Mapping Text to Knowledge Graph Entities using Multi-Sense LSTMs

Kartsaklis, Dimitri; Pilehvar, Mohammad Taher; Collier, Nigel

doi:10.18653/v1/d18-1221

Cited by 43 publications

(55 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Question-answering with knowledge graphs. Our work is also related to the domain of question answering and reasoning in knowledge graphs (Das et al, 2018;Xiong et al, 2018;Hamilton et al, 2018;Xiong et al, 2017;Welbl et al, 2018;Kartsaklis et al, 2018), where either the model is provided with a knowledge graph to perform inference over or where the model must infer a knowledge graph from the text itself. However, unlike previous benchmarks in this domain-which are generally transductive and focus on leveraging and extracting knowledge graphs as a source of background knowledge about a fixed set of entities-CLUTRR requires inductive logical reasoning, where every example requires reasoning over a new set of previously unseen entities.…”

Section: Related Workmentioning

confidence: 99%

CLUTRR: A Diagnostic Benchmark for Inductive Reasoning from Text

Sinha¹,

Sodhani²,

Dong³

et al. 2019

Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conferen

View full text Add to dashboard Cite

The recent success of natural language understanding (NLU) systems has been troubled by results highlighting the failure of these models to generalize in a systematic and robust way. In this work, we introduce a diagnostic benchmark suite, named CLUTRR, to clarify some key issues related to the robustness and systematicity of NLU systems. Motivated by classic work on inductive logic programming, CLUTRR requires that an NLU system infer kinship relations between characters in short stories. Successful performance on this task requires both extracting relationships between entities, as well as inferring the logical rules governing these relationships. CLUTRR allows us to precisely measure a model's ability for systematic generalization by evaluating on held-out combinations of logical rules, and it allows us to evaluate a model's robustness by adding curated noise facts. Our empirical results highlight a substantial performance gap between state-of-the-art NLU models (e.g., BERT and MAC) and a graph neural network model that works directly with symbolic inputs-with the graph-based model exhibiting both stronger generalization and greater robustness. . 2016. A corpus and cloze evaluation for deeper understanding of commonsense stories. In Proceedings of NAACL-HLT, pages 839-849.Stephen Muggleton. 1991. Inductive logic programming. New Generation Computing, 8(4):295-318. . 2016. Ms marco: A human generated machine reading comprehension dataset. arXiv preprint arXiv:1611.09268. Zettlemoyer. 2018. Deep contextualized word representations. In NAACL.

show abstract

Section: Related Workmentioning

confidence: 99%

CLUTRR: A Diagnostic Benchmark for Inductive Reasoning from Text

Sinha¹,

Sodhani²,

Dong³

et al. 2019

Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conferen

View full text Add to dashboard Cite

show abstract

“…• The model not-pretrained is based on the approach of Kartsaklis et al (2018). They recently proposed a method to obtain single-sense and multi-sense vector embeddings during training (in contrast to our use of pre-trained embeddings for both).…”

Section: Baselinesmentioning

confidence: 99%

Using Multi-Sense Vector Embeddings for Reverse Dictionaries

Hedderich¹,

Yates

Klakow³

et al. 2019

Proceedings of the 13th International Conference on Computational Semantics - Long Papers

View full text Add to dashboard Cite

Popular word embedding methods such as word2vec and GloVe assign a single vector representation to each word, even if a word has multiple distinct meanings. Multi-sense embeddings instead provide different vectors for each sense of a word. However, they typically cannot serve as a drop-in replacement for conventional single-sense embeddings, because the correct sense vector needs to be selected for each word. In this work, we study the effect of multi-sense embeddings on the task of reverse dictionaries. We propose a technique to easily integrate them into an existing neural network architecture using an attention mechanism. Our experiments demonstrate that large improvements can be obtained when employing multi-sense embeddings both in the input sequence as well as for the target representation. An analysis of the sense distributions and of the learned attention is provided as well.

show abstract

“…Machine-learned vector space model (Speer and Chin, 2016) has used which combines word embedding that is produced by GloVe (Pennington et al, 2014) and word2vec (Mikolov et al, 2013) using tightly structured semantic networks like conceptNet (Speer and Havasi, 2012). Kartsaklis et al (Kartsaklis et al, 2018) have proposed a method which maps the natural language texts to the knowledge-based entities, they have enhanced LTSM model with a dynamic disambiguation mechanism on the input word embeddings that address polysemy issue. This method has gained state of the art performance in many word-similarity evaluations.…”

Section: Related Workmentioning

confidence: 99%

Çizge Tabanlı WordNet Ağı Üzerinde Anlamsal İlişki Ağırlıklarının Tespiti

Tulu

Orhan

Turan

2019

Gümüşhane Üniversitesi Fen Bilimleri Enstitüsü Dergisi

View full text Add to dashboard Cite

Determination of semantic relatedness between two textual items is one of the crucial phases in many Natural Language Processing applications. In this study, a new approach to lexicon based semantic relation determination methods was experienced using WordNet 3.0 and Men's real-life similarity dataset. Men's test collection was used for the determination of the relation weights and determined weights were used in semantic relatedness computation. RG65 similarity dataset was used for a benchmark of the proposed method and Spearman correlation 0.81 was gained, taking into account that retrieving the relations weight using a large scale dataset and testing them with another real-life dataset promises new perspectives to the determination of the relations weight and to the relatedness computation.

show abstract

Mapping Text to Knowledge Graph Entities using Multi-Sense LSTMs

Cited by 43 publications

References 32 publications

CLUTRR: A Diagnostic Benchmark for Inductive Reasoning from Text

CLUTRR: A Diagnostic Benchmark for Inductive Reasoning from Text

Using Multi-Sense Vector Embeddings for Reverse Dictionaries

Çizge Tabanlı WordNet Ağı Üzerinde Anlamsal İlişki Ağırlıklarının Tespiti

Contact Info

Product

Resources

About