Matches in SemOpenAlex for { <https://semopenalex.org/work/W4385571551> ?p ?o ?g. }
Showing items 1 to 77 of
77
with 100 items per page.
- W4385571551 abstract "Dense retrieval has shown promise in the first-stage retrieval process when trained on in-domain labeled datasets. However, previous studies have found that dense retrieval is hard to generalize to unseen domains due to its weak modeling of domain-invariant and interpretable feature (i.e., matching signal between two texts, which is the essence of information retrieval). In this paper, we propose a novel method to improve the generalization of dense retrieval via capturing matching signal called BERM. Fully fine-grained expression and query-oriented saliency are two properties of the matching signal. Thus, in BERM, a single passage is segmented into multiple units and two unit-level requirements are proposed for representation as the constraint in training to obtain the effective matching signal. One is semantic unit balance and the other is essential matching unit extractability. Unit-level view and balanced semantics make representation express the text in a fine-grained manner. Essential matching unit extractability makes passage representation sensitive to the given query to extract the pure matching information from the passage containing complex context. Experiments on BEIR show that our method can be effectively combined with different dense retrieval training methods (vanilla, hard negatives mining and knowledge distillation) to improve its generalization ability without any additional inference overhead and target domain data." @default.
- W4385571551 created "2023-08-05" @default.
- W4385571551 creator A5004759804 @default.
- W4385571551 creator A5029998682 @default.
- W4385571551 creator A5047897879 @default.
- W4385571551 creator A5051296316 @default.
- W4385571551 date "2023-01-01" @default.
- W4385571551 modified "2023-09-24" @default.
- W4385571551 title "BERM: Training the Balanced and Extractable Representation for Matching to Improve Generalization Ability of Dense Retrieval" @default.
- W4385571551 doi "https://doi.org/10.18653/v1/2023.acl-long.365" @default.
- W4385571551 hasPublicationYear "2023" @default.
- W4385571551 type Work @default.
- W4385571551 citedByCount "0" @default.
- W4385571551 crossrefType "proceedings-article" @default.
- W4385571551 hasAuthorship W4385571551A5004759804 @default.
- W4385571551 hasAuthorship W4385571551A5029998682 @default.
- W4385571551 hasAuthorship W4385571551A5047897879 @default.
- W4385571551 hasAuthorship W4385571551A5051296316 @default.
- W4385571551 hasBestOaLocation W43855715511 @default.
- W4385571551 hasConcept C105795698 @default.
- W4385571551 hasConcept C119857082 @default.
- W4385571551 hasConcept C134306372 @default.
- W4385571551 hasConcept C138885662 @default.
- W4385571551 hasConcept C151730666 @default.
- W4385571551 hasConcept C153180895 @default.
- W4385571551 hasConcept C154945302 @default.
- W4385571551 hasConcept C165064840 @default.
- W4385571551 hasConcept C177148314 @default.
- W4385571551 hasConcept C17744445 @default.
- W4385571551 hasConcept C199539241 @default.
- W4385571551 hasConcept C2776214188 @default.
- W4385571551 hasConcept C2776359362 @default.
- W4385571551 hasConcept C2776401178 @default.
- W4385571551 hasConcept C2779343474 @default.
- W4385571551 hasConcept C33923547 @default.
- W4385571551 hasConcept C41008148 @default.
- W4385571551 hasConcept C41895202 @default.
- W4385571551 hasConcept C52622490 @default.
- W4385571551 hasConcept C86803240 @default.
- W4385571551 hasConcept C94625758 @default.
- W4385571551 hasConceptScore W4385571551C105795698 @default.
- W4385571551 hasConceptScore W4385571551C119857082 @default.
- W4385571551 hasConceptScore W4385571551C134306372 @default.
- W4385571551 hasConceptScore W4385571551C138885662 @default.
- W4385571551 hasConceptScore W4385571551C151730666 @default.
- W4385571551 hasConceptScore W4385571551C153180895 @default.
- W4385571551 hasConceptScore W4385571551C154945302 @default.
- W4385571551 hasConceptScore W4385571551C165064840 @default.
- W4385571551 hasConceptScore W4385571551C177148314 @default.
- W4385571551 hasConceptScore W4385571551C17744445 @default.
- W4385571551 hasConceptScore W4385571551C199539241 @default.
- W4385571551 hasConceptScore W4385571551C2776214188 @default.
- W4385571551 hasConceptScore W4385571551C2776359362 @default.
- W4385571551 hasConceptScore W4385571551C2776401178 @default.
- W4385571551 hasConceptScore W4385571551C2779343474 @default.
- W4385571551 hasConceptScore W4385571551C33923547 @default.
- W4385571551 hasConceptScore W4385571551C41008148 @default.
- W4385571551 hasConceptScore W4385571551C41895202 @default.
- W4385571551 hasConceptScore W4385571551C52622490 @default.
- W4385571551 hasConceptScore W4385571551C86803240 @default.
- W4385571551 hasConceptScore W4385571551C94625758 @default.
- W4385571551 hasLocation W43855715511 @default.
- W4385571551 hasOpenAccess W4385571551 @default.
- W4385571551 hasPrimaryLocation W43855715511 @default.
- W4385571551 hasRelatedWork W1964120219 @default.
- W4385571551 hasRelatedWork W2000165426 @default.
- W4385571551 hasRelatedWork W2114557664 @default.
- W4385571551 hasRelatedWork W2144059113 @default.
- W4385571551 hasRelatedWork W2146076056 @default.
- W4385571551 hasRelatedWork W2385132419 @default.
- W4385571551 hasRelatedWork W2546942002 @default.
- W4385571551 hasRelatedWork W2772780115 @default.
- W4385571551 hasRelatedWork W2811390910 @default.
- W4385571551 hasRelatedWork W3003836766 @default.
- W4385571551 isParatext "false" @default.
- W4385571551 isRetracted "false" @default.
- W4385571551 workType "article" @default.