Personalized chatbots to improve business automation

Chatbots or virtual agents are programs based on artificial intelligence that can hold a conversation with humans to attend to a range of situations, from a telephone call to a comment on social media. Researchers at the UOC have developed Xatkit, a new open source software platform capable of configuring chatbots with advanced functions that can be defined in line with a company's specific needs. In addition, the technology is aimed at enabling the future creation of families of personalized bots, which is a unique characteristic in the sector. The platform is also one of the first to offer virtual agents that speak and understand Catalan, as well as Spanish and English.

Funded as part of the Innovators 2019 call by the Directorate General for Research of the Government of Catalonia, the project is coordinated by Jordi Cabot, ICREA research professor and leader of the Systems, Software and Models group (SOM RESEARCH LAB), at the UOC's Internet Interdisciplinary Institute (IN3). The call includes co-funding by the European Regional Development Fund (ERDF) and is intended for value building and knowledge transfer projects developed by innovators on placements with organizations in the research and innovation system in Catalonia.

An open, highly flexible platform

Chatbots have become widespread in recent years thanks to their ability to automate processes, especially in the customer support and e-commerce sectors. According to Jordi Cabot, "Initially these programs took care of customers' most frequent questions, but step by step they've grown in complexity and have started to take on increasingly more complex conversations as well as management tasks, which involves not only text replies to customers, but also the automatic processing of their requests, including phone calls".

These features enable reduced costs and an improved user experience, meaning customers are able to contact the company 24/7. Not all chatbots are the same, though, as every company offers different services and has its own management environment. In addition, these are highly complex programs that require specialist platforms to define their functions, ensure the quality of their training, and test and monitor their behaviour. The SOM research team has a lot of experience in this type of computer architecture and has applied this knowledge to the construction of a personalized bot development solution. The researcher explained that "Xatkit is aimed at businesses with advanced and highly specific needs in their day-to-day operations. It's an open and very flexible platform that enables the bespoke creation of bots that interact with a company's internal information services and can be integrated with a large number of platforms, from chats and social media to messaging managers". 

Xatkit offers businesses a complete service to advise them in the definition, design and maintenance of virtual agents; however, the use of open source software allows clients to configure chatbots easily on their own and install them on their servers if they prefer. Jordi Cabot said: "We use a bot definition language that aims to facilitate this process. It's one of the only platforms to use the low-code system, the aim of which is to create software by writing little code thanks to programming languages that are more geared towards the domain in which you're working, in this case, the domain of bots."

In addition, the aim in the next few months is to generate families of chatbots, a feature that no other solution provides at present. According to the researcher, "normally, businesses need a lot of virtual agents to meet different needs or similar needs with slight variations, such as multinationals or franchises that have to adapt their services to different countries or languages. Xatkit is the only platform that is explicitly geared towards the problem of generating more than one bot at the same time".

Technology to maximize the quality of conversation

The platform uses the latest trends in natural language processing and understanding to maximize the quality of the conversations. These techniques are used not only to understand what the bot user is trying to say, in other words, the basic function of any chatbot, but also to offer functions such as automatic text summary, sentiment analyses and machine translation, which may be useful as part of a chatbot's functions. Jordi Cabot said: "These features are useful, for example, for visitors to a website who don't want to read the whole content of a post or to ascertain whether a customer is angry or not, a very useful skill in a customer support service where the bot could modulate its answer according to the customer's temper."

Another of the platform's highlighted features is that its chatbots speak Catalan. "Support in English or Spanish is very common, but thanks to integration with an open-source library (https://github.com/axa-group/nlp.js), Xatkit defines bots that understand and speak Catalan, which is very difficult to find at present," the researcher added.

What should educational spaces be like in the twenty-first century?

The Innovators 2019 call has also funded the Smart Classroom project, led by Guillermo Bautista, researcher with the UOC Faculty of Psychology and Education Sciences, which, in light of this grant, has brought researcher Maria Casanovas on board as well. This research takes a new look from a pedagogical perspective at learning spaces in education centres at any level, from nursery school to university, and, for example, at how they can be adapted to any learning need and situation, as well as how the space may lead to an improvement in teaching and learning processes. One of the initial results has been to establish a series of principles that offer guidance in the design of these spaces. The researchers have also published a teaching ten-point plan to adapt classrooms to the exceptional situation caused by the pandemic and to offer guidance on what spaces should be like in order to meet the physical distancing and hygiene measures put in place due to coronavirus.

Leakage detection sensor for vehicle manufacturing

The Government of Catalonia has also awarded funding for a project led by Joan Melià, researcher, and Xavier Vilajosana, full professor, from the Wireless Networks Research Lab (WiNe) at the IN3. The goal of their project is to offer a low cost passive and wireless moisture sensor to detect water leakage in vehicle interiors. This device is aimed at solving one of the most important manual operation problems in the automotive industry: water-tightness testing of vehicles. Positioned in key points around the vehicle, sensors produced using WiLD (Wireless Leakage Detection System in vehicle manufacturing lines) technology can detect leaks automatically, without human intervention. This significantly reduces assembly time while also minimizing workers' exposure to high humidity levels. The grant awarded for the development of this project is part of the Catalan Government's knowledge industry programme, specifically in its Llavor (seed) category, for innovative projects with potential uses in the productive sector.

UOC R&I

The UOC's research and innovation (R&I) contribute to solving the challenges facing the global societies of the 21st century by studying ICTs' interactions with human activity, with a specific focus on e-learning and e-health. Over 400 researchers and 50 research groups work among the University's 7 faculties and 3 research centres: the Internet Interdisciplinary Institute (IN3), the eLearn Center (eLC) and the eHealth Center (eHC).

The United Nations 2030 Agenda's Sustainable Development Goals and open knowledge provide strategic pillars on which the UOC's teaching, research and innovation are built. More information: research.uoc.edu.

The Xatkit project (Massive Generation of chatbots on-demand, ref. 2019 INNOV 00001), the SmartClassroom project (Innovative Educational Methodologies to Design Learning Spaces, ref. 2019 INNOV 00007), and the WiLD project (Wireless Leakage Detection system in the vehicle manufacturing lines, ref. 2019 LLAV 00005) are supported by the Department for Universities and Research in the Government of Catalonia's Ministry of Business and Knowledge and receive funding from the European Regional Development Fund (ERDF).