Why is this called ambiguous?

Each digit holds three or four letters, and this encoder writes only the digit once per letter. So the digit 2 could be A, B, or C, and a sequence like 4663 could spell good, home, gone, hood, and others. Predictive T9 uses a dictionary to pick the most likely word.

How does it differ from multi-tap encoding?

Multi-tap repeats a digit by tap count, so C is 222 and is uniquely recoverable. Ambiguous T9 writes the digit only once per letter, so C and A both reduce to a single 2. That is what makes it shorter but ambiguous.

Which keypad layout is used?

The standard ITU-T E.161 layout: 2 ABC, 3 DEF, 4 GHI, 5 JKL, 6 MNO, 7 PQRS, 8 TUV, 9 WXYZ. Keys 7 and 9 hold four letters each, and 0 represents a space.

What happens to digits and punctuation?

Spaces become 0, the digits 0 to 9 pass through as themselves, and other characters such as punctuation are left unchanged so word boundaries and symbols are preserved in the output.

No. A phone keypad has no concept of case, so uppercase and lowercase letters map to the same digit. The output is purely numeric for alphabetic input.

T9 Ambiguous Code Encoder

The ambiguous T9 encoder converts text into the digit sequence you would press on a phone keypad without recording how many times you tapped. Because each digit covers three or four letters, the result is intentionally ambiguous — exactly the input a predictive T9 dictionary takes in and resolves to a likely word.

How it works

Every letter is replaced by the single digit of the key it sits on, using the ITU-T E.161 layout, and words are kept together:

2 ABC   3 DEF   4 GHI   5 JKL   6 MNO   7 PQRS   8 TUV   9 WXYZ   0 space
HOME -> 4 6 6 3 -> 4663
4663 could also be: GOOD, GONE, HOOD, HOME, INME …

Spaces become the 0 key, digits pass through, and other characters stay as-is.

Example and notes

The word HOME encodes to 4663, the same code as GOOD and GONE. That collision is the whole point: on old phones the software offered each candidate in turn and you cycled to the right one. To read an ambiguous code back you need a word list, whereas the multi-tap encoder produces a sequence that is uniquely decodable on its own.