We welcome additions via Pull Requests.
“There is no fixed, general set of most discriminative features, although n-grams are often used due to ease of implementation and coverage across linguistic levels. The right features depend on various factors, including the data (e.g., genre, length), task (e.g., authorship, machine text detection), and required computational efficiency. There are numerous feature extractions tools available, some of which have been developed specifically for such factors. We recommend reviewing previous work related to your data/task and testing different features on a heldout validation subset to determine which features discriminate best.” from our survey article (Wegmann et al., 2026)
| Tool | Notes / Link |
|---|---|
| LIWC | Pennebaker et al., 2015 |
| LFTK | Lee and Lee, 2023 |
| StyloMetrix | Okulska et al., 2023 |
| NeuroBiber | Alkiek et al., 2025 |
| Multidimensional Analysis Tagger (MAT) | Nini, 2019 |
| StyloSpeaker (for speakers in speech transcripts) | Aggazzotti et al., 2025 |
| Writeprints+ (PAN authorship verification) | Weerasinghe and Greenstadt, 2020 |
| PAN style change detection | Strøm, 2021, Zuo et al., 2019, Zlatkova et al., 2018 |
| elfen (Efficient Linguistic Feature Extraction) | Maurer, 2026 (paper) |
| gram2vec | Zeng et al., 2025 (paper) |
| Tool | Notes / Link |
|---|---|
| Classifying writing styles within a document | Elahi and Muneer, 2018 |
| Supervised Stylometry (SuperStyl) | Camps and Cafiero, 2024 |
| Tool | Notes / Link |
|---|---|
| Stylo (R) | Eder et al., 2016 |
| JStylo (Java) | PSAL, 2013 |
| Coh-Metrix (for text cohesion and readability) | Graesser et al., 2004 |
| Signature | Millican, 2003 |
| idiolect (R) | Nini, 2026 |
| Language | Tool | Paper |
|---|---|---|
| Spanish | Tulengua | Carreras-Riudavets et al., 2025 |
| Multilingual Stylometry (Python and R) | D33 (CLS-INFRA) | Schöch et al., 2024 |
| Historical Persian | parsipy | Farsi et al., 2025 |
| Cross-language/bilingual | DT-grams | Murauer & Specht, 2021 |
| German/Polish/Russian/Ukranian | StyloMetrix | Okulska et al., 2023 |
| Language | Features | Paper |
|---|---|---|
| Latin | POS tag features | Chen et al., 2024 |
| Urdu | stylometric | Nazir et al., 2021 |
| Bengali | stylometric, n-grams | Hossain et al., 2020 |
| Hinglish | n-grams | Sharma et al., 2018 |
| Chinese | tones/rimes | Hou & Huang, 2019 |
| Chinese | character/rhyme/genre/overlapped words | Tang et al., 2019 |
| Arabic poetry | stylometric | Ahmed et al., 2019 |
| EN/FR/IT/SP | n-grams, word2vec, TFIDF (PAN 2019) | Rahgouy et al., 2019 |
| Model | Model Link | Publication | Notes |
|---|---|---|---|
| CISR | https://huggingface.co/AnnaWegmann/Style-Embedding | https://aclanthology.org/2022.repl4nlp-1.26/ | |
| StyleDistance | https://huggingface.co/StyleDistance/styledistance | https://aclanthology.org/2025.naacl-long.436/ | |
| mStyleDistance | https://huggingface.co/StyleDistance/mstyledistance | https://aclanthology.org/2025.findings-acl.869/ | |
| LUAR | https://huggingface.co/rrivera1849/LUAR-MUD | https://aclanthology.org/2021.emnlp-main.70/ | |
| Multilingual Style Representation | https://huggingface.co/Blablablab/multilingual-style-representation-Llama-3.2 | https://aclanthology.org/2025.emnlp-main.1766/ | |
| STAR | https://huggingface.co/AIDA-UPM/star | https://doi.org/10.1016/j.knosys.2024.111867 | |
| StyleJudge | https://huggingface.co/VibrantVista/gte-large-en-v1.5-stylejudge | https://aclanthology.org/2026.conll-main.31/ | trained for as as GRPO reward model |
| Model | Link |
|---|---|
| LISA | https://ajayp.app/posts/2023/11/learning-interpretable-embeddings-via-llms/ |
| Tool | Notes / Link |
|---|---|
| Latent Style Interpretation | GitHub, HuggingFace demo |
| Project | Link | Description |
|---|---|---|
| Learning Invariant Representations of Social Media Users | https://github.com/noa/iur | |
| A Deep Metric Learning Approach to Account Linking | https://github.com/noa/naacl2021 | |
| Style is NOT a single variable: Case Studies for Cross-Style Language Understanding | https://github.com/dykang/xslue | |
| Does It Capture STEL? A Modular, Similarity-based Linguistic Style Evaluation Framework | https://github.com/nlpsoc/STEL | |
| On the State of the Art in Authorship Attribution and Authorship Verification | https://github.com/JacobTyo/Valla | |
| Capturing Classic Authorial Style in Long-Form Story Generation with GRPO Fine-Tuning | https://github.com/Vince-Liuss/literary_style_model | Style-conditioned long-form story generation: trains a style-similarity judge on authorship-verification data, then uses it as a GRPO reward to fine-tune an 8B generator; four per-author models (Twain, Austen, Dickens, Hardy) on HuggingFace |