Interactive Predictive Parsing Framework for the Spanish Language
Resumen
El marco teórico de Parsing Predictivo Interactivo (IPP) permite construir sistemas de anotación sintáctica interactivos. Los anotadores humanos pueden utilizar estos sistemas de ayuda para crear árboles sintácticos con muy poco esfuerzo (en comparación con el trabajo requerido para corregir manualmente árboles obtenidos a partir de un analizador sintáctico completamente automático). En este artículo se presenta la adaptación a la lengua castellana del marco IPP y su herramienta de anotación IPP-Ann, usando modelos obtenidos a partir del UAM Spanish Treebank. Hemos llevado a cabo experimentación simulando al usuario para obtener métricas de evaluación objetivas para nuestro sistema. Estos resultados muestran que el marco IPP aplicado al UAM Spanish Treebank se traduce en una importante cantidad de esfuerzo ahorrado, comparable con el obtenido al aplicar el marco IPP para analizar la lengua inglesa mediante el Penn Treebank.
The Interactive Predictive Parsing (IPP) framework allows the construction of interactive tree annotation systems. These can help human annotators in creating error-free parse trees with little effort (compared to manually post-editing the trees obtained from a completely automatic parser). In this paper we adapt the IPP framework and the IPP-Ann annotation tool for parse of the Spanish language, by using models obtained from the UAM Spanish Treebank. We performed user simulation experimentation and obtained objective evaluation metrics. The results establish that the IPP framework over the UAM Treebank shows important amounts of user effort reduction, comparable to the gains obtained when applying IPP to the English language on the Penn Treebank.
The Interactive Predictive Parsing (IPP) framework allows the construction of interactive tree annotation systems. These can help human annotators in creating error-free parse trees with little effort (compared to manually post-editing the trees obtained from a completely automatic parser). In this paper we adapt the IPP framework and the IPP-Ann annotation tool for parse of the Spanish language, by using models obtained from the UAM Spanish Treebank. We performed user simulation experimentation and obtained objective evaluation metrics. The results establish that the IPP framework over the UAM Treebank shows important amounts of user effort reduction, comparable to the gains obtained when applying IPP to the English language on the Penn Treebank.