Skip to content

Commit cf25636

Browse files
committed
feat: updated community notes blog
1 parent a3271e5 commit cf25636

File tree

1 file changed

+1
-39
lines changed

1 file changed

+1
-39
lines changed

src/blog/2025-04-30-community-notes.mdx

Lines changed: 1 addition & 39 deletions
Original file line numberDiff line numberDiff line change
@@ -7,42 +7,4 @@ date: 2025-04-30
77
tags: devlog
88
---
99

10-
import TopDomains from "../images/ifcn_blog_top_domains_plot.png"
11-
import TopIFCNDomains from "../images/ifcn_blog_top_ifcn_plot.png"
12-
import IFCNIndiaDomains from "../images/ifcn_blog_top_ifcn_pie_india_colored.png"
13-
14-
We wanted to understand the role, if any, played by fact check websites in community notes on X. The data for community notes and user rating can be downloaded officially from [here](https://x.com/i/communitynotes/download-data) and fields of the data have been documented [here](https://communitynotes.x.com/guide/en/under-the-hood/download-data). The data for notes is available from 28th Jan 2021 to 25th Apr 2025. The total number of notes in this timeline are approximately 1.85 million.
15-
16-
We started off by trying to find which websites or domains are linked in a note? We used regex to extract urls from text and then extracted domain names from those urls. The top 20 website domains that are referenced in community notes over time can be seen below.
17-
18-
<img src={TopDomains} alt="TopDomains" width="1200" height="800" />
19-
20-
We found that **81.23%** of all the community notes had URLs in them. Majority of those URLs were of X itself(x.com and twitter.com). Wikipedia came second with 7.2%. This was followed by Youtube, Google, BBC, Reuters, Instagram etc.
21-
22-
Next, we repeated the same analysis to find the number of community notes which have International Fact-Checking Network ([IFCN](https://www.poynter.org/ifcn/)) websites included in them. Since we couldn’t find any available database of IFCN domains, we used the [list from wikipedia](https://en.wikipedia.org/wiki/List_of_fact-checking_websites) and manually created an array with the domains.
23-
24-
<img src={TopIFCNDomains} alt="TopDomains" width="1200" height="800" />
25-
26-
We then ran a similar analysis as above only for Indian IFCN-afflicated organizations. We found that 10 Indian IFCN domains were present in the **1833** notes, which is approx **0.121%** of all community notes.
27-
28-
<img src={IFCNIndiaDomains} alt="TopDomains" width="1200" height="800" />
29-
30-
Here is a distribution with counts of India based IFCN websites.
31-
32-
| Domain | Count |
33-
| :---- | :---- |
34-
| indiatoday.in | 945 |
35-
| altnews.in | 270 |
36-
| factly.in | 255 |
37-
| newschecker.in | 154 |
38-
| factcrescendo.com | 103 |
39-
| youturn.in | 47 |
40-
| newsmobile.in | 24 |
41-
| vishvasnews.com | 18 |
42-
| thip.media | 13 |
43-
| medicaldialogues.in | 4 |
44-
45-
In the future we would like to conduct a [topic analysis](https://www.ibm.com/think/topics/topic-modeling) of all notes that include links to IFCN-affiliated websites. The goal is to understand the broader topics, discussions, or contexts in which fact-checking sources are cited. Additionally, we’d like to compare user approval ratings for notes that include IFCN links versus those that don’t.
46-
47-
The Google Colab Notebook with the code of the above analysis can be found here - [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1tq6wwBsuq-HFnsD_oBmgg2TAF8WxRt92?usp=sharing).
48-
10+
We received feedback that there were erros in the analysis. We will update the blog after reviewing the analysis.

0 commit comments

Comments
 (0)