← Back to MouthMath

How MouthMath counts words

MouthMath turns public speeches, letters, testimony and court opinions into simple word counts. The idea is easy to explain: take a document, count every word in it, and list the words from most used to least used. This article walks through what happens along the way, so you know exactly what the numbers on the site mean.

Step 1: Find the original document

Every count starts with a public document from an official or well-known source. Federal Reserve speeches come from the Federal Reserve's own website or its historical archive. Presidential speeches come from The American Presidency Project at UC Santa Barbara. Floor remarks by members of Congress come from the Congressional Record, court opinions come from the Supreme Court, and CEO material comes from each company's investor pages. Each document on a speaker's page links back to its original source, so you can always read the full text yourself.

Step 2: Keep only the speaker's own words

Many documents include more than one voice. An earnings call has analysts asking questions, and a hearing has members of Congress speaking. Where a source makes it clear, MouthMath counts only the part spoken or written by the person you selected, such as a CEO's prepared remarks at the start of a call. Headings, page numbers and footnote markers are left out where possible.

Step 3: Count every word

Words are counted without regard to capital letters, so “Inflation” and “inflation” are the same word. Punctuation is set aside. Numbers on their own, like a year or a percentage figure, are left out so they don't crowd the list. Each word then gets a count: the number of times it appears in the documents you have chosen.

Step 4: Hide the common words (if you want)

The most used words in almost any English text are small ones like “the”, “and”, “of” and “to”. They matter for reading but tell you little about a speech's subject. MouthMath hides these common words by default, so the list starts with the words that carry meaning. You can turn that off with the “Hide common stopwords” switch to see everything.

Step 5: Rank and link

The remaining words are ranked from most to least used. Tap any word and MouthMath shows where it was used, with the sentence around it, so a count never stands on its own. You can look at one recent document or combine several, and you can switch between the ranked list, the Word Race, the Word Cloud and the Word Ring to see the same counts in different ways.

What the numbers can and can't tell you

A word count is a simple, honest measure: it shows what someone talked about and how often. It does not show tone, intent or whether a statement was right. A word can appear often because it is the main subject, or because it is part of a fixed phrase, like “United States”. Longer documents also produce bigger counts, so when comparing two speakers it helps to look at the share of all words rather than the raw count.

Not every speech is public, and transcripts can contain small errors, so counts are offered on a best-effort basis and updated as new material comes out. Every speaker is counted the same way, with the same rules, and nothing on MouthMath is financial, investment or legal advice.

More articles · All speakers