Ciro D. Esposito, PhD

Detecting emerging topics is crucial for understanding research trends, technological advancements, and shifts in public discourse. While unsupervised topic modeling techniques such as Latent Dirichlet allocation (LDA), BERTopic, and CoWords clustering are widely used for topic extraction, their ability to retrospectively detect emerging topics without relying on ground truth labels has not been systematically compared. This gap largely stems from the lack of a dedicated evaluation metric for measuring emergence detection. In this study, we introduce a quantitative evaluation metric to assess the effectiveness of topic models in detecting emerging topics. We evaluate three topic modeling approaches using both qualitative analysis and our proposed emergence detection metric. Our results indicate that, qualitatively, CoWords identifies emerging topics earlier than LDA and BERTopics. Quantitatively, our evaluation metric demonstrates that LDA achieves an average F1 score of 80.6% in emergence detection, outperforming BERTopic by 24.0%. These findings highlight the strengths and limitations of different topic models for emergence detection, while our proposed metric provides a robust framework for future benchmarking in this area.

This PhD dissertation explores the role of emerging fields, such as Artificial Intelligence (AI), for two key strategic decisions—entrepreneurial resource acquisition and alliance formation—and examines how these fields can be identified using computational methods. Chapter 2 examines the effects of being associated with an emerging field on initial venture financing from both scientists’ and investors’ perspectives. We demonstrate that increasing scientific involvement in a field positively influences capital raised, while increased investor involvement has a negative effect. This contrast likely arises because scientific interest signals innovation potential, whereas greater investor presence may indicate that the most lucrative opportunities have already been exploited. Chapter 3 investigates how the novelty of these fields—both to the firm making a decision and to its competitors—affects alliance formation. We further explore how the scientific background of alliance decision-makers shapes these strategic choices. Our findings show that only fields familiar to competitors of the focal firm positively impact alliance formation, while decision-makers with a scientific background tend to favor more novel solutions. Chapter 4 assesses the effectiveness of various computational methods in identifying fields, introducing a novel quantitative measure of overlap to compare their results. Our findings reveal that each method yields different insights, highlighting the need for careful interpretation as certain outputs may lack substantive meaning, necessitating human judgment. Chapter 5 extends traditional metrics, commonly applied to track topics over time within a single method, by using them to bridge topics, including fields, across different methods. We also assess the performance of traditional versus neural topic modeling in detecting emergence. Our results show that content-based metrics are particularly effective for cross-method matching, with traditional topic modeling demonstrating greater accuracy than neural topic modeling in identifying emerging fields.

Prior studies show that access to academic knowledge plays a crucial role in new venture financing. We extend this research by shifting the focus from the access to academic knowledge to the developmental state of the academic field, where the academic knowledge is generated. Using natural language processing (NLP), we clustered peer-reviewed academic knowledge from Scopus into various fields. We then analyzed a sample of 341 new biotech ventures from Crunchbase to determine if increased past activity by (1) academics and (2) early-stage venture investors in a particular academic field is associated with the early-stage equity financing of new ventures associated with that field. We found that new ventures associated with academic fields for which academic activity has grown in the past receive more early-stage equity capital. However, contrary to our expectations, we also revealed that when a particular academic field shows greater early-stage venture investments in the past, the amount of early-stage equity capital received by subsequent ventures associated with the same academic field decreases. This suggests that while emerging academic fields signal the presence of business opportunities with high reward potential, past increase in the number of investments by peer early-stage investors associated with a particular academic field signals the opposite.