🌟 State-of-the-Art Affective Norms Extrapolation with Transformers 🌟

---

✨ Description:

Welcome to our cutting-edge research hub, where we redefine the boundaries of extrapolating affective norms across multiple languages. Using state-of-the-art transformer neural networks, we unlock the profound emotionality of words, serving as an indispensable resource for experimental stimuli selection and in-depth sentiment analysis.

🔍 Highlights:

• 🌐 Multi-Lingual Mastery: We've taken a multi-faceted approach, extrapolating norms for an impressive lineup of languages including English, Polish, Dutch, German, French, and Spanish.

• 🧠 Advanced Transformer Architecture: Our avant-garde transformer-based neural network architecture stands unmatched in semantic and emotional norms extrapolation, exuding unmatched precision and adaptability for each language.

• 📊 Benchmark Surpassing Results: Not just setting but elevating the standards, our revolutionary method boasts an average improvement of Δr = 0.1 in correlations with human judgments, overshadowing previous approaches.

• 🛠️ Next-Gen Stimuli Selection: Venturing into uncharted territory, we unveil a pioneering, unsupervised stimuli selection method, ensuring words of profound semantic resonance are handpicked.

• 💻 Researcher's Web Tool Access: While our main emphasis remains on the transformative models, we also provide a web application tool for researchers, serving as a practical touchpoint for direct extrapolation.

---

** Join us on this revolutionary journey, where state-of-the-art technology meets linguistic depth, and together, let's reshape the world of sentiment analysis and emotion understanding in a multilingual paradigm. Your feedback is not just welcomed—it's essential as we tread on this path of innovation and excellence. 🎉🚀**

---

📝 Citation: Plisiecki, H., & Sobieszek, A. (2023). Affective Norms Extrapolation Using Transformer-based Neural
Networks and Its Application to Experimental Stimuli Selection. Behavior Research Methods.
(accepted for publication)

---

📂 Repository Structure & Description:

• Root: Contains the primary scripts and utilities of the repository.

  • datasets_prep.py: Preparing and preprocessing datasets.
  • dataset_and_model.py: Dataset structures and model architecture.
  • dutch.py, english.py, french.py, german.py, polish.py, spanish.py: Language-specific training scripts.
  • picking_words.py: Picking words for robustness analyses.
  • results.py: Consolidates results from various models and analyses.
  • stimuli_descent_bert.py: Stimuli selection using a BERT-based descent algorithm.
  • training_loop.py: General training loop for neural networks.
  • utils.py: Utility functions and helpers.

• prediction_results: Stores prediction results from various models.

  • abstractness_check_results.csv: Results for the abstractness checks of words.
  • abstractness_compare_results.csv: Comparative analysis of abstractness.
  • Language-specific results: e.g., dutch_results.csv, english_aoa_results.csv.
  • stimuli_descent_results.csv: Results from stimuli descent using BERT.

• study_2_data: Secondary data for supplementary analyses.

  • AOA_Kuperman.xlsx: Dataset on Age of Acquisition in English language.
  • Kazojc2009.txt: Dataset for word frequency in Polish language.

• training_data: Training, validation, and testing datasets.

  • Language-specific datasets: e.g., train_dutch.parquet, test_french.parquet.
  • Warriner ANEW datasets: e.g., warriner_anew_test.parquet, warriner_anew_train.parquet.