Paper: arXiv 2409.20397
Authors: Fabian Billert, Stefan Conrad
Abstract
We extract the sentiment from german and english news articles on companies in the DAX40 stock market index and use it to create a sentiment-powered pendant. Comparing it to existing products which adjust their weights at pre-defined dates once per month, we show that our index is able to react more swiftly to sentiment information mined from online news. Over the nearly 6 years we considered, the sentiment index manages to create an annualized return of 7.51% compared to the 2.13% of the DAX40, while taking transaction costs into account. In this work, we present the framework we employed to develop this sentiment index.
Complexity vs Empirical Score
- Math Complexity: 4.5/10
- Empirical Rigor: 6.0/10 (capped from 7.0: no out-of-sample test; see penalty flags)
- Quadrant: Street Traders — empirical and applied, lower math complexity (effective empirical rigor capped from 7.0)
Why this score: The paper applies established NLP and convex optimization techniques (low advanced math) but demonstrates strong empirical rigor with a 6-year backtest on real news data, transaction costs, and a live benchmark comparison.
Research Flowchart
flowchart TD A["Research Goal<br>Create a Sentiment-Driven<br>DAX40 Performance Index"] --> B["Data Collection<br>German & English News Articles"] B --> C["Methodology<br>Deep Learning Sentiment Extraction<br>Company Stock Linkage"] C --> D["Computation<br>Rebalancing & Weight Calculation<br>Transaction Cost Deduction"] D --> E["Benchmarking<br>vs. Standard DAX40 (2.13% return)"] E --> F["Key Outcomes<br>7.51% Annualized Return<br>Faster Market Reaction"]
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes a framework to extract sentiment from news headlines and convert it into weights for a sentiment‑driven performance index for the DAX40.
Source sentence
In this work, we present the framework we employed to develop this sentiment index.
Assumes 6
- Penalty factor for transaction costs is set to 1 (one polarity). model
Source sentence
In order to trade only for signigifant sentiment changes, we set the penalty factor to 1, which is one polarity.
- Maximum weight per single company is limited to 10% of the portfolio. execution
Source sentence
As for constraints, we decide to limit the total investment in a single company to 10%.
- Sum of portfolio weights must be between 99% and 99.9% (effectively 100%). execution
Source sentence
we instead require the sum of weights to be between 99% and 99.9%
- Transaction costs are 0.05% of the stock price per trade. execution
Source sentence
We set the trading costs to 0.05% of the price of the stock, which is multiplied with the amount of stocks traded to get the total cost of a transaction.
Show 2 more
- German news articles are translated to English before sentiment classification. data
Source sentence
Rather than using two separate models or retraining the model as a multilanguage model, we translated the german articles to english before classifying their sentiment.
- Balanced accuracy of the sentiment model is 67% for English and 65% for translated German articles. model
Source sentence
We found that for the english articles, we achieve a balanced accuracy of 67%, while for the german-to-english translated articles we still achieve 65%.
Data
DAX40 · equities · daily
- News headlines (German and English) on DAX40 companies text-news unknown
Source sentence
We extract the sentiment from german and english news articles on companies in the DAX40 stock market index
- Financial Phrasebank dataset text-news public
Source sentence
... and was fine‑tuned first with the Financial Phrasebank dataset from (Malo et al. 2014)
- Manually annotated financial news headlines text-news unknown
Source sentence
and then again with manually annotated financial news headlines
Costs
Modeled
Trading costs are set to 0.05% of the stock price per transaction.
Source sentence
We set the trading costs to 0.05% of the price of the stock, which is multiplied with the amount of stocks traded to get the total cost of a transaction.
Finds 3
annual return 7.5%
- The sentiment index achieved an annualized return of 7.51% over the 6‑year period, compared to 2.13% for the DAX40 benchmark. performance
Source sentence
Over the nearly 6 years we considered, the sentiment index manages to create an annualized return of 7.51% compared to the 2.13% of the DAX40, while taking transaction costs into account.
