AI in FinTech - Harnessing Unstructured Data
Vidhya Vijayakumar
Business and Technology Program Leadership | Technology and Operational Risk Management | Specialising in Regulatory, Risk and Governance | Equity & Debt Markets, FX and Security Lending
The future of FinTech is here!
And it is with AI!
But…
How to Manage AI Risk? Join in, and let's explore together.
We are now very familiar with the statistics on the amount of data collected daily, as highlighted in AI in FinTech - Data Monetisation Strategies. As a refresher from Infographic: How much data is made every day?, the world produces 2.5 quintillion bytes of data daily. Of the 2.5 quintillion bytes of data generated daily, about 80% (of all data generated) is estimated to be unstructured (Unstructured Data Insights: Key Statistics Revealed).
What is unstructured data?
Unstructured data includes everything that is ever produced, including chat messages, emails, social media posts, documents, audio, videos, weather data, market news, financial reports and sensor data, to name a few. The following chart illustrates the exponential growth in unstructured data.
Despite the abundance of unstructured data, the unstructured data may hold valuable information that is not easy to extract and process to draw meaningful insights. In fact, according to recent statistics, 95% of businesses believe that processing unstructured data poses a significant problem.
Fortunately, unlocking intelligence from the unstructured treasure trove of data has become achievable with the progress made in artificial intelligence, particularly Natural Language Processing (NLP).
NLP in FinTech
Traditional banks and FinTech use NLP to extract meaningful insights from unstructured data. Some of the key areas are:
(1) Sensitivity Analysis - Scan tens of thousands of documents, social media posts, etc, to develop sensitivity analysis.
(2) Customer Analysis - Chat conversations and recorded audio calls can be analysed to understand the common issues experienced by the customers and to reduce or eliminate the pain points.
(3) Fraud Detection - Scanning chat transcripts, emails and other documents to flag non-compliant and suspicious activities, which the experts can follow up.
Key challenges with NLP
The massive influx of unstructured data, which holds immense potential, presents significant challenges. Some of these are as follows:
(1) Data privacy - Unstructured data is likely to contain sensitive information - therefore, there is a need to be cognizant of privacy requirements in various jurisdictions.
(2) Model bias - The models are built at the back of existing data. If the underlying data is biased for any reason, the AI models NLP uses will likely perpetuate the bias.
(3) Integration - any new NLP system developed needs to integrate with the legacy systems used by the banks. If the tool or application is poorly integrated, extracting meaningful value from the models might become challenging.
NLP is rapidly transforming the FinTech industry, and it is essential to jump on the bandwagon to unlock value from unstructured data. Taking advantage of AI, specifically NLP, to harness unstructured data into meaningful insights is critical to staying competitive.
#FinTech #AI #ArtificialIntelligence #DataManagement #UnstructuredData #NaturalLanguageProcessing #NLP #Ethics #ResponsibleAI #EthicalAI #RiskManagement #Compliance #Governance
References
Manning, Christopher D., Mihai Surdeanu, John Bauer, Jenny Finkel, Steven J. Bethard, and David McClosky. 2014. The Stanford CoreNLP Natural Language Processing Toolkit In Proceedings of the 52nd Annual Meeting of the Association for Computational Linguistics: System Demonstrations, pp. 55-60
Did you find the article informative?
Recommended by LinkedIn
♻️If you liked this, you may also find the previous articles in this series interesting. Here are the links:
Technical Support
3wWell put, Vidhya..!! Agree unstructured data is very useful if it's utilised properly..!!
Business and Technology Program Leadership | Technology and Operational Risk Management | Specialising in Regulatory, Risk and Governance | Equity & Debt Markets, FX and Security Lending
3wDid you like the article? ♻️ If you agree, hit repost and help others …?