New AI-powered signal transforms company earnings call transcripts into differentiated alpha across global markets
ExtractAlpha today announced the launch of its Transcripts AI Model, an upgraded NLP-based stock selection signal designed to extract predictive insights from earnings call transcripts.
Building on traditional transcript sentiment analysis, the model uses contextual embeddings and machine learning to capture deeper language patterns associated with future stock performance. The signal is trained to predict forward equity returns and is designed for systematic investors seeking differentiated, low-turnover alpha from company communications.
The Transcripts AI Model includes two regional modules: Global English, based on English-language earnings calls across the US, Americas ex-US, EMEA, and APAC; and Japan, based on Japanese-language earnings calls from SCRIPTS Asia.
In ExtractAlpha’s research, the model generated strong risk-adjusted performance across global markets while outperforming traditional transcript sentiment approaches. Full research results are available in the accompanying white paper.
“Earnings calls remain one of the richest sources of forward-looking company information,” said Vinesh Jha, CEO of ExtractAlpha. “The new Transcripts AI Model builds on traditional sentiment analysis with advanced AI that captures deeper contextual language patterns, helping investors identify differentiated sources of alpha.”
The Transcripts AI Model is available today as part of ExtractAlpha’s suite of alternative data signals and quantitative stock selection models.
For more information, contact us.