Research

Recent Research Topics

My research interests are Big Data, Smart City, Urban Informatics, content-based and collaborative filtering and recommender systems, user-centric information retrieval, adaptive information retrieval, specifically personalised retrieval, implicit feedback-based search systems, affective, and sentiment information retrieval systems.

Interests

  • NeuroIS
  • Interactive Information Retreival
  • Recommender Systems
  • Mechanism Design
  • Urban Designs

Projects

  • Neuropsychological Aspect of Information Retrieval (NeuroIS)

    The basis of this work was to shed a light on the nature of key concepts in information retrieval, i.e. relevance and information need from a Neuropsychological point of view. From a practical point of view, this work can open a new research direction in detecting and prediction both information need and the relevance judgement of searchers by monitoring their brain activities which could lead to the next generation of information systems. The key contribution of this work was the introduction of the first neuropsychological model for the concept of relevance and information need. These three publications have won two best paper awards in ECIR 2013 and ACM SIGIR 2016. Within this work, I learnt how to design and conduct fMRI and EEG based experiments as well as how to process and analyse such big data.

    Sample Projects:

    • Yashar Moshfeghi, Frank E. Pollick. Search Process As Transitions Between Neural States. 40th Annual ACM World Wide Web Conference (WWW’18), Lyon, France, pages 1683-1692, April 2018.
    • Yashar Moshfeghi, Peter Triantafillou , Frank E Pollick. Understanding Information Need: an fMRI Study. 39th Annual ACM SIGIR Conference (SIGIR’16), Pisa, Italy, Pages 335-344, July 2016. (Best Paper Award).
    • Marco Allegretti, Yashar Moshfeghi, Maria Hadjigeorgieva, Frank E Pollick, Joemon M Jose, Gabriella Pasi. When Relevance Judgement is Happening?: An EEG-based Study. 38th Annual ACM SIGIR Conference (SIGIR’15), Santiago, Chile, pages 719-722, July 2015.
  • Multimedia & Social Media Analytics & IR

    As part of my Postdoctoral position, under LiMoSINe project funded under EC Seventh Framework Program scheme, I worked on the topic of multimedia and social media information retrieval. The focus of the project was to develop models and conduct experiments aiming to create a large-scale image (described in Paule, Sun and Moshfeghi, 2018) and Twitter collections (described in Paule, Moshfeghi, Jose and Thakuriah, 2017) as well as to incorporate semantics and contextual information to improve the effectiveness of image (described in Moshfeghi, Agarwal, Piwowarski and Jose, 2009; McParlane, Moshfeghi and Jose, 2014) and twitter retrieval systems (described in McParlane, Moshfeghi and Jose, 2014). Within this work, I learnt the process involved in creating test collections as well as challenges associated with information objects such as multimedia and social media. I have also learnt how to collaborate and supervise PhD students.

    Sample Projects:

    • Jorge David Gonzalez Paule, Yeran Sun, and Yashar Moshfeghi. On Fine-Grained Geolocalisation of Tweets and Real-Time Traffic Incident Detection. Journal of Information Processing and Management, 2018. (doi:10.1016/j.ipm.2018.03.011)
    • Jorge David Gonzalez Paule, Yashar Moshfeghi, Joemon M. Jose and, Piyushimita Thakuriah. On Fine-Grained Geolocalisation of Tweets. The 3rd ACM International Conference on the Theory of Information Retrieval (ICTIR’17), Amsterdam, Netherlands, pages 313-316. October 2017.
    • Philip J. McParlane, Yashar Moshfeghi, Joemon M. Jose. “Nobody comes here anymore, it’s too crowded”; Predicting Image Popularity on Flickr. Annual ACM ICMR Conference (ICMR’14), Glasgow, UK, April 2014.
  • Mechanism Design for IR Evaluation

    This work resulted out of a pragmatic need to gather relevance assessment from a crowdsource-based platform. The data gathered from such platforms are known to be noisy due to different worker characteristics. Drawing upon game theory models, we formulated a mechanism design in which risk-averse (poor performing) and risk-inclined (good performing) workers will get automatically identified and eliminated resulting in improving the quality of the data gathered. We have optimised our model via performing an exhaustive simulation (described in Moshfeghi, Rosero and Jose, 2016) and tested it in a crowdsource-based platform, i.e. Amazon Mechanical Turk (described in Moshfeghi, Pinto, Pollick and Jose, 2013).  Within this work, I learnt important concepts in game theory such as Nash and Bayesian equilibrium, got familiar with various types of game theatrical models exist both competitive and collaborative ones, and built a mechanism design inspired by Chicken game theory model.

