Market Sentiment Analysis and Its Impact on Crowdfunding Success Using Deep Language Model–Based Natural Language Processing

Abstract

The objective of this study was to examine the effect of market sentiment extracted through deep language model–based natural language processing on the success of crowdfunding campaigns by analyzing the relationship between linguistic sentiment indicators, investor engagement, and funding performance outcomes. This applied quantitative study employed an explanatory–predictive research design using data collected from 312 crowdfunding campaigns launched by entrepreneurs in Tehran between 2022 and 2024. Textual data including campaign descriptions, creator updates, investor comments, and interaction records were extracted from crowdfunding platforms and preprocessed using Persian natural language processing techniques. Market sentiment was operationalized through deep learning–based transformer models fine-tuned for contextual sentiment detection, enabling extraction of emotional tone, trust signals, perceived risk, and collective optimism. Crowdfunding success was measured using funding achievement ratio, funding speed, investor participation level, and success probability. Data analysis combined descriptive statistics, correlation analysis, multivariate regression modeling, structural path analysis, and machine learning prediction methods including gradient boosting, random forest, and deep neural networks to evaluate predictive relationships between sentiment variables and funding outcomes. Inferential analyses revealed that market sentiment significantly predicted crowdfunding success, demonstrating strong positive effects on investor engagement, funding achievement, and overall campaign success probability. Structural modeling confirmed that investor engagement partially mediated the relationship between sentiment and funding performance. Regression results indicated that sentiment was the strongest explanatory variable among all predictors, while machine learning comparisons showed that deep neural network models achieved the highest predictive accuracy in classifying successful campaigns. Positive sentiment accelerated funding speed and strengthened collective participation dynamics, confirming that emotional and linguistic signals function as behavioral drivers influencing investment decisions under uncertainty. The findings indicate that crowdfunding markets operate as sentiment-sensitive digital ecosystems in which collective emotional perception, communicated through textual interaction, significantly shapes financial outcomes.