Titre :	Using Multitask Unified Model (MUM) to improve search and NLU
Type de document :	texte imprimé
Auteurs :	Aymen Abderrahim Redjechta, Auteur ; Aya Lamria Rouabhi, Auteur ; Harrag,Fouzi, Directeur de thèse
Année de publication :	2022
Importance :	1 vol (99 f .)
Format :	29cm
Langues :	Français (fre)
Catégories :	Thèses & Mémoires:Informatique
Mots-clés :	Natural language processing
Index. décimale :	004 Informatique
Résumé :	After BERT, the introduction of MUM is the next significant major shift for Google search. While earlier machine learning-based innovations relied on a number of trained models for various tasks, the objective of MUM is to employ just one model for all indexing, information retrieval, and ranking-related tasks. In this work, we aim to use the MUM concepts to retrieve images and generate them from the Arabic dataset. When it comes to natural language processing (NLP), Arabic is a highly difficult language that is rarely employed the new deep learning techniques. Open Ai CLIP is one the best new deep learning model for Connecting Text and Images wich can map both images and text, alongside with Open Ai Dall for generating images. In our project, we will be fine-tuning the model to extract images and generate them.
Côte titre :	MAI/0619
En ligne :	https://drive.google.com/file/d/1MxPmzAXceOLsvk5T6ry8g7rnYkHHioMe/view?usp=share [...]
Format de la ressource électronique :	pdf

Using Multitask Unified Model (MUM) to improve search and NLU [texte imprimé] / Aymen Abderrahim Redjechta, Auteur ; Aya Lamria Rouabhi, Auteur ; Harrag,Fouzi, Directeur de thèse . - 2022 . - 1 vol (99 f .) ; 29cm.
Langues : Français (fre)

Catégories :	Thèses & Mémoires:Informatique
Mots-clés :	Natural language processing
Index. décimale :	004 Informatique
Résumé :	After BERT, the introduction of MUM is the next significant major shift for Google search. While earlier machine learning-based innovations relied on a number of trained models for various tasks, the objective of MUM is to employ just one model for all indexing, information retrieval, and ranking-related tasks. In this work, we aim to use the MUM concepts to retrieve images and generate them from the Arabic dataset. When it comes to natural language processing (NLP), Arabic is a highly difficult language that is rarely employed the new deep learning techniques. Open Ai CLIP is one the best new deep learning model for Connecting Text and Images wich can map both images and text, alongside with Open Ai Dall for generating images. In our project, we will be fine-tuning the model to extract images and generate them.
Côte titre :	MAI/0619
En ligne :	https://drive.google.com/file/d/1MxPmzAXceOLsvk5T6ry8g7rnYkHHioMe/view?usp=share [...]
Format de la ressource électronique :	pdf