考文献 23

[1]	张嗣宏, 张健 . 以 ChatGPT 为代表的生成式 AI 对通信行业的影响和应对思考[J]. 电信科学, 2023,39(5): 67-75.
	ZHANG S H , ZHANG J . Impact and countermeasures of generative AI represented by ChatGPT on the telecom industry[J]. Telecommunications Science, 2023,39(5): 67-75.
[2]	Lexico. Oxford university press and dictionary.com “cognition”[EB]. 2020.
[3]	王昊奋, 王萌 . “神经 + 符号”：从知识图谱角度看认知推理的发展[J]. 中国计算机学会通讯, 2020,16(8): 52-56.
	WANG H F , WANG M . Neuro-symbolic:the development of cognitive reasoning from the perspective of knowledge graphs[J]. Communications of the CCF, 2020,16(8): 52-56.
[4]	ASHISH V , NOAM S , NIKI P ,et al. Attention is all you need[C]// NIPS2017:Proceedings of the 31st International Conference on Neural Information Processing Systems. California:Curran Associates Inc, 2017: 6000-6010.
[5]	桑基韬, 于剑 . 从ChatGPT看AI未来趋势和挑战[J]. 计算机研究与发展, 2023,60(6): 1191-1201.
	SANG J T , YU J . ChatGPT:a glimpse into AI’s future[J]. Journal of Computer Research and Development, 2023,60(6): 1191-1201.
[6]	JONATHAN S B E . In two minds:dual-process accounts of reasoning[J]. Trends in Cognitive Sciences, 2003,7(10): 454-459.
[7]	KAHNEMAN D . Thinking,fast and slow[M]. New York: Farrar,Straus and Giroux, 2011.
[8]	BUBECK S , CHANDRASEKARAN V , ELDAN R ,et al. Sparks of artificial general intelligence:early experiments with GPT-4[EB]. arXiv preprint,2023,arXiv:2303.12712.
[9]	YAN N . A path towards autonomous machine intelligence[EB]. 2022.
[10]	MAXWELL N , MICHAEL H T , JOSHUA B ,et al. Improving coherence and consistency in neural sequence models with dual-system,neuro-symbolic reasoning[J]. Advances in Neural Information Processing Systems, 2021: 25192-25204.
[11]	HARSH T , NIRANJAN B , TUSHAR K ,et al. Interleaving Retrieval with chain-of-thought reasoning for knowledge-I ntensive multi-step questions[C]// ACL2023:Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics. Toronto:Association for Computational Linguistics, 2023: 10014-10037.
[12]	YAO S , YU D , ZHAO J ,et al. Tree of thoughts:deliberate problem solving with large language models[J]. arXiv preprint arXiv:2305.10601, 2023.
[13]	ZHOU A , YAN K , SHLAPENTOKH-ROTHMAN M , ,et al. Language agent tree search unifies reasoning acting and planning in language models[J]. arXiv preprint arXiv:2310.04406, 2023.
[14]	YORAN O , WOLFSON T , BOGIN B ,et al. Answering questions by meta-reasoning over multiple chains of thought[J]. arXiv preprint arXiv:2304.13007, 2023.
[15]	BESTA M , BLACH N , KUBICEK A ,et al. Graph of thoughts:solving elaborate problems with large language models[J]. arXiv preprint arXiv:2308.09687, 2023.
[16]	SHEN Y L , SONG K T , TAN X ,et al. HuggingGPT:solving AI tasks with ChatGPT and its friends in hugging face[J]. arXiv preprint arXiv:2303.17580, 2023.
[17]	GE Y Q , HUA W Y , MEI K ,et al. OpenAGI:when LLM meets domain experts[J]. arXiv:2304.04370, 2023.
[18]	ETSI. Zero-touch network and service management (ZSM) (Version 2.1.1)[S]. 2023.
[19]	ETSI. Generic autonomic network architecture (GANA) (version 1.1.1)[S]. 2020.
[20]	TMF. Autonomous networks whitepaper (3.0)[R]. 2021.
[21]	刘贤松, 陈凌, 尹俊 . 通信行业语言大模型技术和应用研究[J]. 邮电设计技术, 2023(7): 14-19.
	LIU X S , CHEN L , YIN J . Research on large language model technology and application in communication industry[J]. Designing Techniques of Posts and Telecommunications, 2023(7): 14-19.
[22]	毛志伟, 邢向晖, 孙广生 . 浅谈AIGC在通信设计领域的应用[J]. 邮电设计技术, 2023(7): 25-30.
	MAO Z W , XING X H , SUN G S . Discussion on application of AIGC in communication design field[J]. Designing Techniques of Posts and Telecommunications, 2023(7): 25-30.
[23]	李涛, 王春佳, 李姗姗 . 电信网络智能化方案研究[J]. 电信科学, 2023,39(3): 162-172.
	LI T , WANG C J , LI S S . Research on intelligent scheme of telecommunication network[J]. Telecommunications Science, 2023,39(3): 162-172.

