Over the past decade, computer scientists have developed a variety of machine learning (ML) models that can quickly and efficiently analyze large amounts of data. However, these models should protect users’ privacy and prevent information from being accessed by third parties or developed by developers in order to apply it in real-time situations involving highly sensitive data analysis.
Researchers at the Manipal Institute of Technology, Carnegie Mellon University and Yildes Technical University have recently developed a privacy-enabled model for financial text analysis and classification. Introduced in a previously published paper in ArXiv, this model is based on a collection of natural language processing (NLP) and machine learning methods.
Priam Basu, a researcher who conducted the study, said, “Our paper was based on our previous work, ‘Differential Privacy and Federal Learning for Bench Models.’ Tech Explore. “This book is our humble attempt to integrate both natural language processing (NLP) and machine learning that protect privacy.”
The main objective of the recent work by Basu and his colleagues was to develop an NLP model that would protect the privacy of users and prevent their data from being accessed by others. Such a model can be particularly useful in analyzing bank statements, tax returns, and other sensitive financial documents.
“Machine learning is primarily data-based and gives you understanding and predictions and information based on data,” Basu said. “Therefore, it is very important for us to do research on how to protect user privacy at once.”
The framework developed by Basu and his colleagues is based on two approaches, known as differential privacy and federal learning, combined with bilateral symbol representations from the popular and widely used NLP models Transformer (Bert). Differential privacy technologies add some volume to the data fed into the model. As a result, the data processing party (eg developers, technology companies, or other companies) cannot access real documents and data because the individual element is hidden.
“Federation Learning, on the other hand, is a way of training a model of multiple decentralized devices so that no device can have complete data access at once,” Basu explained. “Bert is a language format that provides contextual embedding for natural language text that can then be used for multitasking, such as classification, sequencing, and semantics.”
Basu and his colleagues used the tactics they developed to train several NLP models for financial text classification. They then evaluated these models in a series of experiments, where they used them to analyze data from the financial phrase bank database. They found that NLP models, as well as other advanced technologies, were implemented for financial analysis, ensuring greater data security, and the results were very promising.
The study by these researchers could have significant implications for several industries, including the financial sector and other areas including sensitive user data analysis. In the future, the new models they have developed will significantly enhance the privacy associated with NLP technology that analyzes personal and financial information.
“Classification and classification based on natural language data is used in many domains and, therefore, we have provided a way to maintain the privacy of user data, which is very important when using money, and the data used therein is highly sensitive and confidential. , “Said Basu. “We now plan to improve the accuracy achieved by our model without incurring significant losses from private trading. Analysis and clustering using FL. ”
Privacy Enabled Activation of Financial Text Classification using differential privacy and federal learning. arXiv: 2110.01643 [cs.CL]. arxiv.org/abs/2110.01643
Privacy and federal learning standardization of differentials for Bert models. arXiv: 2106.13973 [cs.CL]. arxiv.org/abs/2106.13973
© 2021 Science X Network
Excerpt: Financial Text Classification Example (2021, October 13) Retrieved October 13, 2021 from https://techxplore.com/news/2021-10-finanagement-texts-users-privacy.html while protecting users’ privacy.
This document is subject to copyright. No part of it may be reproduced without written permission, except in fairness for personal study or research. Content is provided for informational purposes only.