Font Size: a A A

Research And Implementation Of Anaphora Resolution

Posted on:2018-07-07Degree:MasterType:Thesis
Country:ChinaCandidate:Q ZhangFull Text:PDF
GTID:2348330521451643Subject:Software engineering
Abstract/Summary:PDF Full Text Request
In recent years,the development of Chinese information processing is no longer confined to the simple facts,the research is more complex problems in aspects of speech comprehension,information extraction,text summarization and other applications are used in coreference resolution,can carry out accurate coreference resolution,it is particularly important in natural language processing.From 2015 to 2017,Shanxi University is in charge of the national "863" project in the language of solving problem and answer generation key technology and system research,mainly responsible for reading text based on semantic understanding,in the process of solving the problem of coreference resolution is based on the research of the task.Based on this demand,this paper aims to study and solve the anaphora resolution problem in reading comprehension.This paper is divided into the following three parts:(1)To using the N-Gram language model to solve the problem of anaphora resolutionTo refer to,must consider the collocation relationship between words and words,so this article use N-gram language model for research.Research is the main steps of the replaced with refer to language will be refer to language,and then replace with model to calculate the probability of the sentence,after the highest probability of words or phrases as final refer to language.Because calculation result to be highlighted and the restriction of space consumption,this paper finally chose the binary language model,namely the Bi-gram language model.(2)The maximum entropy model is used to solve the problem of anaphora resolutionAfter using N-gram language model research,found that the results is not ideal,study the causes,and in the use of N-gram language model,did notuse the syntactic and semantic information,lead to point to not clear.So in the next study,using the maximum entropy model,the semantic information as a feature.Study of the main steps is calculated to eliminate to be consistent in characteristics can the probability to determine the final resolution,use in this article refers to the characteristics of a consistency,plurality,gender consistency,distance attribute,syntactic features and semantic features of five.(3)The Construction of anaphora resolution systemIn this paper,the use of the above two methods,the construction of a system of anaphora resolution,the system uses the language of the provincial college entrance examination reading comprehension.It provides the technical support for solving the problem of reading comprehension.
Keywords/Search Tags:Anaphora Resolution, N-Gram, Maximum Entropy
PDF Full Text Request
Related items