Filters








25,430 Hits in 2.8 sec

Graph Embedding Based on Nodes Attributes Representatives and a Graph of Words Representation [chapter]

Jaume Gibert, Ernest Valveny
2010 Lecture Notes in Computer Science  
of the graphs to be processed.  ...  Although graph embedding has recently been used to extend statistical pattern recognition techniques to the graph domain, some existing embeddings are usually computationally expensive as they rely on  ...  We had not faced this problem as the vocabularies we have selected for our databases of graphs were not that large.  ... 
doi:10.1007/978-3-642-14980-1_21 fatcat:63ksqkcwl5f3rpf35neyqp6dmi

Automatic Text Summarization with a Reduced Vocabulary Using Continuous Space Vectors [chapter]

Elvys Linhares Pontes, Stéphane Huet, Juan-Manuel Torres-Moreno, Andréa Carneiro Linhares
2016 Lecture Notes in Computer Science  
In this paper, we propose a new method that uses continuous vectors to map words to a reduced vocabulary, in the context of Automatic Text Summarization (ATS).  ...  Our experiments show that the reduced vocabulary improves the performance of state-of-the-art systems.  ...  Consequently, a higher number of words of the text vocabulary are not in the dictionary of French word embeddings. Table 1 .  ... 
doi:10.1007/978-3-319-41754-7_46 fatcat:mugm7zsuknd5vcupyynbgbac6e

ConceptNet at SemEval-2017 Task 2: Extending Word Embeddings with Multilingual Relational Knowledge

Robyn Speer, Joanna Lowry-Duda
2017 Proceedings of the 11th International Workshop on Semantic Evaluation (SemEval-2017)  
ConceptNet is an open, multilingual knowledge graph that focuses on general knowledge that relates the meanings of words and phrases.  ...  Our submission to SemEval was an update of previous work that builds high-quality, multilingual word embeddings from a combination of ConceptNet and distributional semantics.  ...  Vocabulary Selection Expanded retrofitting produces vectors for all the terms in its knowledge graph and all the terms in the input embeddings.  ... 
doi:10.18653/v1/s17-2008 dblp:conf/semeval/SpeerL17 fatcat:fpbx7w4nmjbipbnuafsim5jpiq

Navigating with Graph Representations for Fast and Scalable Decoding of Neural Language Models [article]

Minjia Zhang, Xiaodong Liu, Wenhan Wang, Jianfeng Gao, Yuxiong He
2018 arXiv   pre-print
This paper proposes a novel softmax layer approximation algorithm, called Fast Graph Decoder (FGD), which quickly identifies, for a given context, a set of K words that are most likely to occur according  ...  However, NLMs are very computationally demanding largely due to the computational cost of the softmax layer over a large vocabulary.  ...  (b) shows the transformation from the word embedding vectors of the NLM vocabulary, x 1 , x 2 , . 3 : 3 Hyperparameter: Small world graph neighbor degree M. 4: for all i in (0..  ... 
arXiv:1806.04189v1 fatcat:syrup2twcvf7dl32hduoafpub4

VGCN-BERT: Augmenting BERT with Graph Embedding for Text Classification [chapter]

Zhibin Lu, Pan Du, Jian-Yun Nie
2020 Lecture Notes in Computer Science  
However, their ability of capturing the global information about the vocabulary of a language is more limited. This latter is the strength of Graph Convolutional Networks (GCN).  ...  In this paper, we propose VGCN-BERT model which combines the capability of BERT with a Vocabulary Graph Convolutional Network (VGCN).  ...  Instead of using only word embeddings of the input sentence in BERT, we feed both the vocabulary graph embedding obtained by Eq. 5 and the sequence of word embeddings to BERT transformer.  ... 
doi:10.1007/978-3-030-45439-5_25 fatcat:6kwtm2vov5dpzbfjfpspri2h5i

VGCN-BERT: Augmenting BERT with Graph Embedding for Text Classification [article]

Zhibin Lu, Pan Du, Jian-Yun Nie
2020 arXiv   pre-print
However, their ability of capturing the global information about the vocabulary of a language is more limited. This latter is the strength of Graph Convolutional Networks (GCN).  ...  In this paper, we propose VGCN-BERT model which combines the capability of BERT with a Vocabulary Graph Convolutional Network (VGCN).  ...  Instead of using only word embeddings of the input sentence in BERT, we feed both the vocabulary graph embedding obtained by Equation 5 and the sequence of word embeddings to BERT transformer.  ... 
arXiv:2004.05707v1 fatcat:s3jtioywffcj3itewjhn2rncw4

Synonym Discovery with Etymology-based Word Embeddings [article]

Seunghyun Yoon, Pablo Estrada, Kyomin Jung
2017 arXiv   pre-print
We test our model in the Chinese and Sino-Korean vocabularies. Our graphs are formed by a set of 117,000 Chinese words, and a set of 135,000 Sino-Korean words.  ...  Our model derives word embedding vectors using the etymological composition of words, rather than the context in which they appear.  ...  Verifying the word embeddings: Synonym discovery To verify the validity of the embeddings, we selected the task of synonym discovery.  ... 
arXiv:1709.10445v2 fatcat:csqiasmqljhhtmyngyoparwlpu

