All translations
Enter a message name below to show all available translations.
Found 3 translations.
| Name | Current message text |
|---|---|
| h English (en) | {{Term|bert}} addressed this limitation by introducing a novel {{Term|pre-training}} objective — '''{{Term|masked language model|masked language modeling}}''' ({{Term|masked language model|MLM}}) — that enables genuine bidirectional {{Term|pre-training}}. Combined with a '''{{Term|next sentence prediction}}''' ({{Term|next sentence prediction|NSP}}) task, {{Term|bert}} learned rich {{Term|contextual embedding|contextual representations}} that could be transferred to downstream tasks through simple {{Term|fine-tuning}}, eliminating the need for task-specific architectures. |
| h Spanish (es) | {{Term|bert|BERT}} abordó esta limitación introduciendo un nuevo objetivo de {{Term|pre-training|preentrenamiento}} — el '''{{Term|masked language model|modelado de lenguaje enmascarado}}''' ({{Term|masked language model|MLM}}) — que permite un {{Term|pre-training|preentrenamiento}} bidireccional genuino. Combinado con una tarea de '''{{Term|next sentence prediction|predicción de la siguiente oración}}''' ({{Term|next sentence prediction|NSP}}), {{Term|bert|BERT}} aprendió ricas {{Term|contextual embedding|representaciones contextuales}} que podían transferirse a tareas posteriores mediante un simple {{Term|fine-tuning|ajuste fino}}, eliminando la necesidad de arquitecturas específicas de cada tarea. |
| h Chinese (zh) | {{Term|bert|BERT}} 通过引入一种新颖的 {{Term|pre-training|预训练}} 目标——'''{{Term|masked language model|掩码语言建模}}'''（{{Term|masked language model|MLM}}）——解决了这一限制，从而实现了真正的双向 {{Term|pre-training|预训练}}。结合 '''{{Term|next sentence prediction|下一句预测}}'''（{{Term|next sentence prediction|NSP}}）任务，{{Term|bert|BERT}} 学到了丰富的 {{Term|contextual embedding|上下文表示}}，这些表示可以通过简单的 {{Term|fine-tuning|微调}} 迁移到下游任务，消除了对任务特定架构的需求。 |