The Adjective Index
How each outlet wrote from 20 September to 27 September, calmest first. Measured over the articles we read in full from each outlet (at least five, each over 150 words).
Calmest this week: SBS News, with a median Niral Score of 9.8 and 18.1 describing adjectives per 1,000 words.
| # | Outlet | Articles | Niral Score (median) | Change | Adjectives per 1,000 words | Sentiment intensity | Ours, and what came out |
|---|---|---|---|---|---|---|---|
| 1 | SBS News | 78 | 9.8 | -1.3 | 18.1 | 4.5 | 7.8 −2 |
| 2 | ABC News | 615 | 10.1 | -0.3 | 16.2 | 3.5 | 8.2 −1.9 |
| 3 | The Register | 127 | 10.4 | -0.4 | 20.1 | 2.3 | 8.3 −2.1 |
| 4 | 7NEWS | 195 | 11.1 | +0.3 | 15.7 | 5.7 | 7.6 −3.5 |
| 5 | TechCrunch | 125 | 11.3 | +0.1 | 16.6 | 1.5 | 9.6 −1.7 |
| 6 | BBC News | 56 | 11.3 | +0.7 | 21.3 | 3.9 | 8.5 −2.9 |
| 7 | The Guardian Australia | 434 | 12.1 | 0 | 19.6 | 4.2 | 9.6 −2.5 |
| 8 | Ars Technica | 73 | 13.3 | +0.8 | 23.4 | 2.5 | 10.2 −3.1 |
| 9 | NASA | 44 | 15 | +0.5 | 29.3 | 1.8 | 12.7 −2.3 |
| 10 | Engadget | 182 | 15.2 | +0.6 | 28.8 | 1.6 | 13 −2.1 |
| 11 | The Motley Fool Australia | 247 | 16.7 | +0.3 | 31.5 | 3.3 | 13.7 −3 |
How to read it
- Niral Score: 0 is a plain report, 100 is loaded. The median article, so one feature does not swing it.
- Adjectives per 1,000 words: describing adjectives in front of a noun, outside quotations, not carrying a fact.
- Change: against the week before. Green is calmer.
- Ours, and what came out: the median Niral Score of our plain-language version of this outlet's articles, and how many points that is below theirs. Both figures are over the same articles — the ones whose licence lets us publish a full rewrite, which is a subset of the count on the left — and the cell is blank under ten of them, because a median over fewer is one article's opinion.
Why Mundane Read is not a row in this table
It would come first every week, and that is the problem. Every row above is an outlet's own reporting. Ours would be their reporting with the loaded words taken out, so it wins by construction — scoring best on the number you edit text against is a tautology, not a finding, and a league table whose author is at the top is a table nobody should believe.
The last column is the honest version of the same question. It is the same article measured twice, so there is nothing to rig: it says what we took out of each outlet, and it is different for each of them.
This measures how articles were written, not whether they were right, and which stories an outlet chose shapes it: a week of sport reads differently from a week of politics. Method · Accuracy · Data (JSON, CC BY 4.0)