I am a Ph.D. student at École Polytechnique Fédérale de Lausanne (EPFL) and a research assistant at Idiap. I work under the supervision of Dr. James Henderson in the Natural Language Understanding group. My project is funded by the NCCR Evolving Language, Swiss National Science Foundation Agreement.
My research focuses on discovering meaningful units (i.e., morphemes) from character sequences in an unsupervised manner. We believe that representing text at morpheme-level would lead to better understanding of language in NLP models.
Broadly, I am interested in morphology induction, subword segmentation and unsupervised object discovery methods.
I received my BSc in Software Engineering (2017) and my MSc in Artificial Intelligence (2019) from Sharif University of Technology.
Melika Behjati, and James Henderson. "Inducing Meaningful Units from Character Sequences with Slot Attention." arXiv preprint arXiv:2102.01223 (2021).
Melika Behjati, Seyed Mohsen Moosavi Dezfooli, Mahdieh Soleymani Baghshah, and Pascal Frossard. "Universal Adversarial Attacks on Text Classifiers." In Proceedings of IEEE ICASSP2019 IEEE International Conference on Acoustics, Speech and Signal Processing (Icassp), no. CONF, pp. 7345-7349. IEEE, 2019.