Add an example of a noncontracting (context-sensitive) grammar.
Chris Pressey
4 months ago
| 15 | 15 | ambiguous, then all possible parses (or generations) can be returned. |
| 16 | 16 | |
| 17 | 17 | It should be understood that `relwrite` trades off performance and small |
| 18 | memory footprint in favour of generality. There is, however, a feature to improve | |
| 19 | the performance in the case of very long derivations. Specifying a search strategies | |
| 20 | enables a **beam search** algorithm which aggressively focuses on derivations with a | |
| 21 | desirable character, e.g. a particular minimum length. | |
| 18 | memory footprint in favour of generality, so in general usage, it works | |
| 19 | best on small inputs. | |
| 20 | ||
| 21 | There are, however, features intended to improve performance in the case of very | |
| 22 | long derivations. Specifying a search strategy enables a **beam search** algorithm | |
| 23 | which aggressively focuses on derivations with a desired propery, e.g. a particular | |
| 24 | minimum length. This does sacrifice completeness however -- only a handful of all | |
| 25 | the possible results will be returned. | |
| 22 | 26 | |
| 23 | 27 | The grammar must be provided in the form of a JSON file. There are example |
| 24 | 28 | grammar files in the `eg/` directory of this repo. |
| 37 | 41 | ./bin/relwrite eg/recursive-grammar.json --parse --start "a penguin sees a penguin then sees a dog" |
| 38 | 42 | ``` |
| 39 | 43 | |
| 40 | Generate a really long string from a non-terminal in a grammar: | |
| 44 | Generate a really long string from a non-terminal in a grammar, without running out | |
| 45 | of memory and only taking a few hours of processor time: | |
| 41 | 46 | |
| 42 | (TK) | |
| 47 | ``` | |
| 48 | ./bin/relwrite eg/recursive-grammar.json --start "<Sentence>" --max-derivations=1 --strategy=expansion --expand-until=3000 | |
| 49 | ``` | |
| 50 | ||
| 51 | Parse a really long string from a non-terminal in a grammar, without running out | |
| 52 | of memory and only taking a few hours of processor time. This assumes the string | |
| 53 | to be parsed is in JSON format in the file `xyz.json`. | |
| 54 | ||
| 55 | ``` | |
| 56 | ./bin/relwrite eg/recursive-grammar.json --parse --start-set-file=xyz.json --max-derivations=1 --strategy=contraction | |
| 57 | ``` | |
| 58 | ||
| 59 | ### Detailed usage | |
| 60 | ||
| 61 | Run `relwrite --help` for a description of all the possible command-line options. Note that | |
| 62 | these are somewhat provisional and subject to change. | |
| 43 | 63 | |
| 44 | 64 | ### Notes |
| 45 | 65 | |
| 46 | `relwrite` uses the term "derivation" as a generic term meaning "parse or generation". | |
| 47 | ||
| 48 | `relwrite` also uses the term "utterance" to mean "any string of terminals and non-terminals". | |
| 66 | `relwrite` uses the term "derivation" as a generic term meaning "a parse or a generated utterance". | |
| 67 | It also uses the term "utterance" to mean "any string of terminals and non-terminals". | |
| 49 | 68 | |
| 50 | 69 | ### TODO |
| 51 | 70 | |
| 52 | 71 | * specify output filename |
| 53 | 72 | * try heuristic for contraction phase: highest proportion of terminals |
| 73 | * `--goal` to assert that a particular final utterance appears | |