This thesis addresses several aspects about the automatic translation from Castilian Spanish to Spanish Sign Language _x0028_LSE_x0029__x002c_ two typologically distant languages with not enough linguistics resources enabling statistical approaches to translation. For this reason_x002c_ a rule-based approach grounded on contrastive grammatical studies on both languages is used. An architecture following the analysis, transfer and generation model has been chosen. Transfer is performed at grammatical functions level, which are delivered by a Spanish dependency parser without incurring into the complexities of a more deeper analysis. The bilingual base lexicon is obtained from the Diccionario normativo de la lengua de signos espan~ola (DILSE-III), which contains the correspondences between Spanish lemmas and their SEA (Sistema de escritura alfabeĀ“tica) representation of signs. The lexicon is extended through two ways: taking advantage of the difference in flexibility between the part-of-speech systems of Spanish and LSE, and exploiting lexical semantic relations, not only synonymy and hyponymy, but also other relations as part-whole. During the structural transfer phase, some nodes of the dependency analysis are transformed, other are removed, and new nodes are inserted. Some classifier predicates are generated in this phase. Surface order generation of signs is obtained by means of the topological ordering of the graph of precedence relations between signs. Pairs of signs having head-dependent relations or sharing the same head are examined in order to determine if its relative ordering is marked or not. The system is evaluated at this point, and results are compared to those obtained with statistical models. Best results are obtained with the rule-based approach, with a 0.30 BLEU (Bilingual Evaluation Understudy) and a 42% TER (Translation Error Rate). A linguistic-oriented analysis of errors is provided. Finally, in the morphological generation phase, glosses with morphological annotations are replaced by the HamNoSys phonological representations produced by a computational morphology. These representations are used for animation synthesis with avatars. The computational morphology that has been implemented uses inflection, introflection and suppletion to model a significant fragment of the LSE morphology. Among the phenomena considered, it has been implemented deictics, nominal plural, aspect marking, verbal agreement, adjectival modification and degree.