Extrinsic Evaluation on Automatic Summarization Tasks: Testing Affixality Measurements for Statistical Word Stemming Carlos-Francisco Méndez-Cruz 1,4 , Juan-Manuel Torres-Moreno 1,2 , Alfonso Medina-Urrea 3 , and Gerardo Sierra 4 1 LIA-Université d’Avignon et des Pays de Vaucluse, France cmendezc@ii.unam.mx, juan-manuel.torres@univ-avignon.fr 2 École Polytechnique de Montréal, Canada 3 El Colegio de México A.C., México amedinau@colmex.mx 4 GIL-Instituto de Ingeniería UNAM, México gsierram@ii.unam.mx Abstract. This paper presents some experiments of evaluation of a sta- tistical stemming algorithm based on morphological segmentation. The method estimates affixality of word fragments. It combines three indexes associated to possible cuts. This unsupervised and language-independent method has been easily adapted to generate an effective morphological stemmer. This stemmer has been coupled with Cortex, an automatic summarization system, in order to generate summaries in English, Span- ish and French. Summaries have been evaluated using ROUGE. The results of this extrinsic evaluation show that our stemming algorithm outperforms several classical systems. Key words: Automatic summarization, Affixality Measurements, Morphologi- cal Segmentation, Statistical Stemming, CORTEX. 1 Introduction Discovering linguistic units in Natural Languages has been a long-standing hu- man task. Now, automatic approaches are used to conduct this work. Despite rule-based approaches for word processing are widely used, there is a renewed in- terest in morphological methods. Corpora of morphologically complex languages, such as agglutinative ones, can be processed in order to discover morphemes. Sim- ple strategies are not suitable for this kind of languages. Also, corpora for lan- guages, which have not been computationally studied, appear every day. Then, this paper presents a method for unsupervised learning of morphology. A morphologist tries to collect morphological units from languages, among other tasks. Roughly, two types of units are considered in this paper: stems and affixes. On the one hand, a stem is the part of a word stripped of affixes, it carries lexical content. On the other hand, affixes carry either grammatical information