Term Function Recognition Of Academic Text

In this article, the definition of term function classification is thoroughly stud-ied. A term function framework in academic text is constructed, which combines both domain-independent lexical functions and domain-specific lexical functions. On the ba-sis of this framework, this paper establishes two methods to automatically recognize the term function in academic text. Considering the domain applicability of term function, the idea of open information extraction is introduced in this paper to investigate the appropriate open information extraction methods applied to academic text analysis.Specifically, the major contributions of this paper are as follows:1. Academic text term function is defined and the mechanism of term function is discussed. Based on the above discussion, a framework is built which combined both domain-independent lexical functions and domain-specific lexical functions.2. Based on condition random fields, a method for recognizing problem function and method function in academic text is proposed in this article. The experiment result shows the effectiveness of this method. Furthermore, a verb roles clustering method based on word2vec is proposed. This method can cluster the terms according to their functional role presented in context.3. An automatic extraction method based on ranking model is established to recognize the core issues and core methods in academic text. A rapid method for constructing labeled data based on English-Chinese parallel corpus is proposed.4. An open information extraction method and system implementation, EXVerb, is raised for academic text. EXVerb identifies the relational data in the academic text via term recognition, syntactic analysis, concept network structure, and relation generation. Comparing to other systems, EXVerb can improve recall significantly on the premise of ensuring approximate precision.Although the research content of this paper plays a certain role in promoting the study of term function analysis, further research is needed to solve some certain prob- lems:1. How to further improve the performance of term function recognition. The precision and recall of the current method is still in relatively low level, so there is much room for improvement. The exploration for new method is needed to improve the precision and recall of automatic recognition.2. How to conduct application of term function analysis. This article only provides an attempt among all the possible applications of academic text term function analysis. Therefore, how to better apply this technique and idea will be a direction for future research.3. In this article, the modeling of the relationship among the academic terms is not included in the term function framework. More exploration from the perspective of term relationship is needed in support of the deep semantic analysis in academic text.4. This article introduces an open information extraction method for academic text. However, this is only a simple attempt. Further study is still needed to improve the performance of the extraction.