TL;DR: Prefer space-separated input. For concatenated strings, parse left-to-right choosing valid 1–26 chunks; use dictionary checks to disambiguate splits.
The concatenation problem
81201215 could be 8-12-0-12-15 (invalid), 8-1-20-12-15 (HATLL?), or other splits. Without spaces, decoding is ambiguous.
Greedy parsing algorithm
From the left, take the longest valid two-digit number if the first two digits form 10–26; otherwise take one digit. Apply recursively on remainder. Compare result to dictionary words.
Dictionary-assisted disambiguation
If greedy parse yields HAT + nonsense tail, try alternate split at first branch. Tools like letterstonumbers.com expect spaces—preprocess concatenated strings by inserting splits manually where word boundaries are known.
Fixed-width encodings
Some legacy systems pad each number to two digits (08 05 12 12 15 for HELLO). Parser reads every two characters. Confirm padding rules on assignment PDFs.
Puzzle designer responsibility
Publish space-separated or fixed-width numbers. Solvers blame themselves for “wrong cipher” when the format was ambiguous—fair puzzles specify format explicitly.
Manual split exercise
Take 918151215. Try splits, list candidate words, pick English phrase “I LOVE” (9 18 15 22 5) only if 22 maps to V—check each chunk validity.
Programming note for solvers
Competitive puzzle coders brute-force splits with DFS and word list pruning. Hand solvers use linguistic intuition—both valid.
Recovering from one wrong split
If decode produces XRQZ gibberish mid-message, backtrack to first implausible bigram and retry branch. One early wrong digit poisons the rest.
Teaching spacing discipline
Grade encoding submissions on spacing format, not just correct letters. Professionals format for decodability.
Prevention beats cure
Encode with spaces, photograph clue sheets clearly, and reject camera OCR that drops spaces between digits when sharing in chat apps.
Greedy parsing walkthrough
From left, take two-digit chunk if 10–26, else one digit. Recurse on remainder. Compare greedy result to dictionary words.
Fixed-width two-digit padding
Some systems encode 0805121215 for HELLO with zero padding. Assignment PDFs should state padding rules explicitly.
Human vs programmatic solvers
Competitive teams brute-force splits with code; hand solvers use linguistic intuition—both valid hunt skills.
Backtracking after gibberish
If decode yields XRQZ mid-message, backtrack to first branch point and retry alternate split.
Prevention beats cure for authors
Space-separate published clues. Photograph props clearly so OCR in chat apps does not eat spaces between digits.
Splitting the digit run 2091514 into TRAIN
TRAIN without spaces is the digit run 2091514. Read left to right and only accept chunks from 1 to 26. 20 is a legal letter, T. The next digit is 9, which is I. Then 15 is O, and 14 is N. That split is 20 9 15 14, TRAIN. A greedy reader who always takes two digits will try 20, then 91, and 91 is illegal, so they back up to 20 and 9. The backup is the whole algorithm. You do not need a program to do it on a six-token word, but you do need to write the failed chunk down so you do not retry it.
The same run can be misread as 2 0 9 15 14, which dies immediately because 0 is not a letter. It can also be paused as 20 9 1 5 14, which is TAIEN, a non-word. Context picks TRAIN if the clue is about transport. Context picks nothing if the clue is a random string. That is why an author who removes spaces owes the solver a dictionary hint or a fixed width. “Two digits every time, with a leading zero on A through I” would have made TRAIN into 20 09 15 14 and removed the argument.
A second run, 3121116, that should not be guessed
3121116 is a deliberately mean string. Two-digit greed takes 31, which is illegal, then backs up to 3, which is C. Next, 12 is L, 11 is K, and 16 is P, so one path is CLK P, or CLKP. Another path takes 3, then 1, then 21, then 11, then 6: C A U K F. Neither is a word you should force. If a puzzle used this run, the fair recovery is to ask the author, not to invent a word that uses most of the digits.
Put TRAIN and this second run on the same worksheet as a pair labeled “has a word” and “does not.” Students learn to stop. The skill is not heroic parsing. The skill is noticing when every legal split is gibberish and when one split is obviously a word. Authors who want the first outcome must keep spaces or pad to two digits. Authors who publish 3121116 as a clue have made a broken clue.
Padded TRAIN as 20 09 15 14 for the next worksheet
Reprint the TRAIN clue with two digits per letter: 20 09 15 14. The leading zero on I removes the split argument. Students can still decode it, and they can see why the unspaced version was a trap rather than a bonus. Keep the unspaced 2091514 on a teacher-only copy so you can show the contrast for two minutes, then collect it. Leaving both versions on the student desk recreates the ambiguity you just explained.
For the non-word run 3121116, the padded form is the honest fix if you still want to use those letters. Decide the letters first, then pad. Do not pad a broken string and hope a word appears. If you cannot name the word before you write the digits, the item is not ready to print.
