As crowdfunding has emerged as a popular source of funding for new ventures, budding entrepreneurs often struggle to deliver a convincing pitch to attract backers. We adopt an n-gram language model to analyze narratives of over 21,000 film projects from Kickstarter and find that the choice of words is critical to crowdfunding success. Using penalized logistic regression, we identify the relative power of phrases to predict funding outcome, resulting in a dramatic reduction in error rate. Consistent with the language expectancy theory, the linguistic analyses show that successful projects usually include words that reflect the credibility of project creators and meaningful social interactions, whereas failed projects exude negativism or uncertainty. While good word choices vary among movie genres, words of lower cognitive complexity dampen the chances of funding. These findings have broad implications for text analysis and natural language generation for persuasive marketing communications.

随着众筹已成为新企业的流行资金来源，崭露头角的企业家往往难以提供令人信服的宣传来吸引支持者。我们采用 n-gram 语言模型来分析来自 Kickstarter 的 21,000 多个电影项目的叙述，发现单词的选择对众筹成功至关重要。使用惩罚逻辑回归，我们确定了预测资金结果的短语的相对能力，从而显着降低了错误率。与语言期望理论一致，语言分析表明，成功的项目通常包含反映项目创建者可信度和有意义的社会互动的词，而失败的项目则散发出消极或不确定性。虽然好词选择因电影类型而异，认知复杂度较低的词会抑制获得资助的机会。这些发现对用于有说服力的营销传播的文本分析和自然语言生成具有广泛的影响。