Acuity Institutional Resources

How does Acuity’s topic model & attention scoring make the most of Dow Jones News?

Written by Acuity Trading | Feb 12, 2024 11:46:04 AM

Acuity Trading's innovative approach to leveraging Dow Jones News data through their topic model and attention scoring system offers investors a powerful tool for market analysis.Operating unsupervised and without the need for labelled data, Acuity's topic model utilises the Latent Dirichlet Allocation (LDA) methodology to identify latent topics within textual articles. By representing documents as mixtures of these latent topics, Acuity's model uncovers the underlying structure of the news corpus.


Read our latest White Paper, 'Topic Model & Attention Scoring' compiled by our data scientists Lucas Manchado and Ariel Duarte López, Ph.D. to find out:

🔵 How Acuity’s topic model operates unsupervised and without the need for labelled data.

🔵 How our attention scores can accurately capture the recognition a topic receives in the media.

🔵 The influence of attention scores on macroeconomic indicators

Click the image below to dive into our latest academic study:

Through meticulous parameter selection and coherence optimisation, Acuity's model identifies 200 topics within Dow Jones news articles, providing a comprehensive breakdown of relevant themes. These topics, manually named for interpretability, cover a wide range of subjects, from Crude Oil to US Defence. By clustering similar topics using hierarchical agglomerative algorithms, Acuity further refines its analysis, enabling both detailed and generalised insights.

Acuity's attention scoring system complements its topic model by quantifying the recognition that each topic receives in the media. Calculated based on topic probabilities, the attention metric measures topic popularity at various granularity levels and geographical zones. This metric, available daily and monthly for the US, facilitates trend analysis and event impact assessment.

To validate the attention scores' accuracy, Acuity examines their correlation with historical events and macroeconomic indicators, focusing on US GDP growth. Through distance correlation, Spearman's rank correlation, and Granger causality analyses, Acuity demonstrates the relationship between attention scores and economic trends. Notably, certain topic clusters exhibit significant correlations with GDP growth, suggesting their potential as predictive indicators.

Acuity Trading, with over a decade of experience, specialises in providing investors with sharper investment data. Their natural language processing (NLP) technology, supported by a dedicated team of scientists and professionals, simplifies access to financial news data and content. Through partnerships with industry leaders, Acuity ensures the delivery of best-in-class alternative data services for investment professionals worldwide.

Acuity Trading's topic model and attention scoring system offer investors a sophisticated yet intuitive approach to market analysis. By uncovering latent topics within Dow Jones news articles and quantifying their popularity, Acuity provides actionable insights for informed decision-making in today's dynamic financial landscape.

To find out more about how Acuity’s topic model & attention scoring, please read here