    Sample Projects:

    • Yashar Moshfeghi, Alvaro F. H. Rosero, and Joemon M. Jose. Identifying Careless Workers in Crowdsourcing Platforms: A Game Theory Approach. 39th Annual ACM SIGIR Conference (SIGIR’16), Pisa, Italy, pages 857-860, July 2016.
    • Yashar Moshfeghi, Alvaro F. H. Rosero, and Joemon M. Jose. A Game Theory Approach for Effective Crowdsource Based Relevance Assessment. ACM Transactions on Intelligent Systems and Technology, 7(4), 55, 2016. (doi:1145/2873063)
  • Beyond Implicit Relevance Feedback

    The basis of this work was to investigate the possibility of considering affective and physiological signals, gathered from searchers through their search processes, as an indicator of their relevance judgment. Within this work, we considered facial expression as our affective signal. In order to capture physiological signals, we consider multiple sensory channels including heart rate monitoring, skin temperature, and neural activity. We also investigated whether affective and physiological signals can be used as a complementary source of information for behavioural signals (i.e. dwell time) to create a reliable signal for relevance judgement prediction (described in Moshfeghi, Rosero and Jose, 2016). We have also tried to use the relevance judgment inferred from affective signals to improve the recommender systems (Moshfeghi and Jose, 2013; Arapakis, Mosfeghi, Ren and Jose, 2009). Within this work, I learnt, modified and used various supervised machine-learning models such as SVM (linear or with Kernel), Decision Tree, Random Forest, Naïve Bayes, Logistic Regression, etc. (Moshfeghi, Rosero and Jose, 2016; Moshfeghi and Jose, 2013; Arapakis, Moshfeghi, Ren and Jose, 2009). In addition, I learnt how to design and conduct user-based experiments as well as how to process and analyse such data.

    Sample Projects:

    • Yashar Moshfeghi, Joemon M. Jose. An Effective Implicit Relevance Feedback Technique Using Affective, Physiological and Behavioural Features. 36th Annual ACM SIGIR Conference (SIGIR’13), Dublin, Ireland, pages 133-142, July 2013.
    • Ioannis Arapakis, Yashar Moshfeghi, Reede Ren, and Joemon Jose. Enriching User Profiling with Affective Features for the Improvement of a Multimodal Recommender System. Presented the ACM International Conference on Image and Video Retrieval (CIVR’09) in Island of Santorini, Greece, pages 21-29, July 2009.
    • Ioannis Arapakis, Yashar Moshfeghi, Reede Ren, David Hanh and Joemon Jose. Relevance Estimation by Facial Expression Analysis. Presented the International Conference on Multimedia and Expo (ICME’09), New York, USA, June 2009.
  • Role of Emotion in Information Retrieval (NLP, IR & Recommendation)

    These papers develop the notion of emotion need, emotion object and emotion relevance, which was the focus of my Doctoral Thesis. In order to investigate the concept of emotion need we have shown that how cognition, emotion, and interaction of searchers differ when they are engaged in search tasks with different intentions (such as seeking, finding, or entertainment) (Arapakis et al., 2009).  In order to investigate the concept of emotion object and emotion relevance, we extracted emotion from textual documents using Natural Language Processing techniques (described in Moshfeghi and Jose, 2013) and use this information as a criterion in novel relevance-matching algorithm specifically developed to improve the effectiveness of information retrieval models, particularly in two domains of collaborative filtering (Moshfeghi and Jose, 2013; Moshfeghi, Zuccon and Jose, 2011) and re-ranking search result page (Moshfeghi, Piwowarski and Jose, 2011). We also investigated the complementary aspect of emotion objects with semantic objects (Moshfeghi and Jose, 2013). Within this work, I learnt, modified and used various unsupervised machine-learning models such as KNN (with and without Gaussian Kernels) (Moshfeghi, Zuccon and Jose, 2011; Moshfeghi and Jose, 2011), Latent Dirichlet Allocation and parameter optimization using Boosted Tree (Moshfeghi and Jose, 2013). In addition, I learnt natural language processing and how to build a text-based emotion extraction system.

    Sample Projects:

    • Yashar Moshfeghi, Joemon M. Jose. On Cognition, Emotion, and Interaction Aspects of Search Tasks with Different Search Intentions. 22nd Annual ACM WWW Conference (WWW’13), Rio de Janeiro, Brazil, pages 931-942, May 2013.
    • Yashar Moshfeghi, Joemon M. Jose. On the Effectiveness of Emotion Extraction Techniques. 10th International Conference in the RIAO Series (OAIR’13), Lisbon, Portugal, pages 197-200, May 2013.
    • Yashar Moshfeghi, Benjamin Piwowarski and Joemon M. Jose. Handling Data Sparsity in Collaborative Filtering using Emotion and Semantic Based Features. Attended and presented the 34th Annual ACM SIGIR Conference (SIGIR’11) in Beijing, China, pages 625-634, July 2011.

Insight Lab

Sakrapee Paisalnan

Sakrapee Paisalnan

PhD Student

Follow Me
Mohamed Amine Belabbes

Mohamed Amine Belabbes

PhD Student

Dominika Michalkova

Dominika Michalkova

PhD Student

Zuzana Pinkosova

Zuzana Pinkosova

PhD Student

Muhammad Abubakar Alhassan

Muhammad Abubakar Alhassan

PhD Student

Soukaina Elaouad

Soukaina Elaouad

PhD Student

Alexandros Ioannidis

Alexandros Ioannidis

PhD Student

Kritsana Khiaomang

Kritsana Khiaomang

PhD Student

Insight Lab

If you are interested in what we are doing, please feel free to find out more in our lab website.