Papers with document encoder
Phrase-Indexed Question Answering: A New Challenge for Scalable Document Comprehension (D18-1)
Copied to clipboard
| Challenge: | Existing QA models rely on learning interaction between document and question . current models require explicit attention to the document before or as it reads it . |
| Approach: | They propose a modular question answering task that enforces complete independence of the document encoder from the question encoder. |
| Outcome: | The proposed model achieves reasonable accuracy but significantly underperforms unconstrained QA models. |
Hierarchical Modeling of Global Context for Document-Level Neural Machine Translation (D19-1)
Copied to clipboard
| Challenge: | Document-level machine translation (MT) remains challenging due to the difficulty in efficiently using document context. |
| Approach: | They propose a hierarchical model to learn document context for document-level neural machine translation . they use a sentence encoder to capture intra-sentence dependencies and a document encoder . |
| Outcome: | The proposed model significantly improves document-level translation performance over strong baselines. |
Event-enhanced Retrieval in Real-time Search (2024.lrec-main)
Copied to clipboard
| Challenge: | Existing embedding-based retrieval models face the "semantic drift" problem . a low adoption rate of retrieval results is evident in real-time search scenarios . |
| Approach: | They propose an embedding-based retrieval approach that enhances real-time retrieval performance by adding contrastive learning to the dual-encoder model. |
| Outcome: | The proposed approach improves the dual-encoder model of traditional EBR. |
Investigating Multi-layer Representations for Dense Passage Retrieval (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Dense retrieval models adopt vectors from the last hidden layer of the document encoder to represent a document, which is in contrast to the fact that representations in different layers of a pre-trained language model contain different kinds of linguistic knowledge and behave differently during fine-tuning. |
| Approach: | They propose to utilize representations from multiple encoder layers to make up the representation of a document, which they denote Multi-layer Representations (MLR). |
| Outcome: | The proposed model outperforms dual encoder, ME-BERT and ColBERT in the single-vector retrieval setting and with other advanced training techniques. |
Topic Detection and Tracking with Time-Aware Document Embeddings (2024.lrec-main)
Copied to clipboard
| Challenge: | Topic Detection and Tracking (TDT) systems aim to cluster news articles into real-world events. |
| Approach: | They propose a neural method that fuses temporal and textual information into a single representation of news documents for event detection. |
| Outcome: | The proposed model outperforms baselines on two benchmark TDT data sets in English. |