TAILIEUCHUNG - Báo cáo khoa học: "Analysis and Repair of Name Tagger Errors"

Name tagging is a critical early stage in many natural language processing pipelines. In this paper we analyze the types of errors produced by a tagger, distinguishing name classification and various types of name identification errors. We present a joint inference model to improve Chinese name tagging by incorporating feedback from subsequent stages in an information extraction pipeline: name structure parsing, cross-document coreference, semantic relation extraction and event extraction. We show through examples and performance measurement how different stages can correct different types of errors | Analysis and Repair of Name Tagger Errors Heng Ji Ralph Grishman Department of Computer Science New York University New York NY 10003 UsA hengj i@ grishman@ Abstract Name tagging is a critical early stage in many natural language processing pipelines. In this paper we analyze the types of errors produced by a tagger distinguishing name classification and various types of name identification errors. We present a joint inference model to improve Chinese name tagging by incorporating feedback from subsequent stages in an information extraction pipeline name structure parsing cross-document coreference semantic relation extraction and event extraction. We show through examples and performance measurement how different stages can correct different types of errors. The resulting accuracy approaches that of individual human annotators. 1 Introduction High-performance named entity NE tagging is crucial in many natural language processing tasks such as information extraction and machine translation. In traditional pipelined system architectures NE tagging is one of the first steps in the pipeline. NE errors adversely affect subsequent stages and error rates are often compounded by later stages. However Roth and Yi 2002 2004 and our recent work have focused on incorporating richer linguistic analysis using the feedback from later stages to improve name taggers. We expanded our last year s model Ji and Grishman 2005 that used the results of coreference analysis and relation extraction by adding feedback from more information extraction components -name structure parsing cross-document coreference and event extraction - to incrementally re rank the multiple hypotheses from a baseline name tagger. While together these components produced a further improvement on last year s model our goal in this paper is to look behind the overall performance figures in order to understand how these varied components contribute to the improvement and compare the remaining .

• Báo cáo khoa học
• Analysis and Repair
• Name Tagger Errors
• Ralph Grishman
• báo cáo khoa học
• mô hình ngôn ngữ
• xử lý ngôn ngữ tự nhiên
• centralized
• intermediate
• repair
• facilities
• conus
• analysis
• air
• space
• expeditionary
• forces
• supporting
• Water absorption
• Constraint stress analysis
• Polymer modified cement mortar
• Patch repair material
• Material selected is polymer modified
• BMC Pediatrics
• Anastomotic stricture
• Birth weight
• Esophageal atresia repair
• Meta analysis
• Risk factors
• BMC Pulmonary Medicine
• Epithelial cells
• Wound repair
• Cluster analysis
• Pathway analysis
• BMC Cancer
• Prostate cancer
• Androgen receptor
• DNA repair
• DNA damage response
• BMC Musculoskeletal Disorders
• Rotator cuff
• Cuff repair
• Single row
• Single row knotless techniques
• Endometrial cancer
• Bioinformatics analysis
• Prognostic model
• BMC Anesthesiology
• Iatrogenic tracheal rupture
• Surgical repair
• Perioperative management
• Airway management
• One lung ventilation
• Interrupted aortic arch
• One stage repair
• Selective cerebral perfusion
• Kaplan Meier survival analysis
• Left ventricular outflow tract obstruction
• BMC Molecular and Cell Biology
• Comet assay
• Genome wide expression
• Network analysis
• Inguinal hernia repair
• Spinal anesthesia
• General anesthesia
• Numerical Analysis
• Augmented Locking
• Plate Fixation
• Proximal Humeral Fractures
• Comminuted fractures
• Lung cancer
• Base excision repair
• DNA damaging chemotherapy
• Physiochemical analysis
• Saudi Arabia
• Egyptian prescriptions including honey
• Tissues for wound repair
• Suppress inflammation
• BMC Bioinformatics
• DNA double strand breaks
• ImageJ plugin
• Automated analysis
• Ionizing radiation
• Radiation biology
• Joint capsule
• Total hip arthroplasty
• Hip diseases
• Colorectal cancer
• Lynch syndrome
• Mismatch repair
• Hereditary colorectal cancer
• Multivariate analysis
• Annular closure device
• Annular repair
• Recurrent disc herniation
• Lumbar intervertebral disc
• Disc herniation
• báo cáo y học
• công trình nghiên cứu về y học
• tài liệu về y học
• cách trình bày báo cáo
• trình bày báo cáo
• tài liệu báo cáo khoa học
• tạp chí y học Medical Sciences
• nghiên cứu y học
• Breast cancer
• Oestrogen receptor
• Multivariate survival analysis
• Clinicopathological factors
• Radiation induced apoptosis
• Biological heterogeneity
• MLPA assay
• Triple negative breast cancer
• Comparative genomic hybridization
• Progesterone receptor
• MSH6 Mutation