Principal Research Engineer at Samsung R&D Institute (UK).
PhD in Data Science from the University of Edinburgh (SMASH group).

I am interested in building computational agents that process natural language, and react in a manner that brings value to our lives. Towards this vision, my recent work has focused on safety for generative artificial intelligence (GenAI), as supported by large language models (LLMs). For instance, check out this paper (under review) about a method for parameter-efficient guardrailing LLMs. I have also worked on problems in Computational Social Science, Figurative Language Comprehension, Machine Translation, and Computer Vision. For more details, check out my publications below.

silviu dot vlad dot oprea at gmail dot com

My CV in PDF format is here.

News​

• 12 August 2024: Check out our paper, LoRA-Guard: Parameter-Efficient Guardrail Adaptation for Content Moderation of Large Language Models (arxiv; under review). We introduce a parameter-efficient method for LLM guardrailing. It outperforms existing approaches with 100-1000x lower parameter overhead, enabling on-device content moderation.
• 20 May 2024: Excited to announce that I've joined Samsung R&D Institute (UK) as a Principal Research Engineer ⏳🧐🤖🚀🙏🏻. Thanks to everyone at Amazon for the last two years.
• 3 March 2023: I passed my PhD viva with no reviewable corrections 🥳! Thanks be to God 🙏🏻; to my supervisors Walid Magdy, Bonnie Webber, and Maria Wolters; and to my examiners Alexandra Birch-Mayne and Rada Mihalcea. Check out my thesis, Computational Sarcasm Detection and Understanding in Online Communication.
• 4 April 2022: Our patent, Processing communications in a computing arrangement for semantic understanding and interpretation of code-switching, by Sourav Dutta, Silviu Vlad Oprea, Salama Hitham, and Hu Peng, was published.
• 30 June 2021: The flood segmentation model that we built at Frontier Development Lab has now been deployed by SpaceX on an actual satellite 🛰. Along the way, we collaborated with the European Space Agency and UNICEF. Our work was covered by this post from the University of Oxford; and by several media outlets: 1, 2, 3, 4, 5. Check out our Nature (Scientific Reports) paper and the video of the rocket launch 🚀

See more news here.

Work​

• 2024 - present: Principal Research Engineer at Samsung R&D Institute UK

  I've recently been working on LLM guardrailing. For instance, check out our paper, LoRA-Guard: Parameter-Efficient Guardrail Adaptation for Content Moderation of Large Language Models (arxiv; under review); we introduce a parameter-efficient method for LLM guardrailing. It outperforms existing approaches with 100-1000x lower parameter overhead, enabling on-device content moderation.

• 2022 - 2024: Applied Scientist at Amazon Alexa AI

  I've worked on improving the ability of large language models (LLMs) to generate responses that would provide Alexa customers with a more delightful experience.

• 2021: Applied Scientist (Intern) at Amazon Alexa AI

  At Amazon, I worked with Elisabeth Kwan, Molly Xia, Christos Christodoulopoulos, Dave Palfrey, and Stephen Teskey on language generation using language models.

• 2020: Research Scientist (Intern) at Huawei

  At Huawei, I worked with Haytham Assem and Sourav Dutta on learning transformations between monolingual word embedding spaces, to enable unsupervised translation and transfer learning to low-resource languages. Check out our COLING 2022 paper based on this work.

• 2019: Researcher at Frontier Development Lab

  At Frontier Developemnt Lab, we built a flood segmentation model. In the process, we collaborated with the European Space Agency and UNICEF. The model has now been deployed by SpaceX on an actual satellite 🛰.

  Our work was covered by this post from the University of Oxford; and by several media outlets: 1, 2, 3, 4, 5. Check out our Nature (Scientific Reports) paper and the video of the rocket launch 🚀

• 2014 - 2017: Engineer at VisualDNA and TheySay

  During this time, I was an engineer at two tech startups. First, a software engineer at VisualDNA, a data science and management platform, where I worked on data aggregation and reporting using Scala and the Scalding interface to Hadoop. After VisualDNA, I spent some time as a contractor. Next, I was an artificial intelligence engineer at TheySay, a startup providing text analytics services, where I used technologies such as Scala and MongoDB.

  Both startups were acquired, see this article about VisualDNA, and this one about TheySay.

