Linguistic Linked Data

Linked-Data Based NLP Architecture

Deep Learning

The figure above gives an overview of the specific research topics organised by research area, showing a graphical overview of their inter-relations:

  • Entity relatedness is a research topic in the knowledge extraction area that exploits both unstructured information and structured information, combining distributional semantics relatedness derived from large amounts of text with relatedness information embedded on the DBpedia Ontology.
  • The taxonomy extraction research topic is concerned with analysing large amounts of unstructured text corpora to produce taxonomic structures that can be used for LOD organisation, generation, and enrichment.
  • Linked data profiling deals with the problem of producing comprehensive descriptions of LD datasets in order to improve the identification of suitable datasets for specific application scenarios.
  • Term and Knowledge Graph Translation is concerned with applying and advancing statistical or neural machine translation techniques for translating relatively limited linguistic descriptions of domain-specific terms or concepts that are available in monolingual or multilingual knowledge bases.
  • The suggestion mining research topic aims at analysing unstructured social media textual content to extract structured suggestions from reviews or other types of user-generated content.
  • Emotion analysis deals with extracting a more nuanced representation, covering dimensional as well as categorical emotion classification schemes.
  • Link discovery is the task of finding links between two datasets, for example linking a dictionary to Wikipedia. This is a vital first step for any system that relies on 'mash-ups' of datasets from different sources. We are developing a new system, NAISC, which applies semantic textual similarity systems and structural constraint optimization to discover links between heterogeneous datasets.
  • As computational activities and the Internet creates a wider multilingual and global community, under-resourced languages are acquiring political as well as economic interest thus creating the need to develop a new machine translation system.
  • Metaphor Analysis is a challenging task for a wide variety of applications such as social media analysis, sentiment analysis, machine translation and text simplification. The significant usage of metaphors communicated on social media makes studying metaphors in such a context of importance. Our research focuses on the computational processing of metaphors in social media. We are employing NLP and machine learning technologies to recognise and interpret metaphoric expressions.
  • Market sentiment analysis is dealing with understanding the evolution of sentiments within a given market, a highly complex process, and aiming at providing insights as well as quantifying the sentiment for each point in time.
  • Health Informatics focuses on collecting, storing, analysing, and presenting data to improve healthcare. Health information can come in many formats such as laboratory test results and clinical reports. Our research utilises unstructured text in the form of informal health reports originated from social networks, search queries, and discussion forums. The knowledge extracted from these sources is then applied to disease monitoring and surveillance.