Machines that truly understand language would be incredibly useful and natural language processing (NLP) is making this possible. NLP algorithms are typically based on machine learning algorithms, which process, analyze, and act on vast amounts of data. However, deep learning-based NLP still suffers from serious shortcomings including poor interpretability (the degree to which humans can understand), inferior scalability, and reduced robustness.
To address the challenges of semantic gaps and data sparsity in large-scale NLP, Zhiyuan Liu, an Associate Professor from the Department of Computer Science at Tsinghua University, has explored representation learning methods for multiple units of natural language, including words, phrases, sentences, documents, networks and knowledge. He has built implicit representations with low-dimensional embedding vectors and knowledge representation with structured knowledge graphs. By taking advantage of information fusion of implicit representation and inference power of knowledge representation, he built a unified representation framework that bridges the gaps in data-driven deep learning and symbol-based linguistics, as well as world knowledge. This framework can support knowledge-guided NLP and is especially important for those knowledge-rich tasks such as reading comprehension, question answering, and text generation.
Liu has accomplished a number of innovative achievements and published several academic papers in leading AI and NLP conferences and journals. Moreover, he has released many popular open source packages for NLP, such as the Chinese lexical analyzer THULAC, the knowledge representation learning package OpenKE, and the network representation learning package OpenNE, which have attracted more than 10 thousand stars on GitHub and are used by hundreds of institutions and companies.
Liu believes that there is great potential for NLP development in China, but there are still too few researchers who devote themselves to natural language processing. As a doctoral advisor in the NLP field, he wants to help talented young students develop an academic and professional interest in the area. “Natural language processing is key to the realization of artificial intelligence. Human language is also full of unknowns. I hope more young students can join in and we can explore it together,” he said.