uspto-trademarks-features
收藏资源简介:
该数据集名为“USPTO 商标标记特征(衍生)”,是从伴生银级语料库中提取的美国专利商标局(USPTO)商标申请数据的客观特征列集合。每行对应一个唯一的商标申请,可通过serial_no字段关联到原始数据集illeatmyhat/uspto-trademarks。特征列包括12个关键字段:coined_mark_score(衡量商标字符序列统计异常性的新造词得分)、coined_token_max(最大令牌异常值)、single_token(指示商标是否为单一词)、is_standard_char(标识是否为标准字符绘图)、is_marketplace_class(判断是否属于亚马逊相关类别,如第9、11、18、20、21、25、28类)、filing_basis(申请基础类型,如使用、意向、外国、马德里)、final_refusal_in(反映是否发出最终驳回的法律程序状态)、abandoned_no_use_in(反映是否因使用声明缺失/缺陷而被放弃的法律程序状态)、owner_country(申请人国家代码)、goods_item_count(商品/服务项目数量)和filing_year(申请年份)。这些特征综合覆盖了商标的文字特性、申请属性、法律结果和基本信息,适用于商标分析、特征工程、市场趋势研究或机器学习任务(如商标分类或异常检测)。数据以Parquet格式存储,采用美国政府作品(公共领域)许可证。
The dataset is named USPTO Trademark Mark Features (Derived), which is a collection of objective feature columns extracted from the companion silver-level corpus, focusing on United States Patent and Trademark Office (USPTO) trademark application data. Each row corresponds to a unique trademark application (linked to the original trademark dataset illeatmyhat/uspto-trademarks via the serial_no field), with each feature column representing a measurement of the application itself, not subjective judgments about the applicant. The data includes 12 key feature fields: coined_mark_score and coined_token_max (measuring statistical anomalies in the entire trademark character sequence), single_token (indicating whether the trademark is a single word), is_standard_char (identifying if it is a standard character drawing), is_marketplace_class (determining if it belongs to Amazon-related classes, such as classes 9, 11, 18, 20, 21, 25, 28), filing_basis (application basis types, e.g., use, intent, foreign, Madrid), final_refusal_in and abandoned_no_use_in (reflecting legal procedural statuses for final refusal and abandonment due to missing/defective statements of use), owner_country (applicant country code), goods_item_count (number of goods/services items), and filing_year (application year). These features comprehensively cover the textual characteristics, application attributes, legal outcomes, and basic information of trademarks, making the dataset suitable for trademark analysis, feature engineering, market trend research, or machine learning tasks (such as trademark classification or anomaly detection). The data is stored in Parquet format and uses the U.S. Government Work (Public Domain) license.
USPTO Trademark Mark Features (Derived) 数据集概述
基本信息
- 数据集名称: USPTO Trademark Mark Features (Derived)
- 许可证: US Government Work(美国政府作品,公共领域)
- 许可证链接: https://www.usa.gov/government-works
数据集内容
该数据集是从配套的银级语料库中派生的、面向每个商标的特征列。每行对应一个申请(通过 serial_no 与 illeatmyhat/uspto-trademarks 数据集关联)。每个列都是对申请的测量,而非对任何申请人的判断。
主要特征列
| 列名 | 含义 |
|---|---|
coined_mark_score |
整个商标的字符n-gram非概率性,范围[0,1](1表示最像随机/造词) |
coined_token_max |
商标各词元的最大n-gram非概率性,范围[0,1](可捕捉到造词词元与普通词组合的情况) |
single_token |
商标是否为单个词 |
is_standard_char |
是否为标准字符图形 |
is_marketplace_class |
是否在亚马逊密集类别(9/11/18/20/21/25/28)中提交 |
filing_basis |
申请基础(使用/意图/外国/马德里) |
final_refusal_in |
是否发出最终驳回(公开的审查历史事实) |
abandoned_no_use_in |
是否因缺少/有缺陷的使用声明而放弃 |
owner_country |
申请人国家代码(公开记录,可能包含杂质) |
goods_item_count |
列出的商品/服务项目数量 |
filing_year |
申请年份 |
数据配置
- 配置名称:
mark_features - 数据文件:
*.parquet
使用方法示例
sql -- 例如:查找近期基于使用、在亚马逊密集类别中提交且造词分数高的商标 SELECT serial_no, mark_id_char, coined_mark_score FROM hf://datasets/illeatmyhat/uspto-trademarks-features/*.parquet WHERE coined_mark_score > 0.85 AND is_marketplace_class AND filing_basis = use AND filing_year >= 2015 ORDER BY coined_mark_score DESC;
方法论与许可
coined_mark_score衡量商标字母序列的统计异常性;其他列是申请的直属性。- 方法论详情请参见:https://github.com/illeatmyhat/uspto-trademark-radar/blob/main/docs/ANALYSIS.md
- 许可证:美国政府作品(公共领域)





