Update annotations for Ekaterina/paper_25.txt

annotations/Ekaterina/paper_25.txt.json

@@ -14,5 +14,13 @@
     "label": "Unsupported claim",
     "user": "Ekaterina",
     "text": "m-multiple MCFG"
+  },
+  {
+    "file": "paper_25.txt",
+    "start": 1114,
+    "end": 3621,
+    "label": "Lacks synthesis",
+    "user": "Ekaterina",
+    "text": "One of the more general classes of mildly context-sensitive systems are multiple context-free grammars (MCFGs), which essentially generalizes the notion of a context-free grammars to operations on tuples of strings. We defer the reader to Seki et al. (1991) for a full definition and discussion of the properties of MCFGs. Instead we provide a simplified, computationally-oriented description that is more in line with our purposes and implementation. An m-multiple MCFG can be thought of as a tuple ⟨A, N , d, C, R, S 0 ⟩, where:\n\n• A is the terminal alphabet • N is a set of non-terminals and d : N → N a function from non-terminals to natural numbers; each non-terminal N is encoding a tuple of strings of fixed arity d(N) and the maximal arity of N decides the grammar's multiplicity • C is a mapping that associates each nonterminal N to a (possibly empty) set of elements from the d(N)-ary cartesian product N) ; put simply, the set of constants C N prescribes all the possible ways of initializing the non-terminal N • R a set of rewriting rules; rules are functions N × • • • × N → N that provide recipes on how to combine a number of non-terminals into a single non-terminal by rearranging and concetenating their contents; we will write:\n\nto denote a rule that combines non-terminals A and B of arities m and n into a non-terminal C of arity k, where each of the left-hand side coordinates x 1 , . . . y n is used exactly once in the right-hand side coordinates z 1 , . . . z k • S 0 the start symbol, a distinguished element of N satisfying d(S 0 ) = 1\n\nThe choice of MCFGs as our formal backbone comes due to their many advantages. Being a subtle but powerful generalization of CFGs, MCFGs have a familiar presentation that makes them easy to reason about, while remaining computationally tractable (Ljunglöf, 2012;Kallmeyer, 2010). At the same time, they offer an appealing dissociation between abstract and surface syntax and lexical choice. A derivation inspected purely on the level of rule type signatures takes the form of an abstract syntax tree that is reminiscent of a traditional CFG parse. Normalizing an MCFG so as to disallow rules from freely inserting constant strings (i.e. wrapping all constants under a non-terminal) allows us to (i) trace back all substrings of the final yield to a single non-terminal and (ii) provide a clear computational interpretation that casts an MCFG as a linear type system, and its derivation as a functional program (De Groote and Pogodalla, 2003)."
   }
 ]