01
You are working
present statement
Diagnostic
Ulebomba: the Flat Text Problem made audible.
First define the problem: flat text is writing without the sound layer. Then hear the
evidence. The Bemba word ulebomba is one written form, but in speech it
resolves into ten readings and ten meanings. This is the gap text-trained systems
cannot close by scaling text alone.
Audible diagnostic
One spelling. Ten meanings. Hear the missing layer.
In standard Bemba writing, ulebomba is one eight-letter form.
In speech, it can resolve into ten distinct readings. The differences live in tone,
vowel length, downstep, phrase boundary, and voice quality — play them below.
02
You are getting wet
present statement
03
You should work
modal statement
04
Are you working?
yes/no question
05
Who is working
relative reading
06
You are really working
emphatic reading
07
Is the one working?
question with focus
08
Is the one getting wet?
question with lexical shift
09
Are you getting wet?
direct question
10
You should be getting wet
modal progressive
Linguistic notes
What the page cannot show.
You are working
A plain declarative reading. The meaning is carried by the contour across the four syllables.
You are getting wet
The same eight letters, but a different lexical meaning because the sound pattern changes.
You should work
The written string stays flat while the spoken form carries the grammatical force.
Are you working?
The phrase-boundary edge makes this a question even though the letters do not change.
Who is working
A subtle shift in pitch and grouping changes the role of the utterance.
You are really working
Voice quality and emphasis sit in the audio layer, not in the orthography.
Is the one working?
A model trained only on text sees the same token sequence again.
Is the one getting wet?
Tone and vowel timing separate a second lexical path from the same spelling.
Are you getting wet?
The page does not mark the difference. The ear does.
You should be getting wet
The final reading completes the diagnostic: ten meanings, one written form.
Why this matters
No frontier LLM can tell these readings apart from the letters alone.
The distinguishing facts are tonal and phonetic. They live in the audio signal and attach to syllables. A model that sees only the written token has not been given the units required to decide meaning.