Projects
From words to meaning: Association networks and semantic processing KU Leuven
The project consists of three parts: (1) to elaborate the existing word association network, (2) to investigate the semantic properties of a network that is constructed from the word association data set and (3) to explore neuropsychological applications of word association data. The first part consists of further developing and enlarging the world’s largest word association data set, in several languages. In the second part, the network’s ...
A Network Based Account of Word Processing and Semantic Cognition KU Leuven
In this project, we propose to study word meaning using a network based on word associations. We want to enlarge our existing word association sets in Dutch and in English, which are already the world’s largest sets in both languages, to the size of an adult’s mental lexicon. Using these data sets, we want to derive three kinds of models for word processing and semantic cognition. The first model is a simple (first-order) associative ...
Single-trial N400 ERP detection for semantic network formation applied to neuromarketing. KU Leuven
N400 studies have been conducted to assess whether consumers judge a new product as member of a known brand (i.e., brand extension). However, as these studies rely in their analysis on predefined mental categories (brands, product categories): what happens when they are not predefined and even context-dependent? Can we measure with the N400 the effect of marketing stimuli on mental category formation? The aim of this project is to to answer ...
Performance Improvement Strategies for Semantic Segmentation KU Leuven
This thesis focuses on semantic segmentation. It presents methods to improve semantic segmentation, such as context-aware padding, error correction, model selection and the use of temporal information.
In terms of padding for convolutional operation, standard convolutional neural networks use padding to maintain a consistent shape for feature maps. Zero padding is simple and efficient but the additional zeros that are generated by zero ...
Study and design of a federated, context-aware and self-learning reasoning framework enabling scalable and efficient advanced semantic reasoning on IoT data streams Ghent University
Internet of Things (IoT) lapels to the ever-growing network of objects connected to the Internet. That Cisco predicted by 2020 the Internet will contain about 50 billion smart objects. Intelligently processing the data produced by synthesis objects will lead to a wealth of advanced applications in domains like smart cities, pervasive health & amp; environmental sensing. The generated data is voluminous, heterogeneous, time-varying and ...
Computational Analysis of Semantic Change Across Different Environments KU Leuven
This network aims to train early-stage researchers to develop and apply innovative methodologies for Computational Analysis of Semantic Change Across Different Environments (CASCADE), i.e. to identify, analyse and interrogate how meaning is expressed in language in diverse contexts, with a shared focus on the impact of time (diachronic text analytics).
CASCADE responds to a skills deficit within the academic, public and commercial ...
Enhancing data analytics for IoT by enabling semantic enrichment of machine learning tasks Ghent University
The recent spread of sensors, actuators and mobile devices, comprising the Internet of Things (IoT), provides ample opportunity to improve our quality of life through data analytics. However, as IoT data is bound by the four Vs of Big Data—volume, variety, velocity, and veracity—deriving meaningful insights becomes challenging. Today, two approaches have been employed side by side. Relying on knowledge graphs (KGs) and logical rules, ...
Semantic Image Analysis Towards a Visual Turing Machine. KU Leuven
In the recent years visual recognition systems have progressed with leaps and bounds in their accuracy. The vision of this research proposal is to go beyond the existing, accurate, yet rough visual recognition machines, and approach an ideal Visual Turing Machine, namely a machine that would pass as a human when analyzing visual data. We identify 3 directions for this vision. First, beyond generic predictions a Visual Turing Machine should ...