项目描述对网络借贷投资决策的影响：多维语言特征分析视角

The description of a project promotes information disclosure, relieves information asymmetry, and then influences the perception of risk, and at last improves lending efficiency. Therefore, in the perspective of peer-to-peer lending, the influence mechanism of project description on investment decision-making is explored, following the lead of ‘acceptance of textual information -> extraction of linguistic features -> perception of transaction exposure -> explanation of investment behavior’, based on the three paradigms of behavior science, information technology and econometrics. In the perspective of behavior science, the acceptance of project description of an investor is analyzed to expose the impact of multisource distributivism on the investor’s behavior of information adoption. In the perspective of information technology, linguistic features of project description are identified to understand information in different angles, and based on which individual credit assessment is improved and risk perception is corrected. In the perspective of econometrics, an econometric model of the influence of multidimensional linguistic features on an investor’s willingness to lend is established with risk perception playing the moderate role, in order to explain the influence mechanism of project description on investment decision-making. At last, analysis of linguistic features is brought into the recommendation system of lending objects, so as to reduce the risk of investment. This research enriches the system of value discovery of project description in the study of peer-to-peer lending in theory, and provides valuable suggestions to improve lending efficiency and reduce investors’ financial risks.

项目描述能够促进信息披露，缓解借贷双方的信息不对称，进而影响风险感知，最终提升借贷效率。为此，基于网络借贷情境，以项目描述为数据，综合行为科学、信息技术、计量分析三种范式，沿着“文本信息的采纳->语言特征的提取->交易风险的感知->投资行为的解释”主线，探究项目描述对投资决策的影响机理。从行为角度，分析投资人对项目描述的采纳意愿，揭示多源分布性对投资者信息行为的影响；从技术层面，通过文本挖掘，识别项目描述的语言特征，实现多维度的信息解读。在此基础上，完善个人征信评估，修正风险感知；从计量分析角度，以感知风险为中介，构建多维语言特征影响投资人出借意愿的计量模型，解释项目描述对投资决策的影响机理。最后，将语言特征与传统的推荐算法相结合，提出借款标的推荐机制，进一步降低投资风险。理论上，丰富了网络借贷环境下文本描述的价值发现体系，实践上，对提升借贷交易效率和降低金融风险提供了有益指导。

在线众筹和网络借贷是互联网金融的重要形式。在网络融资平台上，项目描述能够促进信息披露，缓解投融资双方的信息不对称，进而影响风险感知，最终能够确保融资成功率和提升借贷效率。为此，基于在线众筹和网络借贷情境，以项目描述为实验数据，综合行为科学、信息技术、计量分析三种范式，沿着“文本信息的采纳->语言特征的提取->交易风险的感知->投资行为的解释”主线，探究项目描述对投资决策的影响机理。.（1）揭示了投资人在决策过程中对多源项目描述的差异性采纳机理。利用信息接受理论、技术采纳模型、信任理论等，探究多源分布的项目描述对投资人信息行为的影响，从而揭示投资人在出借决策过程中，针对不同类型借款标的而采取不同来源信息的影响机理，为合理设计和布局项目描述性信息提供理论依据。（2）识别了借款标的描述性信息的多维度语言特征。借鉴语言学和行为科学相关理论，从说服性、情感性、主客观性、导向性、欺诈性、易读性、交互性等维度，定义项目描述的语言特征。运用文本分析技术，提出面向借款描述的语言特征识别与量化方法，实现了借款标的全方位解读。（3）解释了项目描述的多维语言特征对投资人出借行为的影响。基于项目描述的多维语言特征，洞察借款人及其标的背景信息，以此修正借款人信用等级，并预测逾期违约风险。在此基础上，结合传统的投资理论建立计量分析模型，从借款成功率、借款时效、借款成本等方面，揭示了基于项目描述的信息披露机制对投资人出借行为的影响。.理论上，丰富了在线众筹和网络借贷环境下文本描述的价值发现体系，实践上，对提升借贷交易效率和降低金融风险提供了有益指导。