====== Speak with Elegance (or not) ======

<wrap lo>Sunday October 17, 2021</wrap>

When we’re speaking, we have an overwhelming compulsion to fill ‘dead air’ with words. So, we tend to inject lots of ‘ums’ and ‘uhs’.

{{ :caveman_speech-300x200.jpg |}}

I noticed that YouTube automatically generates a transcript from videos. So, I saved a copy of the transcript from the [[https://www.youtube.com/watch?v=f7Me86wdbfo|recent lunch-n-learn I did for Smart Data]], to see how I did.

I threw together a quick parser:

<code python>
#!/usr/bin/python3

def word_check(line_to_check, word_to_check):
	current_count = 0

	if line_to_check.startswith(f"{word_to_check} "):
		current_count = current_count + 1
	if line_to_check.endswith(f" {word_to_check}"):
		current_count = current_count + 1
	current_count = current_count + line_to_check.count(f" {word_to_check} ")

	return current_count

if __name__ == "__main__":
	um_count = 0
	uh_count = 0

	with open("video_transcript.txt") as f:
		contents = f.readlines()
	
		for line in contents:
			line = line.strip()

			um_count = um_count + word_check(line, "um")

			uh_count = uh_count + word_check(line, "uh")

	print(f"You said 'um' {um_count} times")
	print(f"You said 'uh' {uh_count} times")
</code>

Then, I ran it against the transcript of my 90 minute talk. Bear in mind, I was pretty confident that I do a decent job of minding my ‘uhs’, and ‘ums’.

My results?

<code>
You said 'um' 31 times
You said 'uh' 165 times
</code>

Apparently, perception is not always reality.  Who knew?

{{ :ian_malcom-300x192.jpg |}}