The London Stock Exchange Group (LSEG) is ramping up innovation efforts, tapping into technologies including artificial intelligence (AI), machine learning (ML), natural language processing (NLP), and APIs to provide finance professionals with more accurate market intelligence, deeper insights and facilitate regulatory compliance.
These efforts are being undertaken by the newly announced LSEG Labs, a network of innovation labs that work with the group’s customers and partners to apply strategic analysis, emerging technology, data science and design thinking to solve pressing problems faced by the financial industry.
Out of the seven projects currently being undertaken, three are newer ones that the group is putting a particular focus on. These three projects apply ML, NLP, and APIs to extract critical insights from unstructured data, offer more accurate market intelligence, and enable professionals to make better-informed decisions.
Here we take a closer look at three key projects in LSEG Labs.
Financial Language Modelling: Powering NLP in Finance
The Financial Language Modelling project seeks to create new NLP models that have a better understanding of financial language.
The purpose here is to create finance-domain specific models that produce more accurate word embeddings, and, ultimately, improve the performance of downstream tasks such as text classification, topic modelling, auto summarization and sentiment analysis.
So far, two models have been built upon Google’s BERT-BASE architecture, leveraging LSEG’s unstructured financial data: BERT-RNA (pre-trained using Reuters News Archive), which has already delivered better accuracy levels for tasks like classifying financial news for ESG controversies, and identifying news related to COVID-19 as either a risk or opportunities; and BERT-TRAN, which was pre-trained using a large corpus of earnings call transcripts.
Both models are available on the Refinitiv Data Platform. LSEG Labs are also giving a small group of customers early access to use their new models via a test user interface which include tutorials, example training data and use-cases.
Global Infrastructure API: Timely and Accurate Market Intelligence
The Global Infrastructure API project aims to build a global infrastructure database with NLP and knowledge graphs. The goal is to provide investors with more timely and accurate market intelligence, and deeper insights across active and potential investment opportunities.
For this project, LSEG has collaborated with industry participants including investment management firms and sovereign wealth funds to create a new method to link and enrich six Refinitiv infrastructure datasets.
The Global Infrastructure API uses named entity recognition on unstructured textual data as well as Refinitiv’s open, permanent, and universal identifiers for data, PermID, to pull in associated metadata into a knowledge graph to create a richer, context heavy dataset for analysis.
Currently, the graph spans 60,000 projects, 24,000 entities, 30,000 connected deals and 4 million + relations.
FX Impact Intelligence: Identifying Key Events Impacting Currency Pairs
The FX Impact Intelligence project seeks to develop a new app for foreign exchange (FX) professionals to stay on top of the important events impacting currency pairs.
The app uses rules based and ML models to filter for FX market related events. Each FX market news cluster then passes through an impact analysis model which identifies currency price jumps and impact trends based on the similar characteristics of historical news events.
This translates to capabilities and features such as curated feeds of events related to currency markets, as well as daily impact charts that display the impactful events on price charts alongside the daily volume, volatility and spread. The project currently covers 15 currency pairs.
LSEG Labs is engaging a small group of selected customers to test out the FX Impact Intelligence to improve the tool and user experience.
Other LSEG Labs Projects
In addition to the Financial Language Modelling, Global Infrastructure API, FX Impact Intelligence projects, the LSEG Labs are working on four other initiatives.
SentiMine is an advanced discoverability tool targeted at equity analysts that combines NPL, sentiment analysis and deep learning to extract insights from a large volume of unstructured content including research reports, company transcripts and filings.
ESG Controversy Prediction leverages ML and NLP to detect environmental, social and governance (ESG) controversies in companies news.
Mosaic combines real-time analytics, ML, and intelligent search processes across different datasets to explain sudden price moves.
And Trade Discovery is a tool that enables banks to find the data required to comply with the Risk Factor Eligibility Test (RFET), accelerating the process of conducting the test from months to seconds.
Discover more of the LSEG Labs’ ongoing projects by following this link.
Featured image credit: Photo by David Vincent on Unsplash
