This is the live development testing for Tom Guilleaume's sentence rephraser. The project's target untrained accuracy is 75%, which we hope to improve with user corrections guiding the machine learning process. Because the accuracy of the public cannot be trusted, the most "controversial" paramaters to distinguish cases within decision trees are not enabled until I approve them.

Current development stage is parsing the "movable parts" of the sentence - that is, grouping together unconditionally unseparated words. This is then compared to the Noun Phrases (NPs), Verb Phrases (VPs), Prepositional Phrases (PPs), etc. to determine behavior of that "movable part". Once that behavior is set, an array of sentences can be produced that are grammatically correct (assuming the extracted information was correct in the previous stages, i.e. part of speech mostly)


Enough chat, onto the live demo!

The latest (3rd) version, newest algorithm (Redesigned/started 4/1/13, WIP and may throw errors when I am actively working on it)





The latest (3rd) version, non-oop version with major memory issues (started 1/1/13)

The first algorithm (may not parce 100% of sentences depending on complexity)

Another archived version