LocalView Public Meetings Database
收藏资源简介：
LocalView is a database co-created by Soubhik Barari and Tyler Simko to advance the study of local government in the United States. It is the largest existing dataset of local government public meetings— the central policy-making process in American local government — as they are captured on video. To get started, select the file(s) that you'd like for your use case based on the year that the meeting took place. Note: we are no longer supporting file formats other than .parquet for space considerations. For potential use cases or further guidance on downloading the data in bulk, visit the companion website: localview.net. Change Log Scraping, parsing, identifying, and merging together meetings involves a large number of non-trivial decisions, many of which need to be adjusted over time particularly as the YouTube API changes. As such, when such decisions notably deviate from process or the outputs documented in the first version of this database, it will be logged here. Version 2.0 (2023-10) Data updated up until September 2023. ~10,000 new videos added, all belonging to existing channels in database. Change in data format: channelType column changed to channel_type. ST_FIPS correctly padded to be 7 characters (2 digit state code + 5 digit place FIPS code). videos with no caption available from YouTube are explicitly marked as “” in caption_text. caption_text_cleaned is actually consistently cleaned (previously stray timestamps/pause markers in some entries). acs_2018_* columns now prefixed as acs_18. additional ACS variables now available for each place: acs_18_median_gross_rent: Median gross rent in FIPS place. acs_18_median_hh_inc: Median household income in FIPS place. acs_18_median_age: Median age in FIPS place. acs_18_amind: American Indian population in FIPS place. acs_18_asian: Asian population in FIPS place. acs_18_nhapi: Native Hawaiian or Pacific Islander population in FIPS place. census_2015_* columns removed for redundancy. to avoid confusion and possible conflicts, .json and .csv file formats eliminated in favor of .parquet format. municipal voteshare (_dem2pv) variables have been removed from the public use files for a number of reasons: (1) high degree of missingness, (2) no columns estimates available, (3) potential sensitivity The matching process of videos to ST-FIPS and government types is as follows: videos are matched to channels’ previous ST-FIPS codes and/or government types if there is an umambiguous match, otherwise a (1) series of regex matches with video title/description are used to attempt to match video to the government and (2) state/place names are extracted from each video’s title/description/caption and used to match to an ST-FIPS entity if an unambiguous match; only identified videos are then uploaded to the database. to identify the date of the meeting that the video captures, we first try to extract the date from the title, otherwise we try to extract the date from the description, otherwise it is discarded. Version 1.0 (2023-02) See publication for full details on methodology choices for the Version 1.0 database.
LocalView 是由 Soubhik Barari 与 Tyler Simko 联合创建的数据库，旨在推动美国地方政府治理研究。它是目前已公开的规模最大的美国地方政府政务会议视频数据集——而政务会议正是美国地方政府的核心决策流程载体。 用户可根据会议举办年份选择适配自身应用场景的文件。请注意，出于存储空间考量，我们不再支持 .parquet 以外的文件格式。若需了解潜在应用场景或获取批量下载数据的进一步指导，请访问配套网站：localview.net。 ### 变更日志 会议的爬取、解析、识别与合并流程涉及大量关键决策，其中诸多内容需随时间调整，尤其是在 YouTube 应用程序编程接口（YouTube API）更新的场景下。因此，当此类决策与本数据库首版的流程说明或输出结果出现显著偏差时，相关变更将在此处记录。 #### 2.0 版本（2023年10月） 数据更新至2023年9月，新增约10000条视频，均来自数据库内已有的政务频道。 数据格式变更：将`channelType`字段更名为`channel_type`。 修正`ST-FIPS`字段的填充格式，使其统一为7位字符（2位州代码 + 5位区县FIPS代码）。 YouTube无字幕的视频，其`caption_text`字段将显式标注为空字符串。 `caption_text_cleaned`字段现已实现标准化清洗（此前部分条目存在残留时间戳与暂停标记的问题）。 原`acs_2018_*`系列字段现统一前缀为`acs_18_`。 新增针对各区县的多项美国社区调查（ACS）变量： - `acs_18_median_gross_rent`：对应区县的住房总租金中位数 - `acs_18_median_hh_inc`：对应区县的家庭收入中位数 - `acs_18_median_age`：对应区县的人口年龄中位数 - `acs_18_amind`：对应区县的美洲印第安人口数 - `acs_18_asian`：对应区县的亚裔人口数 - `acs_18_nhapi`：对应区县的夏威夷原住民及太平洋岛民人口数 移除冗余的`census_2015_*`系列字段。 为避免混淆与潜在冲突，现已取消`.json`与`.csv`格式支持，仅保留`.parquet`格式。 出于多重原因，公共使用文件中已移除市政选举得票率（`_dem2pv`）相关变量：(1) 缺失值占比极高；(2) 无可用估算列；(3) 存在潜在敏感性问题。 视频与ST-FIPS代码及政府类型的匹配流程如下：若视频所属频道已有明确的ST-FIPS代码或政府类型，则直接匹配；否则，(1) 通过正则表达式匹配视频标题与描述，尝试将视频关联至对应政府主体；(2) 从视频标题、描述或字幕中提取州/区县名称，若能匹配到唯一的ST-FIPS实体，则完成关联。仅完成识别的视频才会被上传至数据库。 对于视频所记录的会议日期，我们优先从视频标题中提取日期，若失败则尝试从视频描述中提取，若仍无法获取则丢弃该视频。 #### 1.0 版本（2023年2月） 关于1.0版本数据库的方法学选择的完整细节，请参阅相关出版物。
