A Research On Methods Of Knowledge Acquisition From Domain-Specific Texts And Their Application In Knowledge Acquisition From Archaeological Texts

In the current Internet age, the volume, sources, and forms of information are undergoing essential changes, due to information explosion and globalization. How to intelligently process the massive information on Web pages has become a very urgent problem. Transformation from information into knowledge and transition from information infrastructure to knowledge infrastructure are important research areas in the 21st century. Both of the two tasks aim at achieving the goal?getting the"Right Information"to the"Right People"in the"Right Language"in the"Right Timeframe"at the"Right Level of Granularity".Currently, one of the hot issues is text mining, which is to extract useful information from free texts. It raises new challenges and exigent needs for knowledge acquisition from text (KAT). The dissertation addresses the methods of domain-specific concept acquisition, domain-specific hyponymy learning, descriptive stream extraction, and instance knowledge acquisition, and their applications in knowledge acquisition from archaeological texts. The main contributions of this dissertation are summarized as follows:(1) Proposing a hybrid approach to extracting domain-specific concepts. We introduce the criteria, which are used to identify domain-specific concepts. A domain-specific concept is not only a word, but also an entity concept, a qualitative concept, or a relational concept. This hybrid approach makes use of rules, statistic, syntactic, and semantic information about texts to identify concepts. It introduces main verbs and semantic roles to extract concepts. We propose a corpus-based learning approach to identifying the main verbs, and a semantic role recognition method on the basis of maiv verbs and knowledge acquisition oriented semantic models. Comparative exrepriemts show that the proposed approach can give satisfactory results(2) Proposing a multi-strategy method of learning domain-specific hyponymous relations. We design three approaches to extracting the hyponymous relations, namely a seed ontology driven approach, a context-mediated approach, and a concept-building–based approach. These methods are implemented by an ontology learning agent. The agent is composed of known knowledge, learning conditions, and learning knowledge, and it has strong extensibility, because it s expressive language is the frame and first-order logic.(3) Proposing an ontology-driven approach to identifying descriptive streams from domain-specific texts. As a new text processing and analysis task, descriptive stream extraction is proposed to identify the topics of texts, topics'descriptive aspects and the corresponding orders. Thus, descriptive stream can benefit Individuals knowledge acquisition. We propose an ontology-driven approach to extracting descriptive streams of texts, and whose result serves for knowledge acquisition. Experimental results show...