- The sentiment‑driven index outperforms the DAX40 after 2018, showing higher cumulative performance despite similar results in 2017‑2018. performance
Source sentence
We can see that the sentiment index shows a similar performance in 2017 and 2018, but manages to outpace the DAX in the following...
- The proposed framework enables weight adjustments at any point in time without incurring large transaction costs, unlike existing monthly‑rebalanced products. theoretical
Source sentence
our more flexible approach which is able to change the constituent weights at any point in time without amassing large amounts of transaction costs.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 13 related papers). Coverage is partial, so this is neither replication nor refutation.
Code
No repository by the authors found in the paper.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.
Cited References (229)
- GL-GCN: Global and Local Dependency Guided Graph Convolutional Networks for aspect-based sentiment classification (2021) — link
- Recent Advances in Natural Language Processing via Large Pre-Trained Language Models: A Survey (2021) — link
- The Effects of Twitter Sentiment on Stock Price Returns (2015) — link
- German’s Next Language Model (2020) — link
- Good debt or bad debt: Detecting semantic orientations in economic texts (2014) — link
- Improving Language Understanding by Generative Pre-Training (2018)
- Web Media and Stock Markets : A Survey and Future Directions from a Big Data Perspective (02.2018) — link
- Sentiment indices and stock returns: Evidence from China (2021) — link
and 221 more
- How does Punctuation Affect Neural Models in Natural Language Inference — link
- Learning Tone and Attribution for Financial Text Mining (2016)
- The Effect of Social Media on Market Liquidity (2015) — link
- What is this article about? Generative summarization with the BERT model in the geosciences domain (2022) — link
- ALBERT: A Lite BERT for Self-supervised Learning of Language Representations (2019) — link
- A tale of two sentiment scales: Disentangling short-run and long-run components in multivariate sentiment dynamics (2019) — link
- Data Preprocessing for Efficient Sentimental Analysis (2018) — link
- Aggregation of Sentiment Analysis Index with Hesitant Fuzzy Sets for Financial Time Series Forecasting (2021) — link
- A dataset for Sentiment analysis of Entities in News headlines (SEN) (2021) — link
- MAD-X: An Adapter-Based Framework for Multi-Task Cross-Lingual Transfer (2020) — link
- Machine Learning Classification Methods and Portfolio Allocation: An Examination of Market Efficiency (2020) — link
- Public Mood–Driven Asset Allocation: the Importance of Financial Sentiment in Portfolio Management (2018) — link
- Sentiment and Knowledge Based Algorithmic Trading with Deep Reinforcement Learning (2020) — link
- Machine learning portfolio allocation (2022) — link
- Leveraging Social Media to Predict Continuation and Reversal in Asset Prices (2021) — link
- An Entity-Driven Framework for Abstractive Summarization (2019) — link
- Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer (2019) — link
- Transformers generalize differently from information stored in context vs in weights (2022) — link
- The Cross‐Section of Expected Stock Returns (1992) — link
- Universal Language Model Fine-tuning for Text Classification (2018) — link
- A picture is worth a thousand words: Measuring investor sentiment by combining machine learning and photos from news (2021) — link
- Causally Estimating the Sensitivity of Neural NLP Models to Spurious Features (2021) — link
- Deep learning for time series classification: a review (2019) — link
- Explanation-Based Human Debugging of NLP Models: A Survey (2021-12) — link
- Re-Evaluating GermEval17 Using German Pre-Trained Language Models (2021) — link
- Stock Price Prediction Using Sentiment Analysis (2021) — link
- FinancialBERT - A Pretrained Language Model for Financial Text Mining (2022-02) — link
- Deep Reinforcement Learning for Asset Allocation in US Equities (2020) — link
- Generalization through Memorization: Nearest Neighbor Language Models (2019) — link
