PhD @ CMU
Graduate Research Assistant @ DialRC
Currently, I'm a PhD student at CMU's Language Technologies Institute. I'm working on dialog systems, advised by Maxine Eskenazi as part of the DialRC group.
Some recent news:
Doing an internship at Amazon this summer! Excited to be working with Dilek Hakkani-Tur and Mihail Eric.
One paper accepted to SIGdial 2020. Details to follow soon.
I was accepted to continue my PhD at CMU! Estimated graduation is 2022.
A paper accepted to EMNLP 2019 titled Multi-Granularity Representations of Dialog. We show that you can learn multipe granularities of representation by controlling the negative candidates during the task of dialog retrieval.
Two papers accepted at SIGdial 2019!. I'll be giving a talk on Structured Fusion Networks and will have a poster on our multi-reference dataset. I'll also be giving a talk at the Dialog for Good workshop about our dialog system for seniors, CMU GetGoing.
I had a paper accepted to ACL 2019 on dialog-specific pretraining objectives. We introduce several novel pretraining objectives that obtain good performance results and show pretraining to result in better generalizability, less data-hungry models and faster convergence.
I had a publication accepted to NeurIPS 2018! My paper introduces the Middle-Out Decoder which uses self-attention to generate sequences from the middle-out rather than left-to-right.
Feel free to contact me if you're interested in my work, have an interested opportunity, or you just want to chat about NLP or deep learning!
Shikib Mehri, Maxine Eskenazi. "Unsupervised Evaluation of Interactive Dialog with DialoGPT," In Special Interest Group on Discourse and Dialogue (SIGdial), 2020
Shikib Mehri, Maxine Eskenazi. "USR: An Unsupervised and Reference Free Evaluation Metric for Dialog Generation," In Association for Computational Linguistics (ACL), 2020
Yulan Feng, Shikib Mehri, Maxine Eskenazi, Tiancheng Zhao. "'None of the Above': Measure Uncertainty in Dialog Response Retrieval," In Association for Computational Linguistics (ACL), 2020
Shikib Mehri, Maxine Eskenazi. "Multi-Granularity Representations of Dialog," In Empirical Methods in Natural Language Processing (EMNLP-IJCNLP), 2019
Shikib Mehri, Alan W Black, Maxine Eskenazi. "CMU GetGoing: An Understandable and Memorable Dialog System for Seniors," In Workshop on Dialog for Good (DiGo) in SIGdial, 2019 [ORAL]
Prakhar Gupta, Shikib Mehri, Tiancheng Zhao, Amy Pavel, Maxine Eskenazi, Jeffrey P Bigham. "Investigating Evaluation of Open-Domain Dialogue Systems With Human Generated Multiple References," In Special Interest Group on Discourse and Dialogue (SIGdial), 2019 [CODE]
Shikib Mehri, Evgeniia Razumovskaia, Tiancheng Zhao, Maxine Eskenazi. "Pretraining Methods for Dialog Context Representation Learning," In Association for Computational Linguistics (ACL), 2019
Maxine Eskenazi, Shikib Mehri, Evgeniia Razumovskaia, Tiancheng Zhao. "Beyond Turing: Intelligent Agents Centered on the User." Preprint, 2019
Shikib Mehri, Leonid Sigal. "Middle-Out Decoding," In Neural Information Processing Systems (NeurIPS), 2018
Shikib Mehri, Giuseppe Carenini. "Chat Disentanglement: Identifying Semantic Reply Relationships with Random Forests and Recurrent Neural Networks," In International Joint Conference on Natural Language Processing (IJCNLP), 2017 [ORAL]
As a scientist intern on the Alexa Conversational AI team, I was mentored by Dilek Hakkani-Tur and Mihail Eric. Two papers on the way with SOTA results 5 dialog tasks (including big gains on MultiWOZ + intent prediction).
A student in the Language Technologies Institute at Carnegie Mellon University. I am supervised by Profs. Maxine Eskenazi
As a scientist intern on the Alexa AI team, I developed novel approaches for conversational speech recognition, yielding significant improvements in both quantitative and qualitative evaluation. A publication is in the works.
Worked with Prof. Leonid Sigal on video captioning. I developed a novel middle-out sequence decoder, attaining state of the art results. A publication has been accepted to NeurIPS 2018.
Teaching assistant for Multi-Modal Deep Learning. Assisted with the design/development of assignments, as well as general TA duties. One of few undergraduates to ever TA a graduate class.
As part of the Windows Feedback Analysis team, I constructed a number of deep learning based classification and semantic similarity models for the purposes of natural language understanding.
On the Machine Translation team, I worked on the implementation of a number of subword Neural Machine Translation models to ultimately improve post and comment translation on Facebook.
As a member of the Wasserman lab at CMMT, I designed and implemented algorithms for read alignment on a graph based representation of the reference genome.
Worked with Prof. Giuseppe Carenini on the problem of thread disentanglement in multiparticipant chats. My approach employed novel strategies (including leveraging a weakly-supervised LSTM classifier) to attain state of the art results. I presented a publication at IJCNLP 2017.
Taught CPSC 110, CPSC 213 (Computer Systems), CPSC 313 (Operating Systems), CPSC 304 (Databases). I ran labs containing 30+ students, lectured in tutorials, invigilated exams, held office hours and graded assignments and exams.
On the Ads Targeting Modeling team, I worked on generating/evaluating user-interest mappings. I also constructed an incredibly accurate classification model to predict the correct categorization of a given interest.
As the Chief Technology Officer of IntelliMed, a funded startup, I led developers in building an application which automates the process of writing pharmacy medication reviews for patients.
Implemented functionality to detect incorrectly configured network switches and to change the assignment strategies for linerate capable ports in order to decrease downtime.
I recently graduated with a major in Honours Computer Science (with co-op). I graduated with a 4.20 cumulative GPA as well as a 4.33 in-major GPA.
One of 20 students to attend this rigorous, highly-accelerated program that condenses five years of highschool into two. I completed the program at the age of 14, leading to an early entrance to university.
CRA Outstanding Undergraduate Researcher Award.
Won 2nd place and crowd favourite at StartUp-Weekend Vancouver, for AlbertAI, a personal assistant that revolutionizes education.
Third place at LumoHacks for our application which utilized NLP to identify reddit/twitter users at risk of depression, and alert their friends.
Excellence scholarship for continuing students.
Charles and Jane Banks Scholarship.
Received the science scholar distinction due to stellar academic performance.
Won the Telus/IEEE Datathon. Our application, SaferSurrey used machine learning in combination with crime data from the Surrey RCMP to identify the safest route home through dangerous parts of the city. We won $3000 from the competition, as well as the opportunity to pitch our idea to the Surrey RCMP.
Placed 3rd at the 2017 Division 2 ACM-ICPC competition in the Pacific Northwest Region.
Obtained fourth place (first amongst undergraduate submissions) DataSense VanData competition. My submission can be found here.
Have received the Dean's List distinction every term I've been at the University of British Columbia.