Automatic generation of product descriptions / Behilil ,Hamza  

Public ISBD Titre : Automatic generation of product descriptions Type de document : texte imprimé Auteurs : Behilil ,Hamza, Auteur ; Harrag,Fouzi, Auteur Année de publication : 2021 Importance : 1 vol (57 f .) Format : 29 cm Langues : Français (fre) Catégories : Thèses & Mémoires:Informatique Mots-clés : Product Description Text Generation Index. décimale : 004 - Informatique Résumé : Product descriptions play an important role in e-commerce websites, where they give to customers the information needed about the product they looking for. These websites witness a daily addition of a large number of new products especially in fashion domain, each product needs to have its specific description. Writing these descriptions manually is a tedious and time consuming process, so our system comes to solve this problem by presenting an NLP system based on a transformer model which is GPT-Neo with the help of a word embedding model (word2vec) to generate these descriptions automatically. Our 14k dataset was collected from a famous arabic e-commerce website called noon and it contains titles and descriptions of different categories in fashion domain. We used our dataset to train and fine-tune word2vec and GPT-Neo models respectively then, we combined them to create our system which takes product title as input and give product description as output. Evaluation of our systems showed promising results with 67% of recall and 72% of precision. Côte titre : MAI/0533 En ligne : https://drive.google.com/file/d/1lTdODus5T0OPVo4IO2g8PInCaCuIANT5/view?usp=shari [...] Format de la ressource électronique : pdf Automatic generation of product descriptions [texte imprimé] / Behilil ,Hamza, Auteur ; Harrag,Fouzi, Auteur . - 2021 . - 1 vol (57 f .) ; 29 cm. Langues : Français (fre) Catégories : Thèses & Mémoires:Informatique Mots-clés : Product Description Text Generation Index. décimale : 004 - Informatique Résumé : Product descriptions play an important role in e-commerce websites, where they give to customers the information needed about the product they looking for. These websites witness a daily addition of a large number of new products especially in fashion domain, each product needs to have its specific description. Writing these descriptions manually is a tedious and time consuming process, so our system comes to solve this problem by presenting an NLP system based on a transformer model which is GPT-Neo with the help of a word embedding model (word2vec) to generate these descriptions automatically. Our 14k dataset was collected from a famous arabic e-commerce website called noon and it contains titles and descriptions of different categories in fashion domain. We used our dataset to train and fine-tune word2vec and GPT-Neo models respectively then, we combined them to create our system which takes product title as input and give product description as output. Evaluation of our systems showed promising results with 67% of recall and 72% of precision. Côte titre : MAI/0533 En ligne : https://drive.google.com/file/d/1lTdODus5T0OPVo4IO2g8PInCaCuIANT5/view?usp=shari [...] Format de la ressource électronique : pdf

Exemplaires (1)

Code-barres	Cote	Support	Localisation	Section	Disponibilité
MAI/0533	MAI/0533	livre	Bibliothéque des sciences	Anglais	Disponible Disponible

---

1 (1 - 1 / 1)