Four students from the Computer Engineering course at Insper developed the operational structure of a chatbot for translating the Guarani Mbya language. The work, presented in June at the Final Engineering Project (PFE), is part of the tools created in an initiative to preserve Brazilian indigenous languages, a partnership between IBM Research, the Artificial Intelligence Center of the University of São Paulo and the communities of indigenous peoples, with funding from the State of São Paulo Research Foundation (FAPESP).
Beatriz Cabral Fernandes, Daniel Costa Delattre, Luís Felipe Cunha Souza and Sophia Kerber Shigueoka — who will start the ninth semester of the program — used an open source framework that allows access to a Mbya Guarani translator via WhatsApp. Linguistic data, such as the Guarani Mbya to Portuguese dictionary, as well as user interaction data were hosted in the IBM cloud. Among the features are spelling correction, phrase translation and search and addition of pronunciations.
“Technically, this project allowed students to access corporate and cloud systems, via IBM Watson, and the application of microservices, web systems and deployment automation”, explains professor Tiago Tavares, group advisor. “But it was also important in the sense of reflecting on the social impact of technologies, because it deals with that kind of delicate situation for which we want to configure positively, but which potentially can lead to bad results.”
From the beginning of the project, students knew that the topic was complex. One of the concerns, for example, was with user authentication and data security. “It was important to ensure that all the phone numbers that sent messages belonged to people in the village, as they could add words and pronunciations”, says student Luís Felipe. “The project touched on ethical issues related to users of engineering projects that helped us evolve personally and professionally.”
Sophia says that learning with the Final Engineering Project exceeded expectations, especially in the application of the agile methodology, a series of project management values and principles born in the technology sector. This methodology encourages constant collaboration between work teams and product users. “We had the opportunity to apply content that, randomly, doesn’t seem to make much sense, but which is important to solve a real problem like that of the chatbot”, says Sophia.
The need arose in the context of teaching the Guarani Mbya language. A tool was sought to allow the translation of words into Portuguese on a user-friendly platform — in this case, teachers and high school students in indigenous communities. The idea was not to develop a specific application, but a feature in a software known by the majority of people. “WhatsApp solves some issues that are key to the proper functioning of the translator, such as running on older cell phones and unstable connections,” says Tavares.
In addition to the technical ability to master several new tools, the group demonstrated communication skills, a skill that was greatly stimulated during the program and in carrying out the Final Engineering Project. The students had weekly meetings with the IBM mentor and had to sporadically record the evolution of the project, later consolidated in the final report. “There was the additional component of thinking about an audience that is not necessarily as excited about a new technology as engineers are”, says the professor. “These are people who will receive the tool within a context not limited to consumption, but which may include technology co-creation.”
In charge of mentoring at IBM Research Brazil, the engineer Julio Nogima highlights the students’ engagement and focus on the company’s proposal and the practical knowledge not so usual for engineering graduates. “We wanted some IBM components to be used, and they quickly understood what had to be done, the concepts of these components and how to integrate them for the chatbot implementation”, says Nogima. “They did not limit themselves to executing what was initially proposed and asked relevant questions to improve the project. This exchange of ideas was enriching for both sides.”
The partnership established in the Final Engineering Project opens up an opportunity for growth for everyone. “On the students’ side, it’s an interesting experience since they receive a real problem to work on, something that will meet a real need for a real customer,” says Nogima. “For the company, it’s a way to explore paths that meet a need, with the help of students trained in computing. It is also a way for the company to contribute to the training of these students.”
The chatbot was presented to teachers and students from the Tenonde Porã Indigenous Land, located in the southern region of the city of São Paulo. Two of the students, Beatriz and Luís, accompanied the IBM team in the prototype presentation and demonstration session in the village.
The first impressions were positive, according to the team. Any platform that relates to people requires months or years of refinement to become pleasant and functional, observes Professor Tavares. And the Final Engineering Project is a four-month project that needs to have a beginning, middle and end. Within the scope of the IBM project, however, the application will be expanded to other indigenous languages based on the success achieved with Guarani Mbya.