ML model to detect malicious use of TakeTwo crowdsourced labeling

### Background on the problem the feature will solve/improved user experience
As an open-source, data crowdsourced solution, there is potential for malicious use/contributions

### Describe the solution you'd like
Develop an ML model(s) that detects malicious use such as:
- seeking to spam
- alter what is considered by racist, by making offensive racist terms seem less racist or identifying non-racist, benign terms as racist with the intent of making the api useless (by classifying everything as racist)
-
### Tasks
Description of the development tasks needed to complete this issue, including tests,

### Acceptance Criteria
Standards we believe this issue must reach to be considered complete and ready for a pull request. E.g precisely all the user should be able to do with this update, performance requirements, security requirements, etc as appropriate.

未指定任何文件、入口点、测试或验收标准。先调查代码库中的 Jupyter notebooks 和现有 data-science workflow，然后在实现之前定义恶意使用信号、评估数据以及可衡量的完成标准。

由索引模型根据 Issue 内容生成。