ArxivAutoJob

ArxivAutoJob

此仓库使用 arxiv_mcp_project，在 GitHub Archive 中通过每日任务收集 arxiv 论文。

更新

2025-08-19： 如果把事实核查交给 SEO，那我们就永远无法实现真正的事实核查。 所以……把事实核查交给受众自己。

对于任何文章，我们只需

• 将主语、谓语和宾语与定语、状语、补语等拆分开来。

• 对于文章中的事实部分，我们尝试根据主语、谓语和宾语重新构建事实。

• 为此，我们可以提问，以验证数字或示例是否属实。

• 我们根据定语、状语、补语等重新构建作者的观点。

• 为此，我们可以筛除观点：我为什么需要你的观点？并看看自己处于什么样的环境中？

• 我们还需要留意销售／营销文章，所以……

• 为了让受众进行自我评估，我们应该在提示词中加入示例。

但问题是，哪些内容应该从第一段传到第二段，作为动态提示词部分？目前我们只采用填充中间部分的方法。

也许可以做：

• 文章分析。
• 事实核查待办清单。
• 不同段落的数据可视化。

对于每周或每月分析：

• 事实核查行为。
• 用于数据可视化的 3D 词云。
• 打破信息茧房的建议。
• 元数据，或用户自己的术语表。

2025-08-04： 我发现了一种分析内容的新思路。

You are a professional linguist and you are currently working on an analysis of an article that will be used to teach your classmates in class. The topic is to show students how to identify communication methods and their harms in communication through practical analysis.

Please note that the author and source of this article are not reliable and not necessarily credible, so the author may be deliberately deceiving people, and a random article as a teaching case can be very helpful for students to browse content on the Internet on a daily basis.

Analysis steps:
As a first step, please score according to the following formula:
1.1 Core Component (CC) = Subject (S) Predicate (P) Object (O)
1.2 Modifier (MC) = Attributive (Attr) Adverb (Adv) Complement (Comp) Other Modifier (OM)
The second step is to sort out the emotional intensity of the modification component, since the general modification part is the author's opinion or comment, please organize the strength of this part (1~10) for scoring.
The third step is to list the possible communication methods in the original text one by one, such as
- Information screening and one-sided presentation, including but not limited to: biased generalization, selective reporting, and survivor bias
- Concept and issue manipulation, including but not limited to: stealing concepts, labeling, stigmatization, emotional kidnapping
- Factual manipulation, including but not limited to: fabricating facts, taking them out of context, and misleading data
- Communication and psychological manipulation, including but not limited to: repeated reinforcement, creating antagonism, and silence spirals
- Media environment manipulation, including but not limited to: topic setting, information overload, and authoritative endorsement
In the fourth step, based on the results of the first and third steps, construct search terms for readers to practice fact-checking.

Next, I will send you the article piece by piece.

json 结构

{
  "core_components": {
    "subject": ["主语1", "主语2"],
    "predicate": ["谓语1", "谓语2"],
    "object": ["宾语1", "宾语2"]
  },
  "modifier_components": {
    "attributes": ["定语1", "定语2"],
    "adverbials": ["状语1", "状语2"],
    "complements": ["补语1", "补语2"],
    "other_modifiers": ["其他修饰语1", "其他修饰语2"],
    "emotional_intensity": 7
  },
  "communication_manipulation_techniques": {
    "information_filtering": [
      {
        "evaluation": "评价内容1",
        "fact_check_keywords": ["关键词1", "关键词2"]
      },
      {
        "evaluation": "评价内容2",
        "fact_check_keywords": ["关键词3", "关键词4"]
      }
    ],
    "concept_manipulation": [
      {
        "evaluation": "评价内容1",
        "fact_check_keywords": ["关键词1", "关键词2"]
      }
    ],
    "fact_manipulation": [
      {
        "evaluation": "评价内容1",
        "fact_check_keywords": ["关键词1", "关键词2"]
      },
      {
        "evaluation": "评价内容2",
        "fact_check_keywords": ["关键词3", "关键词4"]
      }
    ],
    "psychological_manipulation": [
      {
        "evaluation": "评价内容1",
        "fact_check_keywords": ["关键词1", "关键词2"]
      }
    ],
    "media_environment_manipulation": [
      {
        "evaluation": "评价内容1",
        "fact_check_keywords": ["关键词1", "关键词2"]
      },
      {
        "evaluation": "评价内容2",
        "fact_check_keywords": ["关键词3", "关键词4"]
      }
    ]
  }
}

2025-04-07： 我不认为自己应该用／依赖 AI 来阅读所有这些内容。 但我想用 AI 为我提供快速摘要。 将其留在 综合分析 部分。