Be Concise and Precise: Synthesizing Open-Domain Entity Descriptions from Facts

Rajarshi Bhowmik, Gerard de Melo
2019 The World Wide Web Conference on - WWW '19  
Despite being vast repositories of factual information, cross-domain knowledge graphs, such as Wikidata and the Google Knowledge Graph, only sparsely provide short synoptic descriptions for entities.  ...  Given the rapidly increasing numbers of entities in knowledge graphs, a fully automated synthesis of succinct textual descriptions from underlying factual information is essential.  ...  At each time step t, the decoder input is a concatenation of three vectors: the embedding f t of a selected fact in the current time step, the embedding w t −1 of the vocabulary word at the previous time  ... 
doi:10.1145/3308558.3313656 dblp:conf/www/BhowmikM19 fatcat:pgtscgca2rhonnr3kxwtu2yzu4

Vocabulary Selection Strategies for Neural Machine Translation [article]

Gurvan L'Hostis, David Grangier, Michael Auli
2016 arXiv   pre-print
In this paper we experiment with context and embedding-based selection methods and extend previous work by examining speed and accuracy trade-offs in more detail.  ...  Classical translation models constrain the space of possible outputs by selecting a subset of translation rules based on the input sentence.  ...  The two graphs show different trends: On the left, coverage with respect to the reference for the full vocabulary is 95.1%, while selection achieves 87.5% with a vocabulary of 614 words (3rd point on graph  ... 
arXiv:1610.00072v1 fatcat:53jgt76l45f6dhw3sctrmtupbm

Text2Node: a Cross-Domain System for Mapping Arbitrary Phrases to a Taxonomy [article]

Rohollah Soltani, Alexandre Tomberg
2019 arXiv   pre-print
We compared multiple algorithms and architectures for each stage of the training, including GloVe and FastText word embeddings, CNN and Bi-LSTM mapping functions, and node2vec for node embeddings.  ...  Text2Node operates in three main stages: first, the lexicon is mapped to word embeddings; second, the taxonomy is vectorized using node embeddings; and finally, the mapping function is trained to connect  ...  An important distinction between them is the treatment of words that are not part of the training vocabulary: GloVe creates a special out-of-vocabulary token and maps all of these words to this token's  ... 
arXiv:1905.01958v1 fatcat:chrv7er4tbe53mk7tdikt4frqe

Graph of Words Embedding for Molecular Structure-Activity Relationship Analysis [chapter]

Jaume Gibert, Ernest Valveny, Horst Bunke
2010 Lecture Notes in Computer Science  
In this article we make use of a particular graph representation of molecules and propose a new graph embedding procedure to solve the problem of structure-activity relationship analysis.  ...  The embedding is essentially an arrangement of a molecule in the form of a vector by considering frequencies of appearing atoms and frequencies of covalent bonds between them.  ...  the graph of words embedding.  ... 
doi:10.1007/978-3-642-16687-7_9 fatcat:mwsh5zy6snc6xa6rjpcjqmdu7q

Seq2RDF: An end-to-end application for deriving Triples from Natural Language Text [article]

Yue Liu, Tongtao Zhang, Zhicheng Liang, Heng Ji, Deborah L. McGuinness
2018 arXiv   pre-print
Our model learns the mapping from natural language text to triple representation in the form of subject-predicate-object using the selected knowledge graph vocabulary.  ...  mechanism that leverages knowledge graph embeddings.  ...  based on the meanings in the knowledge graph (KG) vocabulary.  ... 
arXiv:1807.01763v3 fatcat:4gbz7mdskvazzjkd5ybtve6jje

A Neural Knowledge Language Model [article]

Sungjin Ahn, Heeyoul Choi, Tanel Pärnamaa, Yoshua Bengio
2017 arXiv   pre-print
By predicting whether the word to generate has an underlying fact or not, the model can generate such knowledge-related words by copying from the description of the predicted fact.  ...  This is mainly because they acquire such knowledge from statistical co-occurrences although most of the knowledge words are rarely observed.  ...  as the developers of Theano (Bastien et al., 2012) , NSERC, CIFAR, Facebook, Google, IBM, Microsoft, Samsung, and Canada Research Chairs for funding, and Compute Canada for computing resources.  ... 
arXiv:1608.00318v2 fatcat:hxug555mmfhxleu77uvmul75om

Integrating Approaches to Word Representation [article]

Yuval Pinter
2021 arXiv   pre-print
level and the out-of-vocabulary phenomenon.  ...  I present a survey of the distributional, compositional, and relational approaches to addressing this task, and discuss various means of integrating them into systems, with special emphasis on the word  ...  Acknowledgments This survey is an adapted version of the introduction my PhD thesis.  ... 
arXiv:2109.04876v1 fatcat:hyzt3j7ibrhe3dg72imc27w4zy

Poincaré GloVe: Hyperbolic Word Embeddings [article]

Alexandru Tifrea, Gary Bécigneul, Octavian-Eugen Ganea
2018 arXiv   pre-print
Words are not created equal. In fact, they form an aristocratic graph with a latent hierarchical structure that the next generation of unsupervised learned word embeddings should reveal.  ...  This connection allows us to introduce a novel principled hypernymy score for word embeddings.  ...  Model: h = (·) 2 , full vocabulary of 190k words. More of these plots for other D 2 spaces are shown in appendix A.3.  ... 
arXiv:1810.06546v2 fatcat:cjexwnqrazdi7abxaearldpfli
« Previous Showing results 1 — 15 out of 25,430 results