Systems - implementation platforms

Home page


At the Linguistics Department, several implementation platforms for feature structure based grammars have been developed or are still being developed or expanded on. A special area of application for feature structure based implementation platforms are naturally HPSG-type grammars and also grammars not originally designed for computational implementation. We would like to say a few introductory words about the systems offered here. 

The Attribute Logic Engine (ALE) is one of the oldest systems still being used for the implementation of HPSG grammars. Originally ALE was created in collaboration between Bob Carpenter and Gerald Penn, but then became the sole work of Gerald Penn and is still being continually upgraded. Gerald Penn's contributions to the SFB 340 in 1996 to 1999 closely connected the Linguistics Department to the history of ALE. 

The ConTroll-System was developed by our Linguistics Department and also dates back to the Tübingen HPSG-projects of the SFB 340. In its beginning, the ConTroll-System was based on the logical foundations of HPSG-grammars created by Paul King with his Speciate Re-entrant Logic (SRL). SRL was in turn relationally expanded in its expressiveness, by Thilo Götz with regard to practical application for larger grammars, and was then realized in cooperation with Detmar Meurers and other Tübingen computational linguists. One of the particular intentions of ConTroll was to portray the foundations of HPSG, as expounded by Pollard and Sag in their book from 1994. For this reason, ConTroll is a pure logic program, which does not contain a parser. We arrived at its current form at the end of 1997. Since then ConTroll has been continually upgraded to be compatible with the latest version of Linux and Solaris. 

The TRALE system is a combination of the experience from the development and application of ALE and ConTroll. The idea is to combine the advantage of efficiency which ALE offers with the original concept of ConTroll, to submit a depiction, as true to form as possible, of theoretical HPSG grammars into computational implementation. The continuing work on TRALE, which focuses on the requirements of current linguistic research of HPSG, is motivated by this depiction. Our objective is to provide task-specific solutions for typical, computationally expensive mathematical constructs of HPSG grammars. It should then become possible to implement even theoretically well grounded grammars efficiently, whose computational treatment goes beyond the capabilities of a pure, general constraint solving system such as ConTroll, without forcing linguists to renounce their specific assumptions about the structure of language.

Last modified: 12.11.2003