Paste any post, comment or review to see how it is scored, word by word.
Each word, emoji and emoticon is looked up in a lexicon rated from −4 to +4. Scores are raised by boosters ("really", "so"), lowered by dampeners ("kinda"), flipped by negations within three words ("not", "isn't", "never"), lifted for ALL CAPS and stretched spelling ("sooo"), and weighted toward the clause after "but". The total becomes a compound score from −1 to +1: +0.05 or above is positive, −0.05 or below is negative. This follows the VADER approach used for social media text.
The lexicon is English. Hindi, Malayalam, Tamil and other scripts are detected and flagged, and their scores rely only on emoji and English hashtags, so review them or correct the label on the Mentions page.
These settings apply only to this alert. Choose the keywords to collect; excluded keywords drop any mention that contains them.
Sentimeter scores mentions you bring in. Export them from your listening tool, platform analytics or a scraper as CSV or JSON. Recognized columns: text (or content, message, comment, review, body), platform (or source), date, author, handle, location (or country), language and reach (or followers). Only text is required. Data stays in this browser.