*The KDD '23 Workshop on Robust NLP for Finance (RobustFin)*
Financial services is a document-rich industry. This has led to great demand for advanced NLP techniques to process financial corpora in order to curate knowledge, augment human intelligence, and support industry professionals in making informed business decisions. Recent advances in the NLP domain have motivated the emergence of key applications such as the automation of due diligence protocols, financial question answering (QA) and reasoning, conversational AI in finance, summarizing business entity earnings calls, and analyzing investment sentiment of social media posts. Due to the unique industry standards and regulations, robustness is a fundamental consideration underlying nearly all financial NLP applications. The concept of robsustness may connote many specific definitions and real-life considerations. In finance, the key concerns of robustness refer to better generalizability, higher explainability, lower risk of catastrophic failure in unforeseen circumstances, and diligent fairness considerations. Common NLP models like large language models trained on generic corpora usually deteriorate when processing financial texts despite having achieved state-of-the-art results in other applications. The business and regulatory requirements for explainability and the reasoning behind decision makings for the financial industry are often uncompromising, while many NLP modeling techniques still lack this characteristic. Based on years of research experience in this domain, the organizers recapitulate the key aspects of these challenges: massive volume of data, large variation in the data format, low signal-to-noise ratio, scarcity of expert annotated datasets, task ambiguity, challenges regarding data integrity and privacy, domain shift, and high performance requirements set by industry and regulatory standards. While these challenges have been studied in many generic contexts, we believe working on this vertical high-stakes domain, i.e. finance, will help consolidate the development of solutions. The workshop welcomes original research submissions related to robustness and explainability research on NLP, with a focus on the financial domain and/or financial corpora, which will include, but not be limited to, the topics listed below. The workshop will welcome both position and regular research papers.Topics of Interest
The topics of the workshop include, but are not limited to, the following areas: · Adversarial attacks and defense to improve NLP model robustness and privacy protection.· Transfer learning applications and their robustness and explainability studies.· Robustness research against concept drift, domain shift, and noisy datasets; NLP model generalization.· Language modeling on financial corpora including tabular and numerical data, and multi-modal modeling; numeracy and quantitative reasoning, fact verification, and QA over tabular data.· Reconciling structured and unstructured knowledge to improve model interpretability; financial knowledge graph construction.· Few-shot learning and prompt methods.· Synthetic or genuine financial textual datasets and benchmarks.· Reasoning and textual entailment in financial applications.· Empirical studies on robust NLP modelsSubmission GuidelinesWe invite submissions of relevant work that be of interest to the workshop. All submissions must be original contributions that have not been previously published and that are not currently under review by other conferences or journals. Submissions will be peer reviewed, single-blinded. Submissions will be assessed based on their novelty, technical quality, significance of impact, interest, clarity, relevance, and reproducibility. All submissions must be in PDF format and follow the current ACM two-column conference format https://www.acm.org/publications/proceedings-template. We accept two types of submissions:· full research paper: no longer than 9 pages (including references, proofs, and appendixes).· short/poster paper: no longer than 4 pages (including references, proofs, and appendixes).Submission will be accepted via Microsoft CMT https://cmt3.research.microsoft.com/RobustFin2023. All accepted papers will be presented in the workshop. At least one author of each accepted submission must attend the workshop to present their work.Important Dates
- Paper abstract due: May 16, 2023 AoE - Paper submission due: June 4, 2023 AoE - Submission notification: June 23, 2023 - Workshop: August 7-8, 2023
Organizing Committee· Sameena Shah - JPMorgan AI Research· Xiaodan Zhu - Queen's University· Gerard de Melo - Hasso Plattner Institute and the University of Potsdam· Armineh Nourbakhsh - JPMorgan AI Research· Xiaomo Liu - JPMorgan AI Research· Zhiqiang Ma - JPMorgan AI Research· Charese Smiley - JPMorgan AI Research· Zhiyu Chen - MetaWorkshop Website
https://robustfin.github.io/2023/
*Contact*
For general inquiries about the workshop, please write to the organizers at robustfin.workshop@gmail.com.