- Sentiment analysis of financial news articles using performance indicators (2018) — link
- Structural Text Segmentation of Legal Documents (2020) — link
- Parameterized Explanations for Investor / Company Matching (2021) — link
- Context-Guided BERT for Targeted Aspect-Based Sentiment Analysis (2020) — link
- Using four different online media sources to forecast the crude oil price (2021) — link
- Stance Prediction for Contemporary Issues: Data and Experiments (2020) — link
- Stock Movement Prediction with Financial News using Contextualized Embedding from BERT (2021) — link
- Training a Broad-Coverage German Sentiment Classification Model for Dialog Systems (2020) — link
- A Comparative Analysis of Forecasting Financial Time Series Using ARIMA, LSTM, and BiLSTM (2019) — link
- A Challenge Dataset and Effective Models for Aspect-Based Sentiment Analysis (2019) — link
- Spectral analysis of time-dependent market-adjusted return correlation matrix (2018) — link
- Utilizing BERT for Aspect-Based Sentiment Analysis via Constructing Auxiliary Sentence (2019) — link
- Intelligent trading systems: a sentiment-aware reinforcement learning approach (2021) — link
- Fine-grained analysis of explicit and implicit sentiment in financial news articles (2015) — link
- Automatic extraction of gene-disease associations from literature using joint ensemble learning (2018) — link
- HHU at SemEval-2017 Task 5: Fine-Grained Sentiment Analysis on Financial Data using Machine Learning Methods (2017) — link
- Return Attribution (2012)
- Ad Headline Generation using Self-Critical Masked Language Model (2021) — link
- Term Expansion and FinBERT fine-tuning for Hypernym and Synonym Ranking of Financial Terms (2021) — link
- Attention Modeling for Targeted Sentiment (2017) — link
- Automated Investor Sentiment Classification using Financial Social Media (2021) — link
- Matching the Blanks: Distributional Similarity for Relation Learning (2019) — link
- S&P 500 Twitter Sentiment Index | S&P Dow Jones Indices (2021)
- BERT-Based Stock Market Sentiment Analysis (2020) — link
- A Primer on Neural Network Models for Natural Language Processing (2016) — link
- Semantic Frames to Predict Stock Price Movement (2013) — link
- Reinforcement Learning: An Introduction, by Sutton, R.S. and Barto, A.G. (1999) — link
- HuggingFace’s Transformers: State-of-the-art Natural Language Processing (2019) — link
- Market sentiment-aware deep reinforcement learning approach for stock portfolio allocation (2021) — link
- Universal Language Model Fine-tuning for Text Classification (2018) — link
- Sentiment Classification by Incorporating Background Knowledge from Financial Ontologies (2022)
- Design and analysis of a large-scale COVID-19 tweets dataset (2021) — link
- Listening to Chaotic Whispers: A Deep Learning Framework for News-oriented Stock Trend Prediction (2018) — link
- One for “All”: a unified model for fine-grained sentiment analysis under three tasks (2021) — link
- exBERT: Extending Pre-trained Models with Domain-specific Vocabulary Under Constrained Training Resources (2020) — link
- Prediction of stock values changes using sentiment analysis of stock news headlines (2021) — link
- CVXPY: A Python-Embedded Modeling Language for Convex Optimization (2016) — link
- TweetSenti: Target-dependent Tweet Sentiment Analysis (2019) — link
- AllenNLP: A Deep Semantic Natural Language Processing Platform (2018) — link
- Predicting Stock Market Indicators Through Twitter “I hope it is not as bad as I fear” (2011) — link
- DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter (2019) — link
- Target-Based Sentiment Analysis using a BERT Embedded Model (2020) — link
- Aspect-based Sentiment Analysis in Question Answering Forums (2021) — link
- Neural Machine Translation of Rare Words with Subword Units (2015) — link
- A BERT based Sentiment Analysis and Key Entity Detection Approach for Online Financial Texts (2021) — link
- Stock Price Prediction using Facebook Prophet (2022) — link
- Tracking Turbulence Through Financial News During COVID-19 (2021) — link
