Towards a Flexible System Architecture for Automated Knowledge Base Construction Frameworks

摘要

Although knowledge bases play an important role in many domains (including in archives, where they are sometimes used for entity extraction and semantic annotation tasks), it is challenging to build knowledge bases by hand. This is owing to a number of factors: Knowledge bases must be accurate, up-to-date, comprehensive, and as flexible and as efficient as possible. These requirements mean a large undertaking, in the form of extensive work by subject matter experts (such as scientists, programmers, archivists, and other information professionals). Even when successfully engineered, manually built knowledge bases are typically one-off, use-case-specific, non-standardized, hard-to-maintain solutions.Recent advances in the field of automated knowledge base construction (AKBC) offer a promising alternative. A knowledge base construction framework takes as input source documents (such as journal articles containing text, figures, and tables) and produces as output a database of the extracted information.An important motivation behind these frameworks is to relieve domain experts from having to worry about the complexity of building knowledge bases. Unfortunately, such frameworks fall short when it comes to scalability (ingesting and extracting information at scale), extensibility (ability to add or modify functionality), and usability (ability to easily specify information extraction rules). This is partly because these frameworks are often constructed with relatively limited consideration for architectural design, compared to the attention given to algorithmic performance and low-level optimizations.As knowledge bases will be increasingly relevant to many domains, we present a scalable, flexible, and extensible architecture for knowledge base construction frameworks. As a work in progress, we extend a specific framework to address some of its design limitations. The contributions presented in this short paper can shed a light on the suitability of using AKBC frameworks for computational use cases in this domain and provide future directions for building improved AKBC frameworks.