Claudio Pinhanez, Paulo Cavalin, Victor Henrique Alves Ribeiro, Ana Appel, Heloisa Candello, Julio Nogima, Mauro Pichiliani, Melina Guerra, Maira de Bayser, Gabriel Malfatti, Henrique Ferreira. Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 2021.

This paper describes how machine learning training data and symbolic knowledge from curators of conversational systems can be used together to improve the accuracy of those systems and to enable better curatorial tools. This is done in the context of a real-world practice of curators of conversational systems who often embed taxonomically-structured meta-knowledge into their documentation. The paper provides evidence that the practice is quite common among curators, that is used as part of their collaborative practices, and that the embedded knowledge can be mined by algorithms. Further, this meta-knowledge can be integrated, using neuro-symbolic algorithms, to the machine learning-based conversational system, to improve its run-time accuracy and to enable tools to support curatorial tasks. Those results point towards new ways of designing development tools which explore an integrated use of code and documentation by machines.

In a setting where two AI agents embodied as animated humanoid avatars are engaged in a conversation with one human and each other, we see two challenges. One, determination by the AI agents about which one of them is being addressed. Two, determination by the AI agents if they may/could/should speak at the end of a turn. In this work we bring these two challenges together and explore the participation of AI agents in multi-party conversations. Particularly, we show two embodied AI shopkeeper agents who sell similar items aiming to get the business of a user by competing with each other on the price. In this scenario, we solve the first challenge by using headpose (estimated by deep learning techniques) to determine who the user is talking to. For the second challenge we use deontic logic to model rules of a negotiation conversation.

The numerous lawsuits in progress or already judged by the Brazilian Supreme Court consists of a large amount of non-structured data. This leads to a large number of hidden or unknown information, since some relationships between lawsuits are not explicit in the available data; and contributes to generate non-intuitive influences between variables, which in addition increases the degree of uncertainty on judicial outcomes. This work proposes an approach to identify possible judgment outcomes that considers the use of similarity calculations and clustering mechanisms based on lawsuits patterns. The similarity problem was tackled by analysing metadata manually extracted from lawsuits; and this work also presents an approach to detect clusters and to compile past votes. From the results, it is possible to verify lawsuits most likely outcomes and to detect their degree of uncertainty.

This paper presents OpenWordNet-PT, a freely available open-source wordnet for Portuguese, with its latest developments and practical uses.We provide a detailed description of the RDF representation developed for OpenWordnet-PT.We highlight our efforts to extend the coverage of our resource and add nominalization relations connecting nouns and verbs.Finally, we present several real-world applications where OpenWordnet-PT was put to use, including a large-scale high-throughput sentiment analysis system.

In this paper we describe a scalable service for customized data validation and content assistance by means of domain-independent, user-provided sets of complex data constraints. We present an integrated architecture and a particular implementation that combines the use of existing grammar- and rule-based schema languages that allows providers to specify rules in a declarative manner. The integrated architecture provides a way to semi-automatically fill in form fields by calling the proposed service, which enumerates domains from the previously stored data validation constraints. Additionally, better error reporting can be achieved by leveraging structure from the rules' definitions. The proposed architecture has been shown to be practical and is in production use by a large organization, successfully fulfilling its role.

Multi-domain IT services are delivered by technicians with a variety of expert knowledge in different areas. Their skills and availability are an important property of the service. However, most organizations do not have a consistent view of this information because creation and maintenance of a skill model is a difficult task, especially in light of privacy regulations, changing service catalogs and worker turnover. We propose a method for ranking technicians on their expected performance according to their suitability for receiving the assignment of a service request without maintaining an explicit skill model describing which skills are possessed by each technician. We find appropriate assignees by making use of similarities between the assignees and previous tasks performed by them.