- INF-UFG at FiQA 2018 Task 1 (2018) — link
- Subjectivity detection for sentiment analysis on Twitter data (2020) — link
- Stock-specific sentiment and return predictability (2020) — link
- OPT: Open Pre-trained Transformer Language Models (2022) — link
- Bayesian model averaging and the conditional volatility process: an application to predicting aggregate equity returns by conditioning on economic variables (2021) — link
- Unsupervised Domain Clusters in Pretrained Language Models (2020) — link
- NumHTML: Numeric-Oriented Hierarchical Transformer Model for Multi-Task Financial Forecasting (2022) — link
- Graph Auto-Encoders for Financial Clustering (2021) — link
- BERT’s sentiment score for portfolio optimization: a fine-tuned view in Black and Litterman model (2022) — link
- SemEval-2014 Task 4: Aspect Based Sentiment Analysis — link
- Visualizing the Financial Impact of Presidential Tweets on Stock Markets (2021) — link
- Image–text sentiment analysis via deep multimodal attentive fusion (2019) — link
- Gated Neural Networks for Targeted Sentiment Analysis (2016) — link
- GermanQuAD and GermanDPR: Improving Non-English Question Answering and Passage Retrieval (2021) — link
- A Robust Predictive Model for Stock Price Prediction Using Deep Learning and Natural Language Processing (2021) — link
- FinSSLx: A Sentiment Analysis Model for the Financial Domain Using Text Simplification (2018) — link
- Evaluation of Sentiment Analysis in Finance: From Lexicons to Transformers (2020) — link
- An Entity-Level Sentiment Analysis of Financial Text Based on Pre-Trained Language Model (2020) — link
- Algorithmic Trading using Sentiment Analysis and Reinforcement Learning
- Understanding Pre-trained BERT for Aspect-based Sentiment Analysis (2020) — link
- Contextual Sentence Analysis for the Sentiment Prediction on Financial Data (2021) — link
- Sentiment correlation in financial news networks and associated market movements (2021) — link
- SemEval-2017 Task 5: Fine-Grained Sentiment Analysis on Financial Microblogs and News (2017) — link
- Long Short-Term Memory-Networks for Machine Reading (2016) — link
- Convolutional Neural Networks for Sentence Classification (2014) — link
- Google’s Neural Machine Translation System: Bridging the Gap between Human and Machine Translation (2016) — link
- Adverse Media Mining for KYC and ESG Compliance (2021) — link
- Portfolio Optimization with 2D Relative-Attentional Gated Transformer (2020) — link
- Contextual String Embeddings for Sequence Labeling (2018) — link
- Automatic Generation of Product Association Networks Using Latent Dirichlet Allocation (2011) — link
- Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks (2019) — link
- Attention Is All You Need (2017) — link
- How to Fine-Tune BERT for Text Classification? (2020) — link
- Quantile connectedness between sentiment and financial markets: Evidence from the S&P 500 twitter sentiment index (2022) — link
- Daily market news sentiment and stock prices (2019) — link
- Stock Market Prediction using Recurrent Neural Network’s LSTM Architecture (2021) — link
- Insights of strength and weakness of evolving methodologies of sentiment analysis (2021) — link
- Parameter-Efficient Transfer Learning for NLP (2019) — link
- Disagreement after News: Gradual Information Diffusion or Differences of Opinion? (2021) — link
- LSTM Architecture for Oil Stocks Prices Prediction (2022) — link
- Deep Residual Learning for Image Recognition (2015) — link
- NewsMTSC: A Dataset for (Multi-)Target-dependent Sentiment Classification in Political News Articles (2021) — link
- The “Sixth” Factor— A Social Media Factor Derived Directly from Tweet Sentiments (2017) — link
- Inside the Mind of Investors During the COVID-19 Pandemic: Evidence from the StockTwits Data (2020) — link
- Aspect-Based Sentiment Analysis using BERT — link
- Predicting Stock Movement Using Sentiment Analysis of Twitter Feed with Neural Networks (2020) — link
- A rewriting system for convex optimization problems (2018) — link
- Investigating the Performance of Fine-tuned Text Classification Models Based-on Bert (2020) — link
