| # String Transformations |
|
|
| Character-level string transformation tasks encoded as ILP problems. Each task requires learning a program that maps an input string to an output string, operating character-by-character. |
|
|
| **Tasks:** 329 tasks (in `<task>/train/`) |
|
|
| **Head relation:** `out(Ex, Pos, Char)` — position `Pos` in the output string of example `Ex` is character `Char`. |
|
|
| **Background relations:** |
|
|
| | Relation | Arity | Description | |
| |----------|-------|-------------| |
| | `in(Ex, Pos, Char)` | 3 | Character `Char` at position `Pos` in the input string of example `Ex` | |
| | `add(P1, P2, P3)` | 3 | `P3 = P1 + P2` (position arithmetic) | |
| | `lt(P1, P2)` | 2 | Position `P1` is less than position `P2` | |
| | `my_succ(P1, P2)` | 2 | `P2` is the immediate successor of `P1` (i.e. `P2 = P1 + 1`) | |
| | `width(Ex, W)` | 2 | The input string of example `Ex` has width `W` | |
| | `change_case(C1, C2)` | 2 | `C2` is the case-swapped version of `C1` | |
| | `is_letter(C)` | 1 | `C` is a letter (a–z or A–Z) | |
| | `is_lowercase(C)` | 1 | `C` is a lowercase letter (a–z) | |
| | `is_uppercase(C)` | 1 | `C` is an uppercase letter (A–Z) | |
| | `is_number(C)` | 1 | `C` is a digit (0–9) | |
| | `is_space(C)` | 1 | `C` is a space character | |
| | `is_comma(C)` | 1 | `C` is a comma | |
| | `is_fullstop(C)` | 1 | `C` is a full stop (period) | |
|
|
| Not all relations appear in every task; the bias file for each task lists the relations available to the learner. |
|
|
| **Source:** Andrew Cropper. "Playgol: Learning Programs Through Play." *Proceedings of the 28th International Joint Conference on Artificial Intelligence (IJCAI)*, 2019. |
|
|