NomBank

“NomBank 是纽约大学的一个注释项目，与科罗拉多大学的 PropBank 项目相关。我们的目标是在 PropBank 语料库（华尔街日报语料库）中标记与名词同时出现的参数集Penn Treebank)，就像 PropBank 记录动词的此类信息一样。作为注释过程的副作用，我们正在生成许多其他资源，包括各种字典，以及称为框架文件的 PropBank 样式词汇条目。这些资源有助于用户用角色标记头部名词的各种参数和附属词（每个名词的每个意义的参数标签集）。纽约大学和科罗拉多大学正在协调努力，以确保角色定义在可能的情况下在各个部分保持一致比如我们在名词“decision”的注解中使用了PropBank的框架文件“decision”作为动词“decision”。然而我们的协调远不止于此。我们坚定地开始了这个项目在 Catherine Macleod 的 Nomlex 项目和支持动词的相关工作的肩膀上。结果证明这是一个很大的提升，因为语料库中大约一半的参数化名词是名词化或具有类名词化属性的名词（例如，“侵略”和“议程”具有类似于动词 ""destroy"" 和 ""schedule"")。 NomBank 迫使我们非常详细地定义名词论证结构，包括以前几乎没有研究的领域。因此，在许多方面，我们正在开辟新天地。我们涵盖的一些示例现象包括：支持动词结构（“John MADE A DECISION”）；跨系词的争论（““他的决定是离开”）；和括号内的 PP 结构（““应公司要求，昨天下午晚些时候，纽约和多伦多证券交易所的 Cineplex Odeon Corp. 股票交易暂停”）。这些现象中的每一个都很有趣，因为一个名词的论元可能出现在以该名词为首的 NP 之外。除了动词的名词化（""decision""、""helper""、""nominee""），我们还包括形容词的名词化（""incompetence""、""ability""、""wisdom"" )、关系名词（""president""、""friend""、""father""）、部分名词（""barrage""、""clump""、""variety""）和其他几种类型争论的名词。"

NomBank is an annotation project at New York University, affiliated with the PropBank project at the University of Colorado. Our goal is to annotate the sets of arguments co-occurring with nouns within the PropBank corpus (Wall Street Journal corpus, part of the Penn Treebank), just as PropBank records such information for verbs. As a side effect of the annotation process, we are generating numerous additional resources, including various dictionaries and PropBank-style lexical entries known as frame files. These resources enable users to annotate various arguments and adjuncts of head nouns with role labels, i.e., the set of argument labels for each sense of a given noun. NYU and the University of Colorado are coordinating efforts to ensure that role definitions are consistent across both projects wherever possible. For example, when annotating the noun "decision", we use PropBank's frame file for the verb "decision". Our coordination extends far beyond this, however. We built this project firmly on the shoulders of Catherine Macleod's Nomlex project and related work on supporting verbs. This proved to be a significant improvement, as approximately half of the argument-taking nouns in the corpus are nominalized nouns or nouns with nominal-like properties—for example, "aggression" and "agenda" exhibit argument structures similar to those of the verbs "destroy" and "schedule". NomBank has compelled us to define noun argument structures in extreme detail, including areas that had seen little prior research. As such, we are breaking new ground in many respects. Some exemplary phenomena we cover include: light verb constructions (e.g., "John MADE A DECISION"); cross-copular argument structures (e.g., "His decision is to leave"); and parenthetical prepositional phrase (PP) structures (e.g., "At the company's request, trading in shares of Cineplex Odeon Corp. on the New York and Toronto stock exchanges was suspended late yesterday afternoon"). Each of these phenomena is intriguing because the arguments of a noun can appear outside the NP headed by that noun. In addition to verbal nominalizations (e.g., "decision", "helper", "nominee"), we also cover adjectival nominalizations (e.g., "incompetence", "ability", "wisdom"), relational nouns (e.g., "president", "friend", "father"), partitive nouns (e.g., "barrage", "clump", "variety"), and several other types of argument-taking nouns.

搜集汇总

背景与挑战

背景概述

NomBank是纽约大学的一个注释项目，与PropBank项目相关，旨在对Penn Treebank语料库中的名词进行参数标记，类似于PropBank对动词的处理。注释过程产生了字典和框架文件等资源，覆盖名词化、关系名词等多种现象，并基于前期工作如Nomlex项目。