- Content analysis of business communication: introducing a German dictionary (2019) — link
- Time horizon-aware modeling of financial texts for stock price prediction (2021) — link
- Stock Price Prediction via Discovering Multi-Frequency Trading Patterns (2017) — link
- Pointer Networks (2015) — link
- SentiHood: Targeted Aspect Based Sentiment Analysis Dataset for Urban Neighbourhoods (2016) — link
- Analysis of News and Stock Market with FinBERT models (2022) — link
- TweetsCOV19 - A Knowledge Base of Semantically Annotated Tweets about the COVID-19 Pandemic (2020) — link
- An Effective Domain Adaptive Post-Training Method for BERT in Response Selection (2020) — link
- Financial Aspect-Based Sentiment Analysis using Deep Representations (2018) — link
- Leveraging BERT to Improve the FEARS Index for Stock Forecasting (2019) — link
- BUZZ - VanEck Social Sentiment ETF (2021)
- Interdependency between the Stock Market and Financial News (2019) — link
- FinEAS: Financial Embedding Analysis of Sentiment (2021) — link
- AdapterFusion: Non-Destructive Task Composition for Transfer Learning (2020) — link
- FinBERT: A Pre-trained Financial Language Representation Model for Financial Text Mining (2020) — link
- BERTopic: Neural topic modeling with a class-based TF-IDF procedure (2022) — link
- The Capital Asset Pricing Model: Theory and Evidence (2004) — link
- A Numerical-Based Attention Method for Stock Market Prediction With Dual Information (2019) — link
- Competition among memes in a world with limited attention (2012) — link
- FinBERT: A Pretrained Language Model for Financial Communications (2020) — link
- A Stochastic Time Series Model for Predicting Financial Trends using NLP (2021) — link
- Portfolio Optimization-Based Stock Prediction Using Long-Short Term Memory Network in Quantitative Trading (2020) — link
- A Five-Factor Asset Pricing Model (2014) — link
- Beyond facts – a survey and conceptualisation of claims in online discourse analysis (2022) — link
- News-based business sentiment and its properties as an economic index (2022) — link
- A Primer in BERTology: What We Know About How BERT Works (2021) — link
- CNNPred: CNN-based stock market prediction using several data sources (2018) — link
- An automated portfolio trading system with feature preprocessing and recurrent reinforcement learning (2021) — link
- Neural Network Predictive Modeling on Dynamic Portfolio Management—A Simulation-Based Portfolio Optimization Approach (2020) — link
- A Deep Reinforcement Learning Framework for the Financial Portfolio Management Problem (2017) — link
- Artificial Intelligence Applied to Stock Market Trading: A Review (2021) — link
- Feature Investigation for Stock Returns Prediction Using XGBoost and Deep Learning Sentiment Classification and Deep Learning Sentiment Classification (2021)
- What to Pre-Train on? Efficient Intermediate Task Selection (2021) — link
- Choosing news topics to explain stock market returns (2020) — link
- Sentiment in German-language News and Blogs, and the DAX
- Statistical and Machine Learning forecasting methods: Concerns and ways forward (2018) — link
- ERNIE: Enhanced Language Representation with Informative Entities (2019) — link
- Stock Price Prediction Under Anomalous Circumstances (2020) — link
- AdapterHub: A Framework for Adapting Transformers (2020) — link
- Defending Against Neural Fake News (2019) — link
- Brazilian stock market performance and investor sentiment on Twitter (2022) — link
- Learning to Determine the Quality of News Headlines (2020) — link
- SciBERT: A Pretrained Language Model for Scientific Text (2019) — link
- Language Models are Unsupervised Multitask Learners
- Exploring user historical semantic and sentiment preference for microblog sentiment classification (2021) — link
- On the Effectiveness of Adapter-based Tuning for Pretrained Language Model Adaptation (2021) — link
- Stock Price Predictions with LSTM Neural Networks and Twitter Sentiment (2021) — link
- Social media and firm equity value (2013) — link
- AdapterDrop: On the Efficiency of Adapters in Transformers (2020) — link
