Improving Document Ranking with Dual Word Embeddings

By Eric T. Nalisnick, Bhaskar Mitra, Nick Craswell, R. Caruana. 2016

174 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Information Retrieval & Search

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Information Retrieval & Search

Citations in the corpus

End-to-End Neural Ad-hoc Ranking with Kernel Pooling (2017, methodology): “Nalisnick et al. propose to match query and documents using both the input and output of the embedding model, instead of only using one side of them [19].”
Pretrained Transformers for Text Ranking: BERT and Beyond (2020, methodology): “ity. Shen et al. [2014] improved upon DSSM by using CNNs to capture context. Rather than learning text representations as part of the model, the Dual Embedding Space Model (DESM) [Mitra et al., 2016, Nalisnick et al., 2016] represents texts using pre-trained word2vec embeddings [Le and Mikolov, 2014”
Man is to Computer Programmer as Woman is to Homemaker? Debiasing Word Embeddings (2016, background): “There have been hundreds or thousands of papers written about word embeddings and their applications, from Web search [27] to parsing Curriculum Vitae [16].”
Uncovering and Mitigating Algorithmic Bias through Learned Latent Structure (2019, background): “For example, ML and artificial intelligence (AI) are already being used to control end-to-end autonomous vehicles [2], how long a criminal should spend in prison [6], the order in which a person is presented the news [27], or even diagnose and treat medical patients [24].”
Learning to Match using Local and Distributed Representations of Text for Web Search (2016, methodology): “[28] used dual embeddings, one for document terms and one for query terms, then ranked based on the all-pairs similarity between vectors.”
Quantifying and Reducing Stereotypes in Word Embeddings (2016, methodology): “…recent project has shown that, when carefully combined with existing approaches, word vectors can significantly improve Web page relevance results (Nalisnick et al., 2016) (note that this work is a proof of concept – we do not know which, if any, mainstream search engines presently incorporate…”
BISON: BM25-weighted Self-Attention Framework for Multi-Fields Document Search (2020, unknown): “Similar with (18; 20; 11) we sample 30 million query-document pairs from Bing’s search log.”
A Dual Embedding Space Model for Document Ranking (2016, unknown): “[32] to appear in WWW’16, April 11 - 15, 2016, Montreal, Canada.”
Retrieve-and-Read: Multi-task Learning of Information Retrieval and Reading Comprehension (2018, background): “C, including the eﬀect of adding adversarial distracting sentences [23]. Neural IR. Neural ranking models are currently a populartopic in IR [30]. There are roughly two groups: embedding space models [19, 32, 41] and interaction-based models [10, 16, 20, 21, 51]. In the embedding space models, the q”
Query Expansion Using Word Embeddings (2016, methodology): “Word embeddings were used to re-weight the query terms [20] and to compare a query directly with a document [16, 12].”
Conformer-Kernel with Query Term Independence for Document Retrieval (2020, methodology): “ng of terms or concepts may be difﬁcult using query-independent latent document representations [Luan et al., 2020], and therefore these models are often combined with explicit term matching methods [Nalisnick et al., 2016, Mitra et al., 2017]. Xiong et al. [2020] have recently demonstrated that the”
Incorporating Query Term Independence Assumption for Efficient Retrieval and Ranking using Deep Neural Networks (2019, methodology): “Several neural IR methods—e.g., [Ganguly et al., 2015, Kenter and De Rijke, 2015, Nalisnick et al., 2016, Guo et al., 2016]—already operate under query term independence assumption.”
Analysis of the Paragraph Vector Model for Information Retrieval (2016, methodology): “[18] represented both queries and documents with a composition of word vectors and performed ranking based on the cosine similarity between them.”
Query Expansion with Locally-Trained Word Embeddings (2016, methodology): “(Nalisnick et al., 2016) proposed to model document aboutness by computing the similarity between all pairs of query and document terms using dual embedding spaces.”
Neural information retrieval: at the end of the early years (2017, methodology): “…(2015), Mitra and Craswell (2015), Sordoni et al. (2015) Bing Query Logs and WebCrawl Mitra (2015), Mitra and Craswell (2015), Mitra et al. (2016), Nalisnick et al. (2016) CLEF03 English Ad-hoc Clinchant and Perronnin (2013) CLEF 2016 Social Book Search Amer et al. (2016) CLEF Medical Corpora…”
EXS: Explainable Search Using Local Model Agnostic Interpretability (2018, methodology): “Next we implemented and trained several neural networks like DRMM [6] and DESM [11] using Ten-sorFlow.”
The Geometry of Culture: Analyzing the Meanings of Class through Word Embeddings (2018, methodology): “Although we do not make use of them in this analysis, the m contexts by k dimensions matrix in Figure 1 also retains a great deal of semantic information and has been used in concert with word embeddings to identify words that are complements versus substitutes in text (Nalisnick et al. 2016; R”
Estimating Embedding Vectors for Queries (2016, methodology): “Following the impressive results achieved by word embedding techniques in NLP tasks, these techniques have begun to be studied in IR tasks [28, 35, 39, 40].”
Quantifying Gender Bias in Different Corpora (2020, background): “Word embeddings have been used in a wide variety of natural language processing applications from web search to resumé parsing [7].”