eSPERTo’s Paraphrastic Knowledge applied to Question-Answering and Summarization Cristina Mota, Anabela Barreiro, Francisco Raposo, Ricardo Ribeiro, Sérgio Curto, and Luísa Coheur L2F/INESC-ID, Rua Alves Redol 9, 1000-029 Lisboa, Portugal {cmota,francisco.afonso.raposo}@ist.utl.pt {abarreiro,sergio.curto,ricardo.ribeiro,luisa.coheur}@inesc-id.pt http://www.l2f.inesc-id.pt Abstract. This paper reports our first attempt of integrating eSPERTo’s para- phrastic engine, which is based on NooJ platform, with two application scenar- ios: a conversational agent, and a summarization system. We briefly describe eS- PERTo’s base resources, and the necessary modifications to these resources that enabled the production of paraphrases required to feed both systems. Although the improvement observed in both scenarios is not significant, we present a de- tailed error analysis to further improve the achieved results in future experiments. Keywords: paraphrasing, question-answering, summarization, Portuguese, NooJ 1 Introduction eSPERTo is a paraphrasing system that comprises a paraphrase generator, a paraphrase acquisition module, and a web interactive application to help Portuguese language learners, translators and editors in revising their texts. This system was developed in the scope of the eSPERTo project whose aim is to build an hybrid paraphrasing system. Its core linguistic resources, extracted from OpenLogos bilingual resources [3], the free open source version of the Logos System [15], were adapted and integrated into NooJ linguistic engine [16]. In this paper, we present a study on the integration of the paraphrase generator into two application scenarios: a conversational agent, and a summarization system. In the first application scenario, eSPERTo’s paraphrases were explored to enrich the Portuguese knowledge base of a question-answering intelligent virtual conversational agent, EDGAR [5]. In the second application scenario, eSPERTo was used in the pre- processing phase to assist automatic text summarization [12]. The benefits of para- phrastic knowledge to these natural language processing tasks have been defined and quantified by several authors [14, 4]. On the one hand, discovering paraphrased answers provides additional evidence that an answer is correct and helps systems to identify se- mantically related questions for the same answer. On the other hand, the identification of paraphrases allows information across documents to be condensed, redundant infor- mation to be identified and eliminated and helps improve the quality of the generated summaries.