A Computational Model of Natural Language Communication Interpretation, Inference, and Production in Database Semantics

\chapter*{Preface to the Second Edition}\markboth{Preface to the Second Edition}{Preface to the Second Edition}The 2nd Edition resembles the 1st in that the communication cycle ismodeled as a cognitive agent's turn-taking.  The first\footnote{We  begin with the hear mode because it provides the content needed for  the selective activation and inferencing in the think mode, and as  input for the speak mode.} step is the hear mode, which mapsunanalyzed modality-dependent external surfaces into cognitivecontent.  The second step is the think mode, which (i) selectivelyactivates content by navigating along the semantic relations ofstructure and (ii) derives new content by means of inferences forreasoning and as blue prints for action.  The third step is the speakmode, which maps the contents activated or derived in the think modeinto unanalyzed modality-dependent external surfaces.Regarding the details of the software solution, however, the 2ndEdition is substantially revised.  The hear mode relies on anautomatically managed \hbox{\it now front} in the agent'scontent-addressable database.  The now front is cleared in regularintervals: items (proplets) which have ceased to be potentialcandidates for concatenation with an analyzed next word are leftbehind in the sediment of memory, such that a current now front willcontain no more than four or five items.  This allows to use thealgorithm of Left-Associative Grammar without rule packages, relyingupon self-organization instead.The absence of rule packages removes the main difference between theoperations (formerly called rules) of the hear mode and inferences.The speak mode is also simplified by using language-dependentlexicalization rules which are embedded into the think modeoperations.  The selective activation of the think mode ischaracterized by a new kind of graph, first developed in CLaTR.  Thecoreference of anaphoric pronouns is now based on the use of addressesinstead of co-indexing, and may be found in chapters 10 and 11 ofCLaTR.  These changes contribute to a standardized presentation ofDatabase Semantics (DBS), which has been extended to the 2nd Editionof CLaTR and is complied with in HBTR.%Special thanks to Ronan Nugent, Senior Editor of Computer Science at%Springer, for his continued support.\noindent\\April 2016\hspace*{8.67cm}Roland Hausser\\