Identifying Terminal Distinguishable Languages

Learning tree languages from text

“…Our results can be also employed for devising learning algorithms for linear languages, which will be a complimentary approach to the ones detailed in [55,62], as well as in [23]. In order to design string language learning algorithms based on tree learning, it seems to be best to prescribe, for each word length n, a skeleton tree S n which defines how to parse strings of length n. There seem to be interesting connections to the idea of employing permutation (families) for learning as detailed in [26].…”

“…With this interpretation, learning algorithms for regular tree languages give rise to learning algorithms for regular string languages. • If we restrict ourselves to Ter applied to derivation trees of even linear grammars (with trivial labels for interior nodes), we basically arrive at a variant of the terminal distinguishable even linear languages discussed in [23] in more detail. Here, focussing on derivation trees for even linear grammars basically entails a reordering of the letters, a technique addressed in particular in [26].…”