Resolving Ambiguity in NLMaps Parses

Generating database queries from natural language is complex. By treating it as
a Sequence-to-Sequence task, one can nevertheless get good performance. In the
task of generating NLMaps Database Queries from natural language input, treating
it as a semantic parsing task to generate the database queries already reaches
accuracies over 90%.Most of the remaining mistakes are created due to ambiguity,
because information is missing and thus multiple possible answers fit the same
input query. To resolve this, this work tries to establish a human machine
dialogue with the user to resolve confusion.