Clasificaci´ on conjunta de frases clave y sus relaciones en documentos electr´ onicos de salud en espa˜ nol Joint classification of Key-Phrases and Relations in Electronic Health Documents Salvador Medina 1 , Jordi Turmo 1 1 TALP Research Center - Universitat Polit` ecnica de Catalunya smedina@cs.upc.edu y turmo@cs.upc.edu Abstract: This paper describes the approach presented by the TALP team for Task 3 of TASS-2018 : a convolutional neural network to jointly deal with classification of key-phrases and relationships in eHealth documents written in Spanish. The results obtained are promising as we ranked in first place in scenarios 2 and 3. Keywords: Relation extraction, joint classification, key-phrase classification, con- volutional nerural networks. Resumen: Este art´ ıculo describe el m´ etodo presentado por el equipo TALP en la Tarea 3 de TASS-2018 : una red neuronal convolucional para tratar conjuntamente la clasificaci´ on de frases clave y sus relaciones en documentos de salud escritos en espa˜ nol. La propuesta qued´ o en primera posici´ on en los escenarios 2 y 3. Palabras clave: Extracci´ on de relaciones, clasificaci´ on conjunta, clasificaci´ on de frases clave, redes neuronales convolucionales. 1 Introduction This article describes the model presented by the TALP Team for solving B and C sub- tasks of Task 3 in the Taller de An´alisis Sem´antico en la SEPLN 2018 (TASS-2018) (Mart´ ınez-C´amara et al., 2018). TASS-2018’s Task 3 consists in recogniting and classifying key-phrases as well as identifying the rela- tionships between them in Electronic Health Documents (i.e., eHealth documents) written in Spanish. Task 3 is divided in sub-tasks A, B and C, which correspond to key-phrase boundary recognition, key-phrase classifica- tion and relation detection, respectively. In this task, a key-phrase stands for any sub-phrase included in eHealth documents that is relevant from the clinical viewpoint and can be classified into Concept or Action. The relationships between them are classified into 6 types: 4 of them are between Concepts (is-a, part-of, property-of and same-as ) whi- le the rest are between an Action and anot- her key-phrase (subject and target ). The pro- posed task is similar to previous competi- tions such as Semeval-2017 Task 10: Scien- ceIE (Gonzalez-Hernandez et al., 2017), but uses a simpler categorization for key-phrases while considering a broader range of possible relationships. Participants in the Semeval-2017 Task 10: ScienceIE (Gonzalez-Hernandez et al., 2017) shared task considered a large plethora of su- pervised learning models, ranging from Con- volutional or Recurrent Neural Networks to Support Vector Machines, Conditional Ran- dom Fields and even rule-based systems, of- ten applying radically different models for each one of the three sub-tasks. Note that some of the teams did not participate in all three sub-tasks, this was in fact the case for the winners of sub-tasks BC (MayoNLP (Liu et al., 2017)) and C (MIT (Lee, Dernoncourt, and Szolovits, 2017)). 1.1 Joint classification of key-phrases and relationships In our implementation we tackle both the classification of key-phrases and the identi- fication of the relationships between them, corresponding to scenarios 2 and 3 of TASS 2018’s Task 3, as a single task. The intuition behind this decision is that the categories of key-phrases are influenced by the relations- hips they hold with other key-phrases. For instance, a verb is an Action key-phrase if and only if it relates to another Action or Concept by either being the subject or tar- get, which means that sometimes phrases are TASS 2018: Workshop on Semantic Analysis at SEPLN, septiembre 2018, págs. 83-88 ISSN 1613-0073 Copyright © 2018 by the paper's authors. Copying permitted for private and academic purposes.