Paulin Melatagia

• PhD
• Professor (Associate) at University of Yaoundé I

soumission à Episciences Under-resourced languages encounter substantial obstacles in speech recognition owing to the scarcity of resources and limited data availability, which impedes their development and widespread adoption. This paper presents a representation learning model that leverages existing frameworks based on self-supervised learning t...

Recently popularized self-supervised models appear as a solution to the problem of low data availability via parsimonious learning transfer. We investigate the effectiveness of these multilingual acoustic models, in this case wav2vec 2.0 XLSR-53 and wav2vec 2.0 XLSR-128, for the transcription task of the Ewondo language (spoken in Cameroon). The ex...

Many sub-Saharan African languages are categorized as tone languages and for the most part, they are classified as low resource languages due to the limited resources and tools available to process these languages. Identifying the tone associated with a syllable is therefore a key challenge for speech recognition in these languages. We propose mode...

Distributed learning allows the implementation of algorithms that require much more computing power or memory capacity than a single machine can provide. However, this poses the problem of the confidence that must be placed in the work of each of the computing nodes. In this paper we are interested in the ability of the distributed gradient descent...

Prosody is a key area of linguistics that explores tonal and rhythmic variations in speech. In tonal languages such as Yemba, prosody plays a crucial role in distinguishing between words with different meanings or different grammatical forms. However, despite the large number of native speakers of this language in Cameroon, there are few resources...

International audience Reinforcement learning algorithms have succeeded over the years in achieving impressive results in a variety of fields. However, these algorithms suffer from certain weaknesses highlighted by Refael Vivanti and al. that may explain the regression of even well-trained agents in certain environments : the difference in variance...

Using the headers of scientific papers, we have built multilayer networks of entities involved in research namely: authors, laboratories, and institutions. We have analyzed some properties of such networks built from data extracted from the HAL archives and found that the network at each layer is a small-world network with power law distribution. I...

International audience Named Entity Recognition (NER) is a fundamental task in many NLP applications that seek to identify and classify expressions such as people, location, and organization names. Many NER systems have been developed, but the annotated data needed for good performances are not available for low-resource languages, such as Cameroon...

Named entity recognition is an important task in natural language processing. It is very well studied for rich language, but still under explored for low-resource languages. The main reason is that the existing techniques required a lot of annotated data to reach good performance. Recently, a new distributional representation of words has been prop...

Named Entity Recognition (NER) is a fundamental task in many NLP applications that seek to identify and classify expressions such as people, location, and organization names. Many NER systems have been developed, but the annotated data needed for learning is not available for low-resource languages, such as Cameroonian languages. In this paper we e...

Neural networks are gaining popularity in software engineering. This paper presents a dedicated API and visual environment to train and use a neural networks on software source code related data. This short paper illustrates the API using two examples involving prediction of source code properties.

We propose a model of growing networks based on cliques formations. A clique is used to illustrate for example co-authorship in co-publication networks, co-occurence of words or collaboration between actors of the same movie. Our model is iterative and at each step, a clique of λη existing vertices and (1 − λ)η new vertices is created and added in...

We present an online algorithm for routing the automorphisms (BPC permutations) of the queueless MIMD hypercube. The routing algorithm has the virtue of being executed by each node of the hypercube without knowing the state of the others nodes. The algorithm is also vertex and link-contention free. We show, using the proposed algorithm, that BPC pe...

In this paper, we study the convergence of a mathematical model of opinion dynamics called the majority model. In this model, at each iteration step, each individual adopts the opinion which exerts on him the maximum social pressure. Under some assumptions on interaction among members of the society, we show that, in parallel mode, attractors of th...

Given an undirected and connected graph G, with a non-negative weight on each edge, the Minimum Average Distance (MAD) spanning tree problem is to find a spanning tree of G which minimizes the average distance between pairs of vertices. This network design problem is known to be NP-hard even when the edge-weights are equal. In this paper we make a...

In this paper, we study memory iteration where the updating considers a longer history of each site and the set of interaction matrices is quasi-palindromic. For parallel and sequential iterations with memory, we define Lyapunov functional which permits us to characterize the periods behaviour and explicitly bounds the transient lengths of quasi-pa...

We study memory iteration where the updating consider a longer history of each site and the set of interaction matrices is palindromic. We analyze two different ways of updating the net-works: parallel iteration with memory and sequential iteration with memory that we introduce in this paper. For parallel iteration, we define Lyapunov functional wh...