In the COST Action Multi3Generation, we advocate for a perspective on NLG that emphasizes thorough linguistic analysis and efficient computational modeling of languages. This approach highlights the significance of creating knowledge-based linguistic resources by researchers with expertise in computational linguistics, enabling them to challenge the inherent limitations of language models effectively. We believe that investing in high-quality resources and well-designed controllable systems is more advantageous than relying on post-editing of uncontrolled output, despite our curiosity about the capabilities of uncontrollable machines compared to humans. To this end, we have initiated initiatives to utilize resources from OpenLogos, focusing on two key features of the Logos Model: the Lexical and Semantico-Syntactic Abstraction Language (SAL) and the Semantico-Syntactic Tables (SemTab). These resources are utilized to develop and enhance new linguistic tools and resources, including paraphrasing tools and paraphrasaries. One notable initiative involved the design, development, and refinement of the eSPERTo paraphrase generation system, along with the establishment of the Parafrasário for the Portuguese language. We aim to extend similar efforts to other languages and NLP systems such as Summarization and Question-Answering.



Skip to content