序号	检索排序	Recall@1	Recall@5	Recall@10	Recall@20
1	BM25+Hanlp分词检索	39.90%	86.64%	95.16%	97.46%
2	BM25+GanymedeNil_text2vec-large-chinese	40.66%	87.10%	95.61%	97.46%
3	Simcse-chinese-roberta-wwm-ext	35.31%	80.11%	91.61%	93.61%
4	GanymedeNil_text2vec-large-chinese	36.53%	84.8%	93.28%	95.92%
5	Hfl_chinese-roberta-wwm-ext	32.46%	76.98%	90.27%	92.74%

序号	大模型推理方案	ChatGPT	ChatGLM-130B	Vicuna-13B	ChatGLM-6B	ChatGLM2-6B
1	MHR + R＆R(TACA)	0.954	0.872	0.549	0.653	0.764
2	SSR + R＆R	0.913	0.840	0.518	0.645	0.749
3	SSR	0.425	0.348	0.205	0.277	0.307

基于双通道理论的通信认知增强技术研究

Research on dual process theory based telecom augmented cognition

在线阅读

PDF下载

可视化

摘要/Abstract

引用本文

使用本文

图/表 10

参考文献 23

相关文章 12

Metrics

推荐阅读 0

[1]	周胜利, 阮琳琦, 徐睿, 张熙康, 赵泉喆, 连远博. 基于关联规则特征提取的网络行为被害性识别集成优化模型[J]. 电信科学, 2023, 39(9): 129-140.
[2]	周胜利, 蒋可怡, 徐博, 徐睿, 张熙康, 赵泉喆, 徐阳东. 面向电信网络诈骗治理的网络安全课程构建效能评估研究[J]. 电信科学, 2023, 39(6): 122-128.
[3]	李涛, 王春佳, 李姗姗. 电信网络智能化方案研究[J]. 电信科学, 2023, 39(3): 162-172.
[4]	赵冰,刘志民,袁益明,韩涛. SoftCOM AI，自动驾驶电信网络解决方案[J]. 电信科学, 2019, 35(4): 103-113.
[5]	林原,王彦斌,张辉,杨家海. 电信网络资源建模方法[J]. 电信科学, 2016, 32(1): 112-118.
[6]	黄挺,谢文旭,王燕川. 云计算在电信网络管理系统中的应用与实践[J]. 电信科学, 2011, 27(4): 119-122.
[7]	黄元飞,古利勇,金丽萍. 网络信息安全形势及相关热点问题探讨[J]. 电信科学, 2009, 1(2): 16-20.
[8]	黄元飞,古利勇,金丽萍. 网络信息安全形势及相关热点问题探讨[J]. 电信科学, 2009, 25(2): 16-20.
[9]	王光全,王宏伟. 利用RF ID技术实现电信智能配线管理[J]. 电信科学, 2007, 23(9): 99-101.
[10]	莫昕玮,朱王奇. 电信网络运行维护集中化模式研究[J]. 电信科学, 2006, 22(4): 69-72.
[11]	刘谦. 基于电信网络的家庭网络标准研究进展[J]. 电信科学, 2006, 22(3): 4-7.
[12]	. 电信网络知识产权管理与保护对策[J]. 电信科学, 2004, 20(7): 70-72.