How to make construction grammars fluid and robust

Abstract

Natural languages are fluid. New conventions may arise and there is never absolute consensus in a population. How can human language users nevertheless have such a high rate of communicative success? And how do they deal with the incomplete sentences, false starts, errors and noise that is common in normal discourse? Fluidity, ungrammaticality and error are key problems for formal descriptions of language and for computational implementations of language processing because these seem to be necessarily rigid and mechanical. This chapter discusses how these issues are approached within the framework of Fluid Construction Grammar. Fluidity is not achieved by a single mechanism but through a combination of intelligent grammar design and flexible processing principles.

Introducing Fluid Construction Grammar

“…Such errors happen very frequently in language, particularly in speech, because some sounds may not have been accurately produced by the speaker, or the hearer's speech system was unable to recognize them. Some words or constructions may have been used inappropriately by the speaker because of sloppiness or due to errors and mistakes in planning (Steels & van Trijp, 2011). Problems may also arise because the language systems of different speakers of the same language are never entirely the same, and so the hearer may be confronted with an ungrammatical sentence which is perfectly grammatical within the ideolect of the speaker.…”

“…Diagnostics test for example whether all words in the input could be integrated into the sentence as a whole, or whether all the meaning that the speaker wanted to express are actually part of the final utterance. Repair strategies then try to fix this situation, possibly by ignoring some of the input Here are two examples which are illustrated elsewhere in more detail (Steels & van Trijp, 2011): 1. It is possible to introduce units for unknown words or words that could not be recognized by the speech system.…”

The Emergence of Morphosyntactic Case Systems

“…The only difference is that the listener may observe an unknown case marker (Repair 1) that can be coupled to the problem of cognitive effort, or that he observes a known case marker in a new distribution (Repair 2). Technical details on processing and coercion can be found in Steels and van Trijp (2011).…”

The evolution of case grammar

“…The implementation has been extended with solutions for feature indeterminacy and ambiguity (van Trijp 2011c), and long-distance dependencies ( van Trijp 2014b); and it has been grounded on humanoid robots . I have also been particularly concerned with fluid and robust language processing (Steels & van Trijp 2011), integrating diagnostics and repairs in the FCG-system (Beuls, van Trijp & Wellens 2012) and exploring reflective architectures for open-ended processing (van Trijp 2012a).…”