企业包容性创新的测度:基于大语言模型的新方法  

Measurement of Enterprise Inclusive Innovation:New Methods Based on Large Language Model

在线阅读下载全文

作  者:梁宇奇 赵云辉 付星星 赵瑞鹏 LIANG Yu-qi;ZHAO Yun-hui;FU Xing-xing;ZHAO Rui-peng

机构地区:[1]内蒙古财经大学工商管理学院

出  处:《科学决策》2024年第12期86-99,共14页Scientific Decision Making

基  金:国家自然科学基金地区项目(72464024);内蒙古自治区杰出青年基金项目(2024JQ19);内蒙古自治区高等学校创新团队发展计划项目(NMGIRT2202)。

摘  要:包容性创新作为企业面向BoP市场进行多元价值创造的全新创新形态,引发学术界的广泛关注。但目前尚未有一种客观、高效且准确的方法测度包容性创新。为解决这一问题,提出一套基于大语言模型的新测度方法。通过构建关键词词典,对2013-2022年31549份上市公司年度报告中包含关键词的有效语句进行人工标记,使用标记结果微调和训练BERT预训练模型,选择训练过程中分类性能最好的模型对全部文本进行分类,最终构建了企业包容性创新指标。相关性和有效性检验结果进一步表明,企业包容性创新具有明显的地区分布特点、行业差异性,与国家政策高度相关,能够较准确地反映企业包容性创新指标构建的合理性。研究成果对于推动企业包容性创新领域研究具有重要意义。Inclusive innovation,as an emerging form of enterprise innovation aimed at creating multifaceted value for the Base of the Pyramid(BoP)market,has garnered significant attention within the academic community.However,there remains a lack of an objective,efficient,and precise method to measure this crucial indicator.To address this gap,a keyword dictionary was constructed,and effective statements containing these keywords were manually labeled from 31,549 annual reports of listed companies spanning from 2013 to 2022.Using the labeled data,the BERT pre-training model was fine-tuned and trained.Then the model exhibiting the highest classification performance during the training process was selected to categorize all texts.Consequently,an enterprise inclusive innovation measurement index was developed.Further correlation and effectiveness tests revealed that enterprise inclusive innovation exhibits distinct regional distribution patterns and industry-specific variations,and is strongly associated with national policies.These findings provide robust validation for the rationality of the constructed enterprise inclusive innovation indicators.The research outcomes are of great significance for promoting research in the domain of enterprise inclusive innovation.

关 键 词:企业包容性创新 金字塔底层(BoP)战略 大语言模型 BERT预训练模型 

分 类 号:F272[经济管理—企业管理]

 

参考文献:

正在载入数据...

 

二级参考文献:

正在载入数据...

 

耦合文献:

正在载入数据...

 

引证文献:

正在载入数据...

 

二级引证文献:

正在载入数据...

 

同被引文献:

正在载入数据...

 

相关期刊文献:

正在载入数据...

相关的主题
相关的作者对象
相关的机构对象