Knowledge mapping of medicinal plants enable ordinary people to differentiate between medicinal plants and learn their pharmacological effects, provide assistances and instructions to medical workers during the use of medicinal plants, and support intelligent queries of the properties of traditional medicinal plants. This paper innovatively introduces artificial neural network to the knowledge mapping of medicinal plants, and provides a practical and valuable reference for scientific development and reasonable use of medicinal plants. Firstly, the entity relationships were designed for medical knowledge map, and the definitions, scales, and examples were given for each type of data in the proposed knowledge map of medicinal plants. Next, the authors detailed the ideas of multi-source knowledge fusion, and the acquisition and storage strategies for entity information of medicinal plants. Then, the attention-based bidirectional gated recurrent network was combined with convolutional neural network to detect the genetic relationships between medicinal plants from the angles of semantics and texts. Finally, this paper explains the semantic retrieval algorithm for medicinal plants, and visualizes the knowledge map. The proposed model and semantic retrieval algorithm were proved effective and superior through experiments. It is concluded that: The smaller the batch size, the higher the recognition accuracy of plant entities, and the better the recognition effect. The research findings provide a reference for knowledge mapping in other fields.