LkbUpdates

Overview

Since the publication of Copestake (2002), the general documentation of the LKB system (and its approach to using typed feature structures in grammar writing), the software has been continuously extended. Although we have in general attempted to preserve the documented functionality and interfaces, there is an array of additional facilities that users may find useful. The following is an overview of some of the major new components. All of them are optional, in the sense that basic grammar writing and processing can be achieved without using extra facilities; in fact, some of the new components target `power' users (building very large-scale grammars) and specific applications (e.g. machine translation). A more detailed summary of code changes to the LKB is available from the CVS source repository logs and the LkbEvolution page.

Lexical Database Integration

Versions of the LKB since about 2004 come with an optional interface to an external database storing the lexicon. See the LkbLexDb page for more information and instructions on how to set up the database and maintaining it.

Semantic Interface (SEM-I) Support

Interfacing the LKB to an OWL Ontology

XML-Based Interface to External (Morphological) Preprocessor

As part of the EU-funded [http://www.project-deepthought.net Deep-Thought] project, the LKB was interfaced to an external finite-state preprocessor, morphological analyzer, and tagger using an XML-based interface. This interface could serve to harmonize existing connections to external preprocessors (e.g. to [http://chasen.aist-nara.ac.jp ChaSen] in the Japanese grammar) over time, and probably should be used as the role model for future integrations. The Simple PreProcessor Protocol is documented on the LkbSppp page.

Finite-State Preprocessing

MRS Rewrite System (Transfer Component)

Feature Structure Unfilling

Unfilling of feature structures is the removal of redundant information from feature structures, in order to make them smaller. When only redundant information is removed, they contain precisely the same information as the original feature structures, but with less room.

One application for unfilling is the display of feature structures, such as the ones associated with chart edges. When they are displayed in full, they can be huge, but a lot of the information is redundant. Unfilled, the feature structure becomes much clearer.

The menus that display feature structures have been extended with menu entries to display unfilled feature structures (in the "Parse Tree" and "Chart" windows). The best way to get familiar with unfilled feature structures is to compare a few with the unfilled variants.

Unfilling more technically

Unfilling is in some sense the opposite operation of "Expand type". The latter takes a type and recursively adds all constraints that have been added to the types. Unfilling takes a feature structure and recursively removes all information that "Expand type" would have added. Unfilling an expanded type, for instance, returns the original type (note though that it is not always that simple).

Linguistic User Interface (LUI)

The Linguistic User Interface (LUI) is an on-going project to build a visualization tool for the most common object types in constraint-based grammars, i.e. trees, feature structures, MRSs, charts, et al. While the LKB comes with built-in browsers for all of these, the current use of the CLIM (Common-Lisp Interface Manager) toolkit in the LKB severely limits portability, ease of use, and programmatic extension. The LkbLui pages describe the current state of play for integration of LUI with the LKB.

Linguistic Server Protocol (LSP)

Robust Minimal Recursion Semantics