Twitterment

Sentiment checker for X

Read the mood of a conversation on X

Paste the replies under a post, the results of a search, or comments you collected about your brand. Twitterment scores each line as positive, neutral or negative, shows the split, and tells you which words drove it.

Paste posts, one per line

0 lines

Runs in your browser. Nothing you paste leaves this page. Up to 2,000 lines per run.

How the scoring works

A hand-built word list

About 360 English words and short phrases carry a weight from −3 to +3. "Awful" is −3, "slow" is −1, "solid" is +2, "masterpiece" is +3. Phrases like "waste of money", "no complaints" and "doesn't work" are matched before single words.

Negation and intensity

A negator such as "not", "never" or "isn't" within three words flips the next scored word and softens it, so "not good" leans negative while "not terrible" is only mildly positive. "Very", "really" and "extremely" push a word harder; "slightly" and "kind of" pull it back.

Emoji, caps and "but"

Around 50 common emoji have their own weights (🔥 and 🙌 are positive, 😡 and 👎 negative). A single word in capitals counts a bit more, exclamation marks add a little, and in "slow shipping but great product" the part after "but" gets more weight.

The final number

Weights are added up and squashed to a score between −1 and +1. Lines at +0.15 or above are positive, −0.15 or below negative, everything else neutral. Questions are halved, since "does it work?" asks rather than judges.

What a word list gets wrong

Lexicon scoring is fast, transparent and easy to check, which is why it is useful for a first pass. It is also blind in ways you should know before you quote a number.

  • Sarcasm reads literally. "Oh great, another outage" contains "great". The checker flags lines that look sarcastic (phrases like "yeah right", "just what I needed", or a 🙄) but does not change their score.
  • Context is invisible. A reply that says "same" or "this is why I left" depends on the post above it, and the checker only sees the line.
  • Slang moves faster than any list. New slang, in-jokes, and words that flip meaning in a community ("sick", "wicked", "mid") will be missed or misread.
  • It is English only. Other languages score as neutral.
  • It cannot tell who is talking. Ten replies from one angry account or a bot network count the same as ten different people.

Use the split to decide where to look, then read the lines. The guide on reading sentiment without fooling yourself covers sample size, bots and pile-ons.

Guides

  1. 01What sentiment analysis is, and how word lists differ from machine learningHow a computer guesses whether a post is happy or annoyed, and why the method you pick changes what you can trust.
  2. 02How to collect posts from X by hand for sentiment analysisSearch operators, copying tricks and a simple routine for building a sample of posts without scraping.
  3. 03Reading sentiment on X without fooling yourselfSample size, bots, brigading and the other ways a sentiment number can say something that isn't true.
  4. 04Social listening on X for small brands, without a big budgetA 30-minute weekly routine for tracking what people say about you on X, using search, a spreadsheet and the checker.
  5. 05How to measure reaction to a launch on XSet a baseline, collect in time windows, separate replies from quote posts, and report reasons as well as the split.
  6. 06A short history of searching Twitter, from Summize to XTwitter's search began as a sentiment startup called Summize. Here is how it grew from there.

Questions

Does Twitterment connect to my X account or collect posts for me?
No. Twitterment never logs in to X and never fetches anything from it. You copy the posts you want to check, paste them into the box, and the scoring runs in your own browser. Nothing you paste is sent to a server.
How accurate is the score?
It is a rough first read. A word list catches plain praise and complaints well, and it handles simple negation like "not good". It misses sarcasm, slang it doesn't know, and replies whose meaning depends on the post above them. Treat the counts as a starting point and read a sample of the posts yourself.
How many posts can I paste at once?
The checker reads up to 2,000 lines per run. Anything past that is ignored so the page stays fast on phones. For a typical reply thread, a few hundred lines is plenty.
Can I use the results in a report?
Yes. The Export CSV button gives you every line with its score, label and the words that moved the score, so you can sort it in a spreadsheet and check the edge cases by hand before you quote any numbers.
Is this an official X tool?
No. Twitterment is independent and is not affiliated with X Corp. It doesn't use the X API and works with any text you paste, whether it came from X or somewhere else.

Twitterment is independent and not affiliated with X Corp. It does not use the X API and never connects to your account.