Contact

Email rlaejrtn16@gmail.com. There is no contact form, because a form would mean a server, and the whole privacy claim of this site rests on there not being one.

Reporting a wrong number

This is the message we most want to receive, and the most useful ones tell us three things: the page, the figure as printed, and what you believe it should be. Every table on this site is generated from a small number of functions and asserted in an automated test suite, so a genuine error usually means either a function is wrong — in which case every page carrying it is wrong and we want to fix it today — or a sentence has drifted from the table it describes. Both are worth an email.

When the disagreement is with a measured figure and not an arithmetic one, the response is to re-run the measurement, never to adjust the wording. The script that produced the speaking-rate dataset writes the data file directly, so re-measuring updates every page that quotes it at once.

Suggesting a talk for the dataset

The speaking-rate set is limited by one hard requirement: the source must publish a full transcript and exact timings for the same delivery. Most speeches fail it — a famous address usually has a text somewhere and a runtime somewhere else, from different sources, with no way to confirm they describe the same performance.

If you know of material that meets the bar — timed captions alongside a complete transcript, single speaker, English — please send it. Broadening the sample beyond one genre is the clearest way this site could get better, and it is the limitation stated most prominently on the speaking rate page.

Using the calculator elsewhere

The embed snippet on the front page works on any site, including course pages and internal wikis. It loads nothing but the calculator, sets no cookies and makes no tracking requests. You do not need permission and there is nothing to sign; if you are putting it in front of a class and want something changed to make it more useful, tell us what.

Questions about the method

These are welcome and usually make the site better. The three that come up most, answered here so an email is not needed:

"Why not use a bigger, more famous dataset?" Because the requirement is strict and most material fails it. A speaking rate needs a word count and a duration measured on the same delivery. A famous address usually has a transcript from one source and a runtime quoted by another, with no way to confirm both describe the same performance — different takes, different edits, applause included or not. Timed captions solve that by stamping the transcript against the recording itself.

"Why is your median higher than the number everyone else uses?" We do not know, because we could not trace where the commonly cited band came from. Our figure is a median of rehearsed conference talks, which are coached and time-constrained and may well be faster than speech in general. Both facts are on the speaking rate page, along with every individual talk, so you can judge the sample without relying on our summary.

"Can I have the raw data?" It is already published. Every talk, its word count, its measured window and its rate are in the table on that page, with each row linking to the source it came from. Anyone can re-derive a row by hand in a couple of minutes, which is the standard we were aiming for.

What we cannot help with

We cannot review your speech, tell you whether your delivery is good, or advise on what you should say. This site does arithmetic on a measured rate and publishes the measurement — it has no opinion about your material, and any page that claimed otherwise would be selling something.

We also cannot tell you what rate your instructor, producer or conference organiser expects. If you have been given a target that disagrees with everything here, use theirs: the calculator takes any rate you type, and matching the brief you were handed matters more than matching a median. The custom-rate box exists precisely for that.

A note on corrections we have already made

In the interest of showing the work: while building this site the automated checks caught two sentences that were plainly wrong. One said a script "needs a rate faster than every talk we measured" when two talks in fact exceeded it. Another described a fifteen-minute slot as forcing a deliberate speaker to cut, when the arithmetic showed every pace fitting comfortably. Both were rewritten to state exact counts, and both now have tests that fail the build if the sentence and the data ever disagree again. If you find a third, that is exactly the email to send.