- Domain-Specific Language Model Pretraining for Biomedical Natural Language Processing (2022) — link
- XLNet: Generalized Autoregressive Pretraining for Language Understanding (2019) — link
- Facebook FAIR’s WMT19 News Translation Task Submission (2019) — link
- From Opinion Mining to Financial Argument Mining (2021) — link
- Misleading the Covid-19 vaccination discourse on Twitter: An exploratory study of infodemic around the pandemic (2021) — link
- HyperPELT: Unified Parameter-Efficient Language Model Tuning for Both Language and Vision-and-Language Tasks (2022) — link
- Handbook of Natural Computing (2012) — link
- Target-Aspect-Sentiment Joint Detection for Aspect-Based Sentiment Analysis (2020) — link
- BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding (2018) — link
- Prefix-Tuning: Optimizing Continuous Prompts for Generation (2021) — link
- Sentiment Analysis — Comparing 3 Common Approaches: Naive Bayes, LSTM, and VADER | by Kevin C Lee | Towards Data Science — link
- Bayesian Topic Regression for Causal Inference (2021) — link
- S&P 500 Twitter Sentiment Select Equal Weight Index | S&P Dow Jones Indices (2021)
- Exploiting BERT for Multimodal Target Sentiment Classification through Input Space Translation (2021) — link
- Deep learning-based feature engineering for stock price movement prediction (2019) — link
- Stock Price Prediction Using News Sentiment Analysis (2019) — link
- Recent Advances in Reinforcement Learning in Finance (2021) — link
- Stock movement predictive network via incorporative attention mechanisms based on tweet and historical prices (2020) — link
- Crude oil price forecasting incorporating news text (2020) — link
- Effective Approaches to Attention-based Neural Machine Translation (2015) — link
- Investors’ mood and herd investing: a quantile-on-quantile regression explanation from crypto market (2021) — link
- LSTM-based sentiment analysis for stock price forecast (2021) — link
- Stock Movement Prediction from Tweets and Historical Prices (2018) — link
- FinBERT: Financial Sentiment Analysis with Pre-trained Language Models (2019) — link
- It’s a New Era of Investing. Welcome to BUZZ. - BUZZ Indexes (2021)
- Stock Price Movement Prediction Using Sentiment Analysis and CandleStick Chart Representation. (2021) — link
- Entity-Aware Abstractive Multi-Document Summarization (2021) — link
- Zero-Shot Aspect-Based Sentiment Analysis (2022) — link
- Target-dependent Twitter Sentiment Classification — link
- Advances in Intelligent Data Analysis XX, 20th International Symposium on Intelligent Data Analysis, IDA 2022, Rennes, France, April 20–22, 2022, Proceedings (2022) — link
- PEGASUS: Pre-training with Extracted Gap-sentences for Abstractive Summarization (2019) — link
- Sentiment Analysis for Fake News Detection (2021) — link
- Scale Efficiently: Insights from Pre-training and Fine-tuning Transformers (2021) — link
- A machine learning approach to detect early signs of startup success (2021) — link
- Fusion of Sentiment and Asset Price Predictions for Portfolio Optimization (2022) — link
- Evaluating the Rationales of Amateur Investors (2021) — link
- Narrative and computational text analysis in business and economic history (2021) — link
- Front Page News: The Effect of News Positioning on Financial Markets (Working Paper)
- Forecasting the CNY-CNH pricing differential: The role of investor attention (2018) — link
- Impact of Public News Sentiment on Stock Market Index Return and Volatility (2021) — link
- Stock price prediction using BERT and GAN (2021) — link
- FinBERT—A Deep Learning Approach to Extracting Textual Information (2020) — link
- Don’t Stop Pretraining: Adapt Language Models to Domains and Tasks (2020) — link
- Identifying the influential factors of commodity futures prices through a new text mining approach (2020) — link
- Learning and Evaluating General Linguistic Intelligence (2019) — link
- Hybrid Deep Learning Models for Sentiment Analysis (2021) — link
- Reformer: The Efficient Transformer (2020) — link
- The Effect of Preprocessing Techniques on Twitter Sentiment Analysis (2016) — link
0 of 229 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.