Bi-Encoder
المُرمِّز الثنائي
بنية تستخدم مُرمِّزَين منفصلَين (أو مُرمِّزًا واحدًا بأوزان مختلفة) لتمثيل الاستعلام والمستند كمتجهَين مستقلَّين، ثم تقيس التشابه بينهما عبر الضرب النقطي. تتيح ترميز المستندات مسبقاً للاسترجاع السريع.
An architecture using two separate encoders (or one with different weights) to independently encode query and document into vectors, then measuring similarity via dot product. Enables pre-encoding documents for fast retrieval.
Also translated asالبنية ثنائية المُرمِّز، نموذج المُرمِّزَين المنفصلَين
First appears in this corpus in: Sentence-BERT: Sentence Embeddings Using Siamese BERT-Networks (2019)
Appears in these papers
- ColBERT: Efficient and Effective Passage Search via Contextualized Late Interaction over BERT2020in the sky ✦
- ColBERT: Efficient and Effective Passage Search via Contextualized Late Interaction over BERT2020in the sky ✦
- Contriever: Unsupervised Dense Information Retrieval with Contrastive Learning2022in the sky ✦
- Contriever: Unsupervised Dense Information Retrieval with Contrastive Learning2022in the sky ✦
- Dense Passage Retrieval for Open-Domain Question Answering2020in the sky ✦
- Sentence-BERT: Sentence Embeddings Using Siamese BERT-Networks2019in the sky ✦
- Sentence-BERT: Sentence Embeddings Using Siamese BERT-Networks2019in the sky ✦