• 2012: Guest Researcher at the National Institute for Standards and Technology

  I worked with Bruce Miller on extending LaTeXML, a TeX parser that he wrote in Perl. The goal of my extenssion was to convert TikZ graphics to SVG. See this paper that mentions my work.

Education​

• 2018 - 2023: PhD in Data Science at the University of Edinburgh

  Check out my thesis, Computational Sarcasm Detection and Understanding in Online Communication.

  In summary, I used computational methods to detect and understand the phenomenon of sarcasm, as it is manifested in online textual communication, together with my supervisors, Walid Magdy, Bonnie Webber, and Maria Wolters.

  More specifically, I built a dataset of texts annotated for sarcasm (ACL 2020 paper), introduced sarcasm detection models (ACL 2019 paper), and also organised a competition encouraging the community to build such models (SemEval 2022 paper). I showed that people of similar socio-demographic backgrounds understand each other's sarcasm more often than people of dissimilar backgrounds (CSCW 2022 paper). Finally, I built a sarcastic chatbot (EMNLP 2021 demo), and investigated when it is appropriate for chatbots to be sarcastic, and how they should formulate their utterances (ACL 2022 paper).

  Along the way, I had fun as an intern at Frontier Development Lab in 2019 (20201 Nature (Scientific Reports) paper), at Huawei in 2020 (COLING 2022 paper), and at Amazon Alexa AI in 2021 (paper in the baking 👨🏻‍🍳). See below, in the Work section.

• 2017 - 2018: MRes in Data Science at the University of Edinburgh

  I used computational methods to detect the presence of sarcasm in tweets, together with my supervisor, Walid Magdy.

• 2012 - 2013: MSc in Computer Science at the University of Oxford

  I worked with Phil Blunsom on building character-level language models for the Romanian language using recurrent neural networks.

• 2009 - 2012: BSc in Computer Science at Jacobs University Bremen

  This is where my interest in natural language processing was triggered, working with Michael Kohlhase.

Teaching​

• 2021: Lab demonstrator for Text Technologies in Data Science at the University of Edinburgh.
• 2010 and 2011: Teaching assistant for Programming in C/C++ at Jacobs University Bremen.

Media coverage​

Our paper, LoRA-Guard: Parameter-Efficient Guardrail Adaptation for Content Moderation of Large Language Models, under review, has been covered by:

• MarcTechPost: Samsung Researchers Introduce LoRA-Guard: A Parameter-Efficient Guardrail Adaptation Method that Relies on Knowledge Sharing between LLMs and Guardrail Models
• Blog post: Introducing LoRA-Guard: A Breakthrough in AI Content Moderation

Our paper, Towards global flood mapping onboard low cost satellites with machine learning, published in Nature (Scientific Reports) in 2021, was covered by the following articles:

• University of Oxford: Artificial Intelligence pioneered at Oxford to detect floods launches into space
• The Watchers: WorldFloods – AI pioneered at Oxford for global flood mapping launches into space
• Innovation News Network: A look at historic breakthroughs in flood mapping from space
• Homeland Security News Wire: Detecting Floods from Space Using Artificial Intelligence
• Chinese Academy of Sciences: AI加持遥感技术能否为防汛"备料"
• China Science Communication: AI加持，遥瞰洪涛

Patents​

European patent office

• Processing communications in a computing arrangement for semantic understanding and interpretation of code-switching

  Sourav Dutta, Silviu Vlad Oprea, Haytham Assem, and Hu Peng

  Patent WO2022069030A1 issued from application PCT/EP2020/077336. 2022.

  ABShtml

  A method of processing a communication in a computing arrangement that increases accuracy of semantic understanding and improves meaningful interpretation of code- switched regions in the communication. The method includes using the computing arrangement to analyze the communication to identify a predominant language used in the communication, and also to identify one or more code switched regions occurring in the communication. The method further includes using an artificial intelligence (AI) engine of the computing arrangement to translate the one or more languages used in the respective one or more code switched regions into one or more equivalent expressions of the predominant language. The one or more code switched regions of the communication are then replacing or supplemented with the one or more equivalent expressions included into the communication.

My Google Patents page is here.

