Dr Ziqi Zhang
My research addresses methods that enable machines to extract human knowledge from text, to represent such knowledge in a structured representation that is understandable and usable by machines. This ultimately enhances our capability of processing and sense-making of very large scale data, improving decision making. Specifically, this includes but is not limited to:
- Information Extraction: how to automatically turn unstructured, natural language text into structured representation that could support machine understandability and reasoning. This could include the extraction of terms, concepts, named entities, and relations between them from texts.
- Disambiguation: how to teach machines to automatically infer the meaning of a word or phrase within certain context.
- Lexical semantics: how to represent the ‘meaning’ of a word, name, phrase, or sentence; how to measure the relatedness and similarity of these meanings (semantic relatedness and similarity).
- Knowledge base construction: the use of all the above technology in the automatic creation of structured ‘databases’ that support machine understandability and reasoning; and methods of mapping such knowledge bases (ontology alignment, ontology mapping). An example of a knowledge base is the Google Knowledge Graph, or DBpedia.
- Semantic Web and Linked Data: the use of all the above technology to enable the vision of tomorrow’s Web where machine understandable data are put on the Web, shared and reused across application, enterprise, and community boundaries.
Current PhD students
Abdulkareem Alqusair: Product category extraction and linking in the area of semantic web
Daisy Da Moura Semedo: Mining health information on the Social Web: towards an understanding of the influence of social media on public healthcare
Omaima Fallatah: Mapping and aligning large Knowledge Bases
Paul Fenn: Social Media as a tool to enhance Higher Education learning and teaching experiences
Phil Webster: Semantic Web For Knowledge Management
Zhixue Zhao: Learning from unbalanced data and limited data for automated hate speech detection