With the advent of advanced large language models (LLMs), financial professionals now have access to unprecedented insight into earnings call transcripts.
These models enable the precise evaluation of sentiment and tone in CEO remarks, helping analysts detect optimism or concern in statements on topics like inflation, supply chains, or marketing strategies. The result is a scalable, data-driven alternative to traditional, intuition-led assessments.
LSEG recently delved into how firms can leverage AI to unlock investment and risk management opportunities in earnings call transcripts.
By applying proprietary and fine-tuned LLMs, it is now possible to classify emotions and sentiments across every sentence, speaker, and document. Each transcript can be analysed for over 1,000 topics and more than 4,000 event types, with references to millions of products, organisations and individuals. The insights can be used for alpha generation, ESG research, and risk mitigation.
Traditionally, institutional investors and quantitative firms have relied on subjective interpretation of executive tone. With the help of transcript data and LLM sentiment classification, this product enables a more objective approach—removing human bias and making it easier to scale sentiment analysis across firms and sectors.
LSEG MarketPsych Transcript Analytics is a new data feed developed in collaboration between LSEG and MarketPsych, a company known for AI-driven solutions that extract actionable insights from financial text. Their partnership spans nearly 15 years and has produced a range of sentiment and thematic data feeds, ESG analytics, NLP tools, and predictive models used by financial services firms in more than 25 countries.
The solution is powered by LSEG Transcripts, which cover over 16,000 public companies globally. Users can access more than 1,000 predefined topics and conduct free-text searches through an API that tags sentiment and topics at a granular level. The dataset includes time references, verb tenses, parts of speech and a total of 13 speaker emotions. It also utilises MarketPsych’s fine-tuned roBERTa-based classifiers—one of the most accurate NLP classifiers available.
The API supports custom queries on both historical and real-time data, enabling in-depth research, rapid testing, and integration into production workflows.
Several compelling use cases have emerged. Among US-listed firms, those in the top 10% for positive sentiment during earnings calls show stronger stock price performance the following month—especially companies showing high optimism. This correlation allows investors to build models or select stocks based on sentiment signals.
Transcript Analytics also supports ESG research. An added ESG classifier enables monitoring of how often terms such as “carbon”, “climate”, or “emissions” are mentioned and whether they are discussed positively or negatively.
For risk management, Transcript Analytics can detect shifts in executive tone over time and flag companies with negative sentiment around key issues such as “fines”.
These examples highlight just a few ways that investors and financial institutions can use this AI-powered solution to gain an edge through more advanced and scalable earnings call analysis.
For more information, read the full story here.
Keep up with all the latest FinTech news here
Copyright © 2025 FinTech Global
