Encyclopedic Knowledge Based Question Auto-generation System

There is too much information in the internet as the amount of users gets a highspeed of increment. Many people get use to finding solutions of problems which theymeet not only in job but also in daily life by using the search engine. If they are nottrained how to use a search engine effectively or they do not know how to searchinformation in the internet, it will cost them a lot of time to find what they want orselect the goals from varieties of files. The interactive question answering system isdesigned to solve the problems described before. By importing some nature languageprocessing technologies, the question answering system can understand what the userare looking for and return an answer to the user after processing on the question.Question generation is able to help the question answering system in the field ofsupplying the pairs of questions and their answers when the question answering systemis lack of communications in that field. The field can be either specific or general. Thetechnology of generating questions automatically in English has achieved a high levelwhile the research in Chinese area is rare which has been part of question answeringsystem, dialogue system and tutoring system. This research is aimed at adding pairs ofquestion and answering to the database which is part of the knowledge system. We wantto achieve that goal by automatically generating questions from a given text.The paper will show how we generate questions automatically in the followings.1. The way to generate Chinese questions automatically by computers.Nowadays, the system of question generation can not deal with a whole sentencelike human beings do. The preprocessing of a sentence is necessary for every questiongeneration system. Our system is designed in a distributed way such that the functionsof information processing can be done by several parts to speed the system up. Thepreprocessing part is divided into two main classes which consist of seven kinds ofmaterials. We design an algorithm which combines the syntactic information and patternmatching to generate wh-questions.2. The automatic classification of generated questions.The generated question will be classified into six categories due to the namedentities recognized from a Chines sentence and the pattern. The questions are whoquestions, what time questions, where questions, what organization questions, whatquestions and why questions.3. System evaluation and improvingThere are a serial of evaluation standards for English system which can beborrowed to value our system. At the same time, we invited some user to test our systemand give their opinions for the system. We will improve our system based on these information.