STTR Phase I: A Technology for Learning to Infer from Unlabeled Financial Data

The broader/commercial impact of this Small Business Technology Transfer (STTR) Phase I project is to enable long-term, above-average profit returns from investing into the U.S. stock market. Currently, investors rely on the financial models that can forecast the future business performance of company only by looking through a rear-view mirror, and, consequently, create risks for the speculative stock trading that does not deliver any actual goods or services. If successful, this STTR project will make an important social impact that is controlling speculation by discouraging stock trading of company without material changes to its valuation. The research and educational impact of the current project is that selected results from the proposed studies beneficial to the fundamental research will be made available to the U.S research community for analysis, data mining, and search and will be also used to enhance contents of undergraduate and graduate courses. The economic impact is that the proposed efforts will create new financial technology related jobs in the North Alabama region. This Small Business Technology Transfer Phase I project will develop and validate innovative technology capable of learning to infer from time series financial data in a resource-scarce environment. This technology will address the following technical hurdles: (a) well-documented deficiencies of machine reasoning of the qualitative parts of financial reports and earning call transcripts containing information that is much richer than just the financial ratios; (b) current reliance of the language-based models including ChatGPT on human annotation in resource-scarce environment, (c) difficulties with transfer learning for extensive, specialized documents, and (d) scarcity of labeled financial text. The technology will be based on innovative algorithms that use language-based models to augment original unlabeled data and utilize such augmented data for predicting the company valuation far ahead of the existing models. The feasibility of the proposed technology will be conducted in collaboration with academic and industrial partners. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.