Observed versus latent features for knowledge base and text inference

By Kristina Toutanova, Danqi Chen. 2015

1,193 citations recorded by the source corpus.

In this paper we show the surprising effectiveness of a simple observed features model in comparison to latent feature models on two benchmark knowledge base completion datasets, FB15K and WN18. We also compare latent and observed feature models on a more challenging dataset derived from FB15K, and additionally coupled with textual mentions from a web-scale corpus. We show that the observed features model is most effective at capturing the information present for entity pairs with textual relations, and a combination of the two combines the strengths of both model types.

Subfields: Artificial Intelligence

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

Citations in the corpus

Modeling Relational Data with Graph Convolutional Networks (2017, methodology): “This improvement is especially large when we consider the more challenging FB15k-237 dataset (Toutanova and Chen 2015).”
A Review of Relational Machine Learning for Knowledge Graphs (2015, background): “, [117]), as both families focus on different aspects of relational data.”
Representing Text for Joint Embedding of Text and Knowledge Bases (2015, methodology): “We also limit the candidate entities to ones that have types consistent with the position in the relation triple (Chang et al., 2014; Yang et al., 2015), where the types are approximated following Toutanova and Chen (2015).”
Convolutional 2D Knowledge Graph Embeddings (2017, background): “• Previous work by Toutanova and Chen [31] noted that FB15k and WN18 contain many redundant, reversible relations, but they did not investigate the severity of this problem.”
Answering Visual-Relational Queries in Web-Extracted Knowledge Graphs (2017, background): “Any KG derived from FB15k such as FB15k-237[Toutanova and Chen, 2015] can also be associated with the crawled images.”
Leveraging Term Banks for Answering Complex Questions: A Case for Sparse Vectors (2017, methodology): “Khashabi et al. (2016) applied ILP to knowledge in a tabular form, using the same tables as Jauhar et al. (2016).”
Quaternion Knowledge Graph Embeddings (2019, background): “FB15K-237 [Toutanova and Chen, 2015] is a subset of FB15K, with inverse relations removed.”
A Survey of Large Language Models (2023, neutral): “Knowledge Completion WikiFact [307], FB15k-237 [308], Freebase [309], WN18RR [310], WordNet [311], LAMA [312], YAGO3-10 [313], YAGO [314]”
Adaptive Convolution for Multi-Relational Learning (2019, background): “(3) We systematically compare the efﬁciency and effectiveness of ConvR and ConvE on FB15K-237, showing that ConvR can perform substantially better with a good variety of parameter settings.”
Know-Evolve: Deep Temporal Reasoning for Dynamic Knowledge Graphs (2017, background): “(Nickel et al., 2016a) and (Yang et al., 2015; Toutanova & Chen, 2015) contains comprehensive reviews and empirical comparison of relational learning techniques respectively.”
End-to-End Structure-Aware Convolutional Networks for Knowledge Base Completion (2018, neutral): “The FB15k-237 (Toutanova and Chen 2015) dataset contains knowledge base relation triples and textual mentions of Freebase entity pairs, as used in the work published in (Toutanova and Chen 2015).”
DRUM: End-To-End Differentiable Rule Mining On Knowledge Graphs (2019, methodology): “In the experiment we use WN18RR [10], a subset of WordNet, and FB15K-237 [41], which both are more challenging versions of WN18 and FB15K [4] respectively.”
Canonical Tensor Decomposition for Knowledge Base Completion (2018, methodology): “To avoid clutter, we did not include in our table of results algorithms that make use of external data such as types (Krompass et al., 2015), external word embeddings (Toutanova & Chen, 2015), or using path queries as regularizers (Guu et al., 2015).”
Revisiting Evaluation of Knowledge Base Completion Models (2020, methodology): “For each target triple, we compute a local score by ﬁnding all paths from the subject to object and score them in the context of the relation of target triple (a simpler version of this local score is studied in Toutanova and Chen [2015]).”
Composition-based Multi-Relational Graph Convolutional Networks (2019, methodology): “In our experiments, we utilize FB15k-237 (Toutanova & Chen, 2015) and WN18RR (Dettmers et al., 2018) datasets for evaluation.”
Knowledge Graph Completion: A Review (2020, background): “Semantic information can be used as additional information to assist model feature extraction to improve performance [110].”
Open-World Knowledge Graph Completion (2017, background): “But FB15k is fraught with reversed-or synonym-triples (Toutanova and Chen 2015) and does not provide suf-ﬁcient textual information for content-based KGC methods to use.”