A Linguistically Motivated Computational Framework for Irish Sign Language

Abstract

This research is concerned with the development of a linguistically motivated computational framework for Irish Sign Language (ISL) and defining a lexicon architecture that is sufficiently universal and robust to accommodate the linguistic phenomena pertinent to sign languages, in particular to this research ISL, in linguistic terms. To date, there is no definition for the architecture of the ISL lexicon in computational terms. ISL is a visual gestural languages articulated in 3D space with no written or aural form. ISL is a linguistically complete, very rich and complex language. Communication across sign languages occurs using visual-gestural modality, encompassing Manual Features (MFs) and Non-Manual Features (NMFs). MFs include hand shapes, hand locations, hand movements and orientation of the palm of the hands. NMFs include the use of eye gaze, facial expression, mouthing, head and upper body movements. The visual gestural realisation of a word in SL involves the simultaneous and parallel expression of a varied number of MFs and NMFs, each with their own duration, orientation and relative configuration and movement. The four research questions that this dissertation addresses are: RQ1. To what extent can RRG account successfully for ISL sentence structure? RQ2. How do we motivate the phonological-morphological interface in ISL? RQ3. How might lexical entries look for ISL classifiers, ISL verbs and ISL nouns within the RRG lexicon? RQ4. What is the appropriate linguistically motivated computational architecture for ISL and where within this architecture should the ISL grammatical morpheme repository reside? We will seek to implement a humanoid avatar, constrained in 3D space similar to human muscular skeletal constraints and we will utilise this avatar to motivate the design of our lexicon architecture for ISL. We endeavour to utilise Role and Reference Grammar (RRG), a structural functionalist theory of grammar and a functional model of language in the development of our linguistically motivated computational framework. We will utilise the Signs of Ireland (SOI) corpus in our analysis of ISL. In our pursuit of defining a lexicon architecture that is sufficiently universal and robust in nature to accommodate ISL we argue that the theory of qualia structures defined within the theory of the Generative Lexicon (GL) must to be extended to cater for SLs and their associated linguistic phenomena. We have argued that semantic properties, which contribute to the meaning of a sentence, will need to be extended to accommodate ISL. We motivate a new level of lexical meaning termed Articulatory Structure Level, such that the computational phonological parameters associated with this visual gestural language are sufficiently accommodated. This level of lexical meaning will represent the essential (computational) phonological parameters of an object as defined by the lexical item. These parameters will be used to account for various linguistic phenomena pertaining to ISL MFs and NMFs, which are necessary to adequately represent ISL within our computational framework. We refer to our newly developed framework as the Sign_A framework, with the ?A? within this term representing Articulatory Structure Level. We leverage our proposed Articulatory Structure Level for lexical meaning to accommodate the linguistic phenomena consistent with ISL and to develop a lexicon architecture capable of accommodating ISL in computational linguistic terms.