Coordination Network Toolkit

A small Python command line tool and set of functions for studying coordination networks in Twitter and other social media data.

Rationale and Background

Social media activity doesn't occur in a vaccuum. Individuals on social media are often taking part in coordinated activities such as protest movements or interest-based communities.

Social media platforms are also used strategically to boost particular messages in line with political campaign goals or for commercial profit and scamming. This involves multiple accounts posting or reposting the same content, repeatedly and within a short time window (e.g. within 1 minute).

This software provides a toolkit to detect coordinated activity on social media and to generate networks that map the actors and their relationships. It provides a general purpose toolkit for multiple types of coordinated activity on any type of social media platform.

Firstly, it includes functionality for co-tweeting and co-retweeting (Keller et al., 2019; Schafer et al., 2017), where accounts post exactly the same text (co-tweets) or repost the same post within a short time window (co-retweets). Secondly, it includes functionality for co-link analysis, where multiple accounts post the same URLs repeatedly and in a short time window of each other (Giglietto et al., 2020). Thirdly, it adds two new types of network types: co-reply, where accounts are replying to the same post repeatedly together; and co-similarity, where accounts post similar text (but not exact duplicates), which relaxes the strict assumption of co-tweeting.

Five types of coordination networks supported:

1. Co-retweet: reposting the same post

2. Co-tweet: posting identical text

3. Co-similarity: posting similar text (Jaccard similarity or user-defined)

4. Co-link: posting the same link

5. Co-reply: replying to the same post

Installation and Requirements:

This tool requires a working Python 3.6 (or later) environment.

This tool can be installed from pip - this will handle installing the necessary dependencies.

pip install coordination_network_toolkit

Once you have installed it, you can use the toolkit in one of two ways:

1. As a command-line tool (run compute_networks --help to find out how)

2. As a Python library (import coordination_network_toolkit)

For more information, see .

一款用于研究Twitter及其他社交媒体数据中协作网络的轻量Python命令行工具与函数集。

设计原理与研究背景

社交媒体活动并非孤立存在。社交媒体用户往往会参与各类协作活动，例如抗议运动或兴趣社群。

社交媒体平台也会被策略性地利用，以配合政治竞选目标推广特定信息，或是用于商业牟利与诈骗活动。此类行为通常表现为多个账号在极短时间窗口（例如1分钟内）反复发布或转发同一内容。

本软件提供了一套工具集，可检测社交媒体上的协作行为，并生成映射参与主体及其关联关系的网络。该工具集具备通用性，可用于分析任意类型社交媒体平台上的多种协作活动。

首先，工具支持共同发推（co-tweeting）与共同转发（co-retweeting）功能（Keller等人，2019；Schafer等人，2017）：即多个账号在短时间窗口内发布完全一致的文本（共同发推），或转发同一帖子（共同转发）。其次，工具支持链接共现分析（co-link analysis）功能，即多个账号在彼此间隔极短的时间内反复发布相同URL（Giglietto等人，2020）。此外，工具新增了两类新型网络分析模式：共同回复（co-reply），即多个账号协同反复回复同一帖子；以及文本相似共现（co-similarity），即多个账号发布相似但并非完全一致的文本，该模式放宽了共同发推的严格匹配假设。

本工具支持五类协作网络分析：
1. 共同转发（Co-retweet）：转发同一帖子
2. 共同发推（Co-tweet）：发布完全相同的文本
3. 文本相似共现（Co-similarity）：发布相似文本（支持雅卡尔（Jaccard）相似度或自定义相似度计算）
4. 链接共现（Co-link）：发布相同链接
5. 共同回复（Co-reply）：回复同一帖子

安装与环境要求

本工具需运行于Python 3.6及以上版本的有效环境中。

可通过pip进行安装，该方式将自动处理必要依赖的安装：
`pip install coordination_network_toolkit`

安装完成后，可通过两种方式使用该工具集：
1. 作为命令行工具（运行`compute_networks --help`查看使用方法）
2. 作为Python库（通过`import coordination_network_toolkit`导入使用）

如需获取更多信息，请参考相关文档。