Publications​

Safety and Bias

• LoRA-Guard: Parameter-Efficient Guardrail Adaptation for Content Moderation of Large Language Models

  Hayder Elesedy, Pedro M. Esperança, Silviu Vlad Oprea, and Mete Ozay

  Under review. 2024.

  ABSBIBpdf

  Guardrails have emerged as an alternative to safety alignment for content moderation of large language models (LLMs). Existing model-based guardrails have not been designed for resource-constrained computational portable devices, such as mobile phones, more and more of which are running LLM-based applications locally. We introduce LoRA-Guard, a parameter-efficient guardrail adaptation method that relies on knowledge sharing between LLMs and guardrail models. LoRA-Guard extracts language features from the LLMs and adapts them for the content moderation task using low-rank adapters, while a dual-path design prevents any performance degradation on the generative task. We show that LoRA-Guard outperforms existing approaches with 100-1000x lower parameter overhead while maintaining accuracy, enabling on-device content moderation.

  @misc{elesedy2024loraguardparameterefficientguardrailadaptation,
        title={LoRA-Guard: Parameter-Efficient Guardrail Adaptation for Content Moderation of Large Language Models}, 
        author={Hayder Elesedy and Pedro M. Esperança and Silviu Vlad Oprea and Mete Ozay},
        year={2024},
        eprint={2407.02987},
        archivePrefix={arXiv},
        primaryClass={cs.LG},
        url={https://arxiv.org/abs/2407.02987}, 
  }
  

Figurative language comprehension

• Sarcasm Detection is Way Too Easy! An Empirical Comparison of Human and Machine Sarcasm Detection

  Ibrahim Abu Farha, Steven Wilson, Silviu Vlad Oprea, and Walid Magdy

  Findings of the Association for Computational Linguistics. 2022.

  ABSBIBpdf

  Recently, author-annotated sarcasm datasets, which focus on intended, rather than perceived sarcasm, have been introduced. Although datasets collected using first-party annotation have important benefits, there is no comparison of human and machine performance on these new datasets. In this paper, we collect new annotations to provide human-level benchmarks for these first-party annotated sarcasm tasks in both English and Arabic, and compare the performance of human annotators to that of state-of-the-art sarcasm detection systems. Our analysis confirms that sarcasm detection is extremely challenging, with individual humans performing close to or slightly worse than the best trained models. With majority voting, however, humans are able to achieve the best results on all tasks. We also perform error analysis, finding that some of the most challenging examples are those that require additional context. We also highlight common features and patterns used to express sarcasm in English and Arabic such as idioms and proverbs. We suggest that to better capture sarcasm, future sarcasm detection datasets and models should focus on representing conversational and cultural context while leveraging world knowledge and common sense.

  @inproceedings{abu-farha-etal-2022-sarcasm,
      title = {Sarcasm Detection is Way Too Easy! An Empirical Comparison of Human and Machine Sarcasm Detection},
      author = {Abu Farha, Ibrahim and Wilson, Steven and Oprea, Silviu Vlad and Magdy, Walid},
      booktitle = {Findings of the Association for Computational Linguistics: EMNLP 2022},
      month = dec,
      year = {2022},
      address = {Abu Dhabi, United Arab Emirates},
      publisher = {Association for Computational Linguistics},
      url = {https://aclanthology.org/2022.findings-emnlp.387},
      doi = {10.18653/v1/2022.findings-emnlp.387},
      pages = {5284--5295},
  }
  
• SemEval-2022 Task 6: iSarcasmEval, Intended Sarcasm Detection in English and Arabic

  Ibrahim Abu Farha, Silviu Vlad Oprea, Steven Wilson, and Walid Magdy

  Proceedings of the 16th International Workshop on Semantic Evaluation (SemEval-2022). 2022.

  ABSBIBpdfvideo

  iSarcasmEval is the first shared task to target intended sarcasm detection: the data for this task was provided and labelled by the authors of the texts themselves. Such an approach minimises the downfalls of other methods to collect sarcasm data, which rely on distant supervision or third-party annotations. The shared task contains two languages, English and Arabic, and three subtasks: sarcasm detection, sarcasm category classification, and pairwise sarcasm identification given a sarcastic sentence and its non-sarcastic rephrase. The task received submissions from 60 different teams, with the sarcasm detection task being the most popular. Most of the participating teams utilised pre-trained language models. In this paper, we provide an overview of the task, data, and participating teams.

