About me and my research

I am PhD candidate in Computational Linguistics. My research is in grounded language generation and understanding with neural language models. The aim is to combine linguistic representations with uncertain perceptual representations in a single framework.

I think fusion of different modalities is one of the core questions in modelling meaning. In a sense, understanding is about changing several representations or mapping them into an actionable representation. Modelling language essentially deals with high level representations of concepts and their internal meanings in relation with each other. When we deal with grounded language in visual perception, several different levels of representations for both vision and language models is required. This is why neural language models are powerful tools to encode knowledge about the world as representations of concepts. This knowledge must be able to explain “when” and “where” system needs “what” kind of composition and fusion of representations. Modelling spatial language deals with all these aspects. This is why the focus of my research is mainly on spatial language.

In addition to my research as PhD student, I have been TA for following courses in Master in Language Technology (MLT) programme in last three/four years:

  • Artificial intelligence: Cognitive systems (Formerly: “Embodied and Situated Language Processing”) (November-January)
  • Statistical Methods for NLP (January to March)
  • Computational Semantics (April-June)

List of publications

My google scholar page might have the latest updated list. Here is an extended list on GU library. Or, maybe you find the following easier direct access to additional materials and codes:

  • Ghanimifard, Mehdi, and Simon Dobnik. “What goes into a word: generating image descriptions with top-down spatial knowledge.” Proceedings of the 12th International Conference on Natural Language Generation. 2019. pdf webpage
  • Ek, Adam, and Mehdi Ghanimifard. “Synthetic Propaganda Embeddings to Train a Linear Projection.” Proceedings of the Second Workshop on Natural Language Processing for Internet Freedom: Censorship, Disinformation, and Propaganda. 2019. pdf
  • Santín, José Miguel Cano, Simon Dobnik, and Mehdi Ghanimifard. “Interactive visual grounding with neural networks.” pdf
  • Ghanimifard, Mehdi, and Simon Dobnik. “What a neural language model tells us about spatial relations.” Proceedings of the Combined Workshop on Spatial Language Understanding (SpLU) and Grounded Communication for Robotics (RoboNLP). 2019. pdf code
  • Ghanimifard, Mehdi, and Simon Dobnik. “Knowing when to look for what and where: Evaluating generation of spatial descriptions with adaptive attention.” Proceedings of the European Conference on Computer Vision (ECCV). 2018. pdf code
  • Bizzoni, Yuri, and Mehdi Ghanimifard. “Bigrams and BiLSTMs Two neural networks for sequential metaphor detection.” Proceedings of the Workshop on Figurative Language Processing. 2018. pdf code
  • Dobnik, Simon, Mehdi Ghanimifard, and John Kelleher. “Exploring the functional and geometric bias of spatial relations using neural language models.” (2018). pdf code
  • Bizzoni, Yuri, Stergios Chatzikyriakidis, and Mehdi Ghanimifard. ““Deep” Learning: Detecting Metaphoricity in Adjective-Noun Pairs.” Proceedings of the Workshop on Stylistic Variation. 2017. pdf code webpage
  • Ghanimifard, Mehdi, and Simon Dobnik. “Learning to Compose Spatial Relations with Grounded Neural Language Models.” IWCS 2017-12th International Conference on Computational Semantics-Long papers. 2017. pdf code
  • Ghanimifard, Mehdi, and Richard Johansson. “Enriching Word Sense Embeddings with Translational Context.” Proceedings of the International Conference Recent Advances in Natural Language Processing. 2015. pdf code