Identification of Plants assisted by Natural Language
Thesis submitted as a requirement for the degree of Bachelor of Engineering (Computer Engineering), UNSW. October 21, 2008.
This thesis project aims to design and implement the foundations of a website that will help people identify flowering plants. It will focus on improving the usability of existing systems by giving users the opportunity to record and make use of their own observations. In addition, the prototype is able to integrate dichotomous keys and standalone descriptions into a random-access interactive key. The system will process botanical descriptions with the help of an OWL ontology that integrates plant structures, botanical characteristics and the corresponding keywords. The extracted information is then used to semantically search for relevant descriptions. Automatic inference can also be done in easily understood cases.
There is an astounding variety of life on Earth and with 250 000 to 400 000 species the flowering plants comprise just a small fraction of it. Still, their diversity is immense and like all other lifeforms the world’s plants have been divided into hierarchical groups to make their study possible. The highest-level groups of flowering plants are known as families. These are subdivided into genera which are further subdivided into species. A plant’s genera and species names make up its scientific name. For example, Golden Wattle plants have the scientific name Acacia pycnantha. As with almost all other entities, giving each species a unique scientific name is the crucial first step in studying plants. Determining the groups to which a particular plant belongs (ultimately its scientific name) is called “identification”. It is of course of central importance to botany, but is also necessary in fields like ecology, environmental conservation, herbology and agriculture. Moreover, identifying plants lets people perceive them in finer detail than they otherwise would. This can then lead one to appreciate the diversity and complexity of nature to a greater extent. However, botanical identification can be quite challenging and presents several barriers, especially to novices and amateur enthusiasts. This project will attempt to address some of these challenges by adopting a fresh approach to the user-interface.
Please see submit/ThesisB.pdf :)