Dissertation > Excellent graduate degree dissertation topics show

Answer Selection and Answer Validation for Complex Chinese Question Answering

Author: XuBingZuo
Tutor: ZhongYiXin
School: Beijing University of Posts and Telecommunications
Course: Pattern Recognition and Intelligent Systems
Keywords: Answering System Sentence Similarity Computing Information entropy The answer screening Answer Verification
CLC: TP391.6
Type: Master's thesis
Year: 2009
Downloads: 85
Quote: 0
Read: Download Dissertation

Abstract


Answering system as one of the hot spots of natural language processing and information retrieval field, with high academic study and practical value. At present, the answering system gradually focus on the processing of complex issues. Complex question answering system from relevant documents often get similar or even the same answer. Given more of the same (more of the same) the answer does not improve system performance. This makes in order to achieve the de-emphasis as the answers to the primary purpose of the filters become necessary. In the same time, the candidate answer correctly or not need to go through further validation. Answers screening and verification of the factual issues (factoid) is already quite mature. However, due to complex problems factual questions the answer does not have a clear target answer type, the shorter length (usually a word, the characteristics of the phrases, numbers, time, or named entities) are usually longer has a certain grammatical structure of the sentence. Most screening methods do a thesaurus and simple rules classified analysis, verification through semantic dictionary type check, can not meet the complex processing needs of the candidate answers. This paper focuses on the screening and answer validation for the answer to the complex problem of the Chinese. For answers to screening, its basic modules sentence similarity calculation is based the HowNet Dictionary word similarity computation, and further improved edit distance-based sentence similarity calculation. Here, we propose improved edit distance method based on a weighted named entity. Thereafter, for the answer validation problem, the paper proposes a named entity frequency statistics in a set of candidate answers, information entropy to calculate the value of information, to answer rearrangement and verification. And which named entity information value is used to calculate the sentence similarity calculation of weight of the entity, the named entities occupy a high weight to calculate two sub-in value of the information to provide problem makes with the problem (Question) on the repeatability of the target to achieve the syntax layer and a certain degree of weight to pragmatic level. NTCIR-7 Q \u0026 A system task evaluation corpus, we will answer screening and verification algorithm based named entity frequency statistics based on network data-driven method comparison experiments. The results show that the proposed method in the defined class, the issue of the relationship class performance better than the comparison method, but the poor performance in the event class.

Related Dissertations

  1. The Urban Public Transport Evaluation in Level of Services Based on IE-UM,F570
  2. Research on Case-based Reasoning in Decision Support System,TP182
  3. The Innovation Research on the Evaluation of the Core Competence of Enterprise,F272
  4. Research on Chinese Question Answering System Model Base on Concept Expansion,TP391.1
  5. Research on the Technology of Automatic Term Extraction,TP391.1
  6. Edge Detection Based on Ant Colony Optimization and maximum dimensional information entropy image segmentation,TP391.41
  7. Restricted Domain Question Answering System Text Retrieval,TP391.1
  8. Research areas Automatic Entity Relation Extraction,TP391.3
  9. Research on the Methods of Crowd Density Estimation and Motion State Analysis Based on Video Sequence,TP391.41
  10. Data Reduction Algorithm Based on Information Entropy,TP18
  11. The Evaluation Study of High Technology Enterprise Organization Flexibility Based on A Theory of the Unascertained Measure,F224
  12. Consistency of judgment matrix based on discriminant method,O151.21
  13. Network Intelligent Answering System Research and Implementation,TP393.09
  14. Quantitative evaluation model based on information entropy of the classroom observation,G632.4
  15. Research on Wireless Perimeter Intrusion Monitoring System and Recognition of Its Intrusion Signals,TP393.08
  16. Based on non- parametric model for image segmentation method,TP391.41
  17. Forecasts and trusted routing method based on information entropy IP network abnormal behavior,TN915.06
  18. Jiangxi Public Security Cadres and E-learning System Design and Implementation of Communication,TP311.52
  19. Enhancement of K-nearest Neighbor Algorithm Based on Attribute Value Information Entropy,TP18
  20. Set Pair Analysis Methods for Risk Analysis of Flood Disaster,X43
  21. Researches on Methods for Video Logo Detection and Inpainting,TP391.41

CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Teaching machine, learning machine
© 2012 www.DissertationTopic.Net  Mobile