RePEcLLMs & Text

Measuring sentiment news with transformer-based language models

FinBERT-based sentiment measures agree more closely with human raters than dictionary methods and better distinguish positive, neutral, and negative financial news articles.

Read the paper PDF

View as Markdown Open in Claude Open in ChatGPT Open in PerplexitySite index for AI (llms.txt)

Featured in No. 135 on 16 Oct 2026 · 7 days after release

Confusion matrices comparing transformer-based and dictionary sentiment measures against human labels — Figure 5: Confusion matrices by sentiment measure. Rows correspond to human sentiment la- bels and columns to predicted labels. Transformer-based measures show substantial agreement with human evaluations across all sentiment categories, whereas vocabulary-based measures classify most articles as n…

Released: 9 Oct 2026
First featured: No. 135 · 16 Oct 2026
Published in: Not yet, as far as Semantic Scholar knows
Fanfare: 3 of 5
Identifier: RePEc:ecb:ecbwps:20263283
Authors: Stefano Borgioli et al.

Citations and venue from Semantic Scholar (ODC-BY), refreshed weekly. Summary: Quant Letter (CC BY 4.0).

Related papers

From News to Returns: A Granger-Causal Hypergraph Transformer on the Sphere

Visualizing Public Opinion on X: A Real-Time Sentiment Dashboard Using VADER and DistilBERT

Advanced Deep Learning Techniques for Analyzing Earnings Call Transcripts: Methodologies and Applications

A Systematic Literature Review on Sentimental Analysis using Machine Learning for Preprocessing, Feature Extraction and Classification

No Train, all Gain: Self-Supervised Gradients Improve Deep Frozen Representations

SYNTHEVAL: Hybrid Behavioral Testing of NLP Models with Synthetic CheckLists