I am a spatial data scientist, with research and contributions cutting across geographic information retrieval, machine learning, geovisualization, and visual analytics. I conduct integrative research that brings together data science with social/environmental science to inform best practices for a more sustainable and equitable society. My research primarily focuses on method development, spanning various domains including social media analytics, energy (resilience and production), crisis management, situational awareness, precision agriculture, and digital humanities; and I am always excited about applying my expertise in other domains.
A large portion of my research has focused on developing methods for enabling the use of geographical information embedded in textual data, a field of research formally known as Geographic Information Retrieval (GIR). GIR integrates GIScience, information extraction/retrieval, natural language processing (NLP), and spatial indexing/search for (geographic) data extraction, storage, analysis and visualization. While the overwhelming majority of all data in digital form exists as text, unstructured text has not been well-supported by either GIScience theories or existing GIS tools.
While computational, my approach to research and development is human-centered, from visual/system design to algorithm integration and evaluation, to domain deployment and field studies. My visualization-related work in the past couple of years has focused on visual analytics for human-in-the-loop machine learning aiming to (1) develop flexible, performant computational methods leveraging human expertise for dynamic situations (that do not lend themselves to one-off training/deployment), and (2) helping users understand machine learning methods output/biases when applied to geospatial data, falling under what is commonly known as explainable artificial intelligence. Examples include our recent projects on (a) human-in-the-loop learning of topic-relevance in social media data for real-time situational awareness, and (b) interactive feature exploration/selection in hyperspectral imagery using domain knowledge for building optimized regression methods for forecast precision agriculture.
I have developed visual analytics and GIR methods extracting and disambiguating place references in (social media) text in a scalable manner to support situational awareness in crisis management. My dissertation research culminated in two systems and one annotated dataset. Most notably, I developed GeoTxt, a geoparsing software that identifies and geolocates place references in text, which has been used by multiple research projects in various universities. In addition, I developed GeoAnnotator, an interactive semi-automatic annotation system for creating geo-labeled datasets for training/evaluation of machine learning models or spatial linguistics studies.
More recently, my text- and GIR-related work includes interactive learning of topic relevance in social media data, interactive identification of social (media) spambots and troll campaigns, geolocation estimation of social media posts, and developing spatio-textual embeddings for GIR.
Prospective students: I am actively recruiting students. Please feel free to reach out to me with your CV, explaining how your research background and expertise may align with my work, and what areas you'd be interested in working on in the future.
Recent Courses Taught
- Spring 2020 GEOG 3023 Statistics and Geographic Data
- Fall 2019 GEOG 3023 Statistics and Geographic Data
Snyder, L., Lin, Y., Karimzadeh, M., Goldwasser, D., & Ebert, D. S. (in press, 2020). Interactive Learning for Identifying Relevant Tweets to Support Real-time Situational Awareness. IEEE Transactions on Visualization and Computer Graphics.
Khayat, M., Karimzadeh, M., Zhao, J, Ebert, D. S. (in press, 2020). VASSL: A Visual Analytics Toolkit for Social Spambot Labeling. IEEE Transactions on Visualization and Computer Graphics.
Zhao, J., Karimzadeh, M., Masjedi, A., Wang, T., Zhang, X., Crawford, M. M., & Ebert, D. S. (2019). Interactive Feature Selection and Exploration of Regression Models for Hyperspectral Images. IEEE VIS 2019 Conference.
Snyder, L. S., Karimzadeh, M., Chen, R., & Ebert, D. S. (2019). City-level Geolocation of Tweets for Real-time Visual Analytics. Proceedings of the 3rd ACM SIGSPATIAL International Workshop on AI for Geographic Knowledge Discovery, 85–88.
Zhao, J., Karimzadeh, M., Xu, H., Malik, A., Afzal, S., Wang, G., Elmqvist, E., & Ebert, D. (in press, 2020). Route Packing: Geospatially-Accurate Visualization of Route Networks. In Proceedings of the 53rd Hawaii International Conference on System Sciences.
Karimzadeh, M., MacEachren, A.M. (2019). GeoAnnotator: A Collaborative Semi-Automatic Platform for Geo-Annotation. The ISPRS International Journal of Geo-Information special issue on Human-Centered Geovisual Analytics and Visuospatial Display Design.
Karimzadeh, M., Pezanowski, S., Wallgrün, J. O., MacEachren, A. M., Wallgrün, J. O. (2019). GeoTxt: A scalable geoparsing system for unstructured text geolocation. Transactions in GIS, 23(1), 118–136.
Wallgrün, J.O., Karimzadeh, M., Pezanowski, S., MacEachren, A.M. (2018). GeoCorpora: Building a Corpus to Test and Train Microblog Geoparsers. International Journal of Geographic Information Science,Volume 32, Issue 1, 1-29.
Wallgrün, J.O., Klippel, A., and Karimzadeh, M. (2015). Towards Contextualized Models of Spatial Relations. Proceedings of the 9th Workshop on Geographic Information Retrieval,Ross S. Purves and Christopher B. Jones (Eds.) Paris, France. ACM, New York, NY, USA.