Aylin Caliskan: Quantifying the Translator Effect: Identifying authors and machine translation tools in translated text
- 1. Quantifying the Translator Effect: Identifying authors and machine translation tools in translated text Aylin Caliskan Computer Science PhD Student at Drexel University
- 2. Authorship Attribution Identifying the author of an anonymous text Software JStylo Training corpus Lexical, character, syntactic, semantic features Machine learning tools
- 3. Brennan-Greenstadt Adversarial Stylometry Corpus 21 st century writing Diverse topics 13 authors Native English authors Minimum of 5000 words per author Exclude obfuscation and imitation texts
- 4. Translator Effect Original text: “This paper set out to explore what communities who seek greater local control over their lives face as the world begins the herculean task of reorganizing its energy systems.” Google's translator: “In this paper, we explore the world to look for more local control of community life from what it is like to face a very difficult task to begin to rebuild their energy systems.”
- 6. Authorship Attribution in Translated Text
- 7. Effective features in authorship attribution and translator attribution
- 8. Conclusion
- 9. Future Work Different types of translators One way translations Arabic, Russian, Spanish, and Turkish corpus
- 10. Thanks for listening. Any questions? Feel free to contact me: aylin.caliskan@drexel.edu