Research & Development
Our Research
Language Technologies for Intelligent Contact Centers
Companies use contact centers to interact with customers aiming to provide them with technical support and assistance. automated self-service solutions such as chatbots, virtual assistants fail to provide intelligence assistance and continuous interactions with users. In addition, it is unclear how complex natural-language questions can be answered or routed to humans. This project developed an advanced multi-component natural-language processing framework to allow automated call routing, which cannot be identified using simple speech-supported call centers.
Source: iText Routing: Reconocimiento
Inteligente de Intenciones de usuarios para enrutamiento de llamadas.
Intelligent Fraud Detection for Power Supply Companies
Usually, fraud detection in non-registered power supply is based on
decisions of various specialists and human analysts, who have low prediction accuracy in the task of predicting frauds. This project developed several AI methods that combined allow power supply fraud detection fraud with almost 3-4 times effectiveness than human analysts.
Source: Desarrollan en Chile Sistema Inteligente para detectar Fraudes Eléctricos

Brain Computer Interfaces for Emotion Recognition
Automatic EEG emotion recognition is usually restricted to a small number of emotions classes mainly due to signal’s features and
noise, EEG constraints and subject-dependent issues. In order to address
these issues, a novel feature-based emotion recognition model was developed for EEG-based Brain–Computer Interfaces using kernel-based automatic classifiers on
standard EEG datasets.
Source: Improving BCI-based
Emotion Recognition by Combining EEG Feature Selection and Kernel Classifiers.

Automated Essay Assessment using Natural-Language Discourse Analysis
This project combined discourse analysis, embedding learning (i.e., LSA, BERT) and semantic and syntactic models for automated essay assessment. The approach combines shallow linguistic features and discourse patterns in order to predict an essay’s score by using machine learning techniques Unlike current approaches, the method directly measures an essay coherence by using corpus-based semantics and text centering techniques so as to determine discourse patterns underlying high-quality essays when compared with human assessed essays.
Autonomous Multi-Agent Robotics
This project developed a behavior-based system for a team of four-legged autonomous robots (AIBO Sony). This involved game dynamic
strategies, novel real-time computer vision systems, and negotiation techniques
for coordinating intelligent agents on dynamic game conditions.
Source: UdeCans Team Description.

Linguistically-motivated Opinion Mining
Social networks messaging typically contains a lot of
implicit linguistic information. This may significantly impact several tasks
including opinion mining and sentiment analysis, as opinion retrieval tasks
will fail to obtain all the relevant messages. In order to address these
issues, this project developed a novel adaptive approach for opinion retrieval
that combines natural-language co-referencing techniques, and memory-based learning to resolving implicit co-referencing within informal opinion texts.