Towards a Seamless Integration of Word Senses into Downstream NLP Applications

Pilehvar, Mohammad Taher; Camacho-Collados, José; Navigli, Roberto; Collier, Nigel

doi:10.18653/v1/p17-1170

Cited by 32 publications

(19 citation statements)

References 58 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The latter also employ an attention-based approach for creating vectors based on the context for predicting prepositional phrase attachments. Pilehvar et al (2017) use the same DeConf multi-sense embedding for integrating them in a downstream application. In contrast to our work, they require, however, a semantic network to do the disambiguation.…”

Section: Related Workmentioning

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

Section: Related Workmentioning

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

show abstract

“…NASARI has proved to be effective in various NLP tasks, including not only semantic similarity and WSD (Shalaby and Zadrozny 2015;Camacho-Collados et al 2016b;Tripodi and Pelillo 2017), but also sense clustering (see Sect. 5.2.2), knowledge-base construction and alignment (Lieto et al 2016;Espinosa-Anke et al 2016a;Camacho-Collados and Navigli 2017;Cocos et al 2017), object recognition (Young et al 2016) and text classification (Pilehvar et al 2017). …”

Section: Nasarimentioning

confidence: 99%

“…Knowledge resources such as Wikipedia or WordNet suffer from the high granularity of their sense inventories. A meaningful clustering of senses within these sense inventories could help boost the performance in different applications (Hovy et al 2013;Mancini et al 2017;Pilehvar et al 2017). In the following we explain how to deal with this issue in Wikipedia.…”

Section: Sense Clusteringmentioning

confidence: 99%

SenseDefs: a multilingual corpus of semantically annotated textual definitions

Camacho-Collados

Bovi²,

Raganato

et al. 2018

Lang Resources & Evaluation

Self Cite

View full text Add to dashboard Cite

Definitional knowledge has proved to be essential in various Natural Language Processing tasks and applications, especially when information at the level of word senses is exploited. However, the few sense-annotated corpora of textual definitions available to date are of limited size: this is mainly due to the expensive and time-consuming process of annotating a wide variety of word senses and entity mentions at a reasonably high scale. In this paper we present SENSEDEFS, a large-scale high-quality corpus of disambiguated definitions (or glosses) in multiple languages, comprising sense annotations of both concepts and named entities from a