An overview, empirical application, and discussion of the future research potential of Q&A models in B2B contexts

We present a proof of concept for using automated text analytic techniques to extract key information from lengthy B2B legal documents using BERT (Bidirectional Encoder Representations from Transformers), a machine learning-based natural language-processing framework. Our methodological contributions overcome the text length limitations of applying BERT to long legal documents. We identify franchise disclosure documents (FDDs) as an initial use case for these methodologies and a fruitful avenue for further exploration. From FDDs, we successfully extract answers to questions about firm structure, contractual obligations, finances, and litigation disclosures while also overcoming the technical challenges of applying BERT to large bodies of text. Question-and-answer techniques such as these, deployed in a B2B context, potentially can increase transparency and clarity for prospective exchange partners, addressing concerns in the literature about legal document readability and associated problems of information asymmetry and disclosure misunderstanding. Our discussion identifies promising contexts and an agenda for future scholarship focused on question-and-answer in B2B research.