Named Entity Recognition for the Horticultural Domain

Abstract

Named Entity Recognition (NER) is one of the fundamental tasks in natural language processing and knowledge engineering, as well as a prerequisite step of many downstream applications. Horticulture, a major branch of agricultural science, means the cultivation, processing, and sale of fruit, nuts, vegetables, and ornamental plants as well as numerous additional services. NER for the horticulture domain means to find key biological traits and state-of-Art experimental methods for the horticulturists, new cultivation methods and useful tools for farmers, as well as other important information for planners and policy makers to trigger decision-making procedures. In this paper we designed an NER tagging-set of 7 fine-grained types, and since there is no publicly shared annotated corpus available in horticulture domain, we constructed training and testing corpora manually. Thus, we realized Horticulture NER in scientific literature abstracts with CRF method. Results showed that our system's accuracy and precision were satisfactory but still have room for improvement.