MASC (Manually Annotated Sub-Corpus)

手动注释子语料库 (MASC) 由大约 500,000 个单词的当代美国英语书面和口语数据组成，这些数据来自美国开放国家语料库 (OANC)。所有 MASC 都包括人工验证的句子边界、token、lemma 和 POS 注释；名词和动词块；命名实体（人、地点、组织、日期）； Penn Treebank 语法；共指；和话语结构。 MASC 项目已经为部分子语料库生成了额外的手动生成或验证的注释，包括 FrameNet 框架元素的全文注释和带有 WordNet 3.1 感知标签的 100K+ 句子语料库，其中十分之一也被注释为FrameNet 框架元素。其他项目为各种其他语言现象提供了全部或部分子语料库的注释，包括 PropBank、TimeBank、Pittsburgh opinion 和其他几个项目。与大多数免费提供的包含各种语言注释的语料库不同，MASC 包含来自广泛流派的均衡文本选择。

The Manually Annotated Sub-Corpus (MASC) consists of approximately 500,000 words of written and spoken contemporary American English data sourced from the Open American National Corpus (OANC). The entire MASC corpus includes manually verified sentence boundaries, tokens, lemmas, and part-of-speech (POS) annotations; noun and verb chunks; named entities (persons, locations, organizations, dates); Penn Treebank syntax; coreference; and discourse structure. The MASC project has generated additional manually curated or validated annotations for select sub-corpora, including full-text annotations of FrameNet frame elements and a corpus of over 100,000 sentences annotated with WordNet 3.1 sense tags, one-tenth of which are also annotated for FrameNet frame elements. Additional projects have provided annotations for full or partial sub-corpora covering a wide range of other linguistic phenomena, including PropBank, TimeBank, Pittsburgh Opinion, and several other initiatives. Unlike most freely available corpora with diverse linguistic annotations, MASC features a balanced selection of texts drawn from a broad range of genres.

搜集汇总

背景与挑战

背景概述

MASC（手动注释子语料库）是一个包含约50万单词的当代美国英语数据集，源自美国开放国家语料库（OANC）。该数据集提供了人工验证的多种语言注释，如句子边界、词性标注、命名实体和语法结构，并涵盖广泛流派的文本，以支持语言学研究。