The notion of concrete datatypes (conctypes) was worked out in earlier papers. Briefly, such a conctype looks like a context-free grammar where the nonterminals correspond to types. In the first part of the paper, the author considers this idea. Easy arithmetical examples show that systematic use of unambiguous conctypes would lead to complicated syntax and, therefore, to complicated computations.

How can we take notice of ambiguity? The second part of the paper deals with the use of precedences to solve recursive ambiguity, an extension of the ambiguity in parsing arithmetical expressions. More precisely, the author suggests a technical definition of precedence grammars, for which some parse trees will be legal and others illegal. The correct parse trees can be characterized through an associate attribute grammar, which is formally described. Finally, the author shows how to incorporate precedence rules in the general context-free parsing algorithm of Earley.

The paper is part of the proceedings of a conference, so the author had no room to give full particulars, and it is difficult to assess the effectiveness of the approach. I hope for a more detailed account.