Information extraction on novel text using machine learning and rule-based system

摘要

Novel consists of around 30,000 to 50,000 words in total. It usually tells a story about entities and its relation one another such as, Person, Location or Organization. In order to apprehend those information, reading the whole novel is compulsory. However, it is a time-consuming task. This research proposes a solution - automatic extraction of entity relation by means of Information Extraction (IE) technique. This technique is divided into two steps. First, all the entities are retrieved from the text input, by using Named Entity Recognition (NER). Afterward, all relations is extracted by Relation Extraction (RE) process. This research implements an IE system to both NER and RE, which employs supervised machine learning approach combined with rule-based system. The main purpose of this research is to determine which features and algorithm of the machine learning are adequate to acquire the best result, and which rules are the most suitable for novel characteristics.