  @inproceedings{abu-farha-etal-2022-semeval,
      title = {{S}em{E}val-2022 Task 6: i{S}arcasm{E}val, Intended Sarcasm Detection in {E}nglish and {A}rabic},
      author = {Abu Farha, Ibrahim and Oprea, Silviu Vlad and Wilson, Steven and Magdy, Walid},
      booktitle = {Proceedings of the 16th International Workshop on Semantic Evaluation (SemEval-2022)},
      month = jul,
      year = {2022},
      address = {Seattle, United States},
      publisher = {Association for Computational Linguistics},
      url = {https://aclanthology.org/2022.semeval-1.111},
      doi = {10.18653/v1/2022.semeval-1.111},
      pages = {802--814},
  }
  
• iSarcasm: A Dataset of Intended Sarcasm

  Silviu Vlad Oprea, and Walid Magdy

  Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics. 2020.

  ABSBIBpdfvideo

  Online social networks (OSN) play an essential role for connecting people and allowing them to communicate online. OSN users share their thoughts, moments, and news with their network. The messages they share online can include sarcastic posts, where the intended meaning expressed by the written text is different from the literal one. This could result in miscommunication. Previous research in psycholinguistics has studied the sociocultural factors the might lead to sarcasm misunderstanding between speakers and listeners. However, there is a lack of such studies in the context of OSN. In this paper we fill this gap by performing a quantitative analysis on the influence of sociocultural variables, including gender, age, country, and English language nativeness, on the effectiveness of sarcastic communication online. We collect examples of sarcastic tweets directly from the authors who posted them. Further, we ask third-party annotators of different sociocultural backgrounds to label these tweets for sarcasm. Our analysis indicates that age, English language nativeness, and country are significantly influential and should be considered in the design of future social analysis tools that either study sarcasm directly, or look at related phenomena where sarcasm may have an influence. We also make observations about the social ecology surrounding sarcastic exchanges on OSNs. We conclude by suggesting ways in which our findings can be included in future work.

  @inproceedings{oprea-magdy-2020-isarcasm,
      title = {i{S}arcasm: A Dataset of Intended Sarcasm},
      author = {Oprea, Silviu Vlad and Magdy, Walid},
      booktitle = {Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics},
      month = jul,
      year = {2020},
      address = {Online},
      publisher = {Association for Computational Linguistics},
      url = {https://aclanthology.org/2020.acl-main.118},
      doi = {10.18653/v1/2020.acl-main.118},
      pages = {1279--1289},
  }
  
• Exploring Author Context for Detecting Intended vs Perceived Sarcasm

  Silviu Vlad Oprea, and Walid Magdy

  Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics. 2019.

  ABSBIBpdfvideo

  We investigate the impact of using author context on textual sarcasm detection. We define author context as the embedded representation of their historical posts on Twitter and suggest neural models that extract these representations. We experiment with two tweet datasets, one labelled manually for sarcasm, and the other via tag-based distant supervision. We achieve state-of-the-art performance on the second dataset, but not on the one labelled manually, indicating a difference between intended sarcasm, captured by distant supervision, and perceived sarcasm, captured by manual labelling.

  @inproceedings{oprea-magdy-2019-exploring,
      title = {Exploring Author Context for Detecting Intended vs Perceived Sarcasm},
      author = {Oprea, Silviu Vlad and Magdy, Walid},
      booktitle = {Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics},
      month = jul,
      year = {2019},
      address = {Florence, Italy},
      publisher = {Association for Computational Linguistics},
      url = {https://aclanthology.org/P19-1275},
      doi = {10.18653/v1/P19-1275},
      pages = {2854--2859},
  }
  

Computational social science

• Should a Chatbot be Sarcastic? Understanding User Preferences Towards Sarcasm Generation

  Silviu Vlad Oprea, Steven Wilson, and Walid Magdy

  Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics. 2022.

  ABSBIBpdfvideo

  Previous sarcasm generation research has focused on how to generate text that people perceive as sarcastic to create more human-like interactions. In this paper, we argue that we should first turn our attention to the question of when sarcasm should be generated, finding that humans consider sarcastic responses inappropriate to many input utterances. Next, we use a theory-driven framework for generating sarcastic responses, which allows us to control the linguistic devices included during generation. For each device, we investigate how much humans associate it with sarcasm, finding that pragmatic insincerity and emotional markers are devices crucial for making sarcasm recognisable.

  @inproceedings{oprea-etal-2022-chatbot,
      title = {Should a Chatbot be Sarcastic? Understanding User Preferences Towards Sarcasm Generation},
      author = {Oprea, Silviu Vlad and Wilson, Steven and Magdy, Walid},
      booktitle = {Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
      month = may,
      year = {2022},
      address = {Dublin, Ireland},
      publisher = {Association for Computational Linguistics},
      url = {https://aclanthology.org/2022.acl-long.530},
      doi = {10.18653/v1/2022.acl-long.530},
      pages = {7686--7700},
  }
  
• The Effect of Sociocultural Variables on Sarcasm Communication Online

  Silviu Vlad Oprea, and Walid Magdy

  Proceedings of the ACM on Human-Computer Interaction. 2020.

  ABSBIBpdfhtml

  Online social networks (OSN) play an essential role for connecting people and allowing them to communicate online. OSN users share their thoughts, moments, and news with their network. The messages they share online can include sarcastic posts, where the intended meaning expressed by the written text is different from the literal one. This could result in miscommunication. Previous research in psycholinguistics has studied the sociocultural factors the might lead to sarcasm misunderstanding between speakers and listeners. However, there is a lack of such studies in the context of OSN. In this paper we fill this gap by performing a quantitative analysis on the influence of sociocultural variables, including gender, age, country, and English language nativeness, on the effectiveness of sarcastic communication online. We collect examples of sarcastic tweets directly from the authors who posted them. Further, we ask third-party annotators of different sociocultural backgrounds to label these tweets for sarcasm. Our analysis indicates that age, English language nativeness, and country are significantly influential and should be considered in the design of future social analysis tools that either study sarcasm directly, or look at related phenomena where sarcasm may have an influence. We also make observations about the social ecology surrounding sarcastic exchanges on OSNs. We conclude by suggesting ways in which our findings can be included in future work.

  @article{oprea-magdy-2020-the-effect,
      author = {Oprea, Silviu Vlad and Magdy, Walid},
      title = {The Effect of Sociocultural Variables on Sarcasm Communication Online},
      year = {2020},
      issue_date = {May 2020},
      publisher = {Association for Computing Machinery},
      address = {New York, NY, USA},
      volume = {4},
      number = {CSCW1},
      url = {https://doi.org/10.1145/3392834},
      doi = {10.1145/3392834},
      journal = {Proceedings of the ACM on Human-Computer Interaction},
      month = may,
      articleno = {29},
      numpages = {22},
      keywords = {online communication, sarcasm, social media, sociocultural background},
      }
  

Controllable text generation

• Chandler: An Explainable Sarcastic Response Generator

  Silviu Vlad Oprea, Steven Wilson, and Walid Magdy

  Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing: System Demonstrations. 2021.

  ABSBIBpdfvideo

  We introduce Chandler, a system that generates sarcastic responses to a given utterance. Previous sarcasm generators assume the intended meaning that sarcasm conceals is the opposite of the literal meaning. We argue that this traditional theory of sarcasm provides a grounding that is neither necessary, nor sufficient, for sarcasm to occur. Instead, we ground our generation process on a formal theory that specifies conditions that unambiguously differentiate sarcasm from non-sarcasm. Furthermore, Chandler not only generates sarcastic responses, but also explanations for why each response is sarcastic. This provides accountability, crucial for avoiding miscommunication between humans and conversational agents, particularly considering that sarcastic communication can be offensive. In human evaluation, Chandler achieves comparable or higher sarcasm scores, compared to state-of-the-art generators, while generating more diverse responses, that are more specific and more coherent to the input.

  @inproceedings{oprea-etal-2021-chandler,
      title = {Chandler: An Explainable Sarcastic Response Generator},
      author = {Oprea, Silviu Vlad and Wilson, Steven and Magdy, Walid},
      booktitle = {Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing: System Demonstrations},
      month = nov,
      year = {2021},
      address = {Online and Punta Cana, Dominican Republic},
      publisher = {Association for Computational Linguistics},
      url = {https://aclanthology.org/2021.emnlp-demo.38},
      doi = {10.18653/v1/2021.emnlp-demo.38},
      pages = {339--349},
  }
  

Machine translation

• Multi-Stage Framework with Refinement Based Point Set Registration for Unsupervised Bi-Lingual Word Alignment

  Silviu Vlad Oprea, Sourav Dutta, and Haytham Assem

  Proceedings of the 29th International Conference on Computational Linguistics. 2022.

  ABSBIBpdf

  Cross-lingual alignment of word embeddings are important in knowledge transfer across languages, for improving machine translation and other multi-lingual applications. Current unsupervised approaches relying on learning structure-preserving transformations, using adversarial networks and refinement strategies, suffer from instability and convergence issues. This paper proposes BioSpere, a novel multi-stage framework for unsupervised mapping of bi-lingual word embeddings onto a shared vector space, by combining adversarial initialization, refinement procedure and point set registration. Experiments for parallel dictionary induction and word similarity demonstrate state-of-the-art unsupervised results for BioSpere on diverse languages – showcasing robustness against variable adversarial performance.

  @inproceedings{oprea-etal-2022-multi,
      title = {Multi-Stage Framework with Refinement Based Point Set Registration for Unsupervised Bi-Lingual Word Alignment},
      author = {Oprea, Silviu Vlad and Dutta, Sourav and Assem, Haytham},
      booktitle = {Proceedings of the 29th International Conference on Computational Linguistics},
      month = oct,
      year = {2022},
      address = {Gyeongju, Republic of Korea},
      publisher = {International Committee on Computational Linguistics},
      url = {https://aclanthology.org/2022.coling-1.92},
      pages = {1089--1097},
  }
  

Computer vision

• Towards global flood mapping onboard low cost satellites with machine learning

  Gonzalo Mateo-Garcia*, Joshua Veitch-Michaelis*, Lewis Smith*, Silviu Vlad Oprea, Guy Schumann, Yarin Gal, Atılım Güneş Baydin, and Dietmar Backes

  Nature (Scientific Reports). 2021.

  ABSBIBhtml

  Spaceborne Earth observation is a key technology for flood response, offering valuable information to decision makers on the ground. Very large constellations of small, nano satellites— ’CubeSats’ are a promising solution to reduce revisit time in disaster areas from days to hours. However, data transmission to ground receivers is limited by constraints on power and bandwidth of CubeSats. Onboard processing offers a solution to decrease the amount of data to transmit by reducing large sensor images to smaller data products. The ESA’s recent PhiSat-1 mission aims to facilitate the demonstration of this concept, providing the hardware capability to perform onboard processing by including a power-constrained machine learning accelerator and the software to run custom applications. This work demonstrates a flood segmentation algorithm that produces flood masks to be transmitted instead of the raw images, while running efficiently on the accelerator aboard the PhiSat-1. Our models are trained on WorldFloods: a newly compiled dataset of 119 globally verified flooding events from disaster response organizations, which we make available in a common format. We test the system on independent locations, demonstrating that it produces fast and accurate segmentation masks on the hardware accelerator, acting as a proof of concept for this approach.

  @article{Mateo-Garcia2021,
      author = {Mateo-Garcia, Gonzalo and Veitch-Michaelis, Joshua and Smith, Lewis and Oprea, Silviu Vlad and Schumann, Guy and Gal, Yarin and Baydin, At{i}l{i}m G{"u}ne{c{s}} and Backes, Dietmar},
      title = {Towards global flood mapping onboard low cost satellites with machine learning},
      journal = {Scientific Reports},
      year = {2021},
      month = mar,
      day = {31},
      volume = {11},
      number = {1},
      pages = {7249},
      issn = {2045-2322},
      doi = {10.1038/s41598-021-86650-z},
      url = {https://doi.org/10.1038/s41598-021-86650-z},
      }
  

* indicates equal contribution. Check the full list of publications on my Google Scholar profile.

Talks​

This list does not include conference presentations of my papers.

• Talk about my work on sarcasm detection and understanding at Oakland University, MI, USA (online).
• Talk about my work on sarcasm detection and understanding at the Technical University of Cluj-Napoca, Romania (online).