Making Pepper Understand and Respond in Romanian | Semantic Scholar (2025)

A modular architecture based on APIs allowing an integration engine (either in the robot or external) to chain together the available modules as needed is designed, targeting the ROBIN project's goals ofdialogue in closed micro-worlds.

The creation of a core vocabulary to be learned by a robot is extracted from two tokenised and lemmatized scenarios pertaining to two imagined microworlds in which the robot is supposed to play an assistive role.

An unsupervised syllabification approach that incorporates logical reasoning into the reinforcement learning training process, achieving knowledge-guided syllabification in low-resource Lao language is proposed.

An ASR system for the Romanian language that uses a multi-layer neural network architecture to transcribe the input speech, augmented with a statistical language model to improve the correctness of the transcriptions is presented.

CoRoLa contains original, IPR-cleared texts and is representative for the contemporary phase of the language, covering mostly the last 20 years, and can be queried using the KorAP corpus management platform.

This paper describes a study aiming at finding an objective evaluation procedure of the dialogue with a social robot to build an empathic robot and it focuses on elderly people, the end-users expected by ROMEO2 project.

The semantic network for Romanian is presented, started to add a series of new lexical relations (language specific) based on an in-depth analysis of the Romanian derivational morphology, and the results of a word sense disambiguation algorithm based on wordnets are presented.

It is proved that CoRoLa provides superior word representations compared to other known Romanian corpora, such as the Wikipedia corpus.

This paper is a case report that explains the developmental process of the application that contains three educational programs that children can select in interacting with Pepper, a personal robot developed by SoftBank Robotics Corp. and Aldebaran Robotics SAS.

This paper focuses on the development of small robust deep-learning models that are designed to provide high quality text-to-speech functionality (TTS) on smart devices, without requiring network access.

The best result is obtained from splicing the baseline 40-dimensional speaker adapted features again across 9 frames, followed by reducing the dimension to 200 or 300 using another LDA, which is about 3% absolute better than the best GMM system.

Char2Wav is an end-to-end model for speech synthesis that learns to produce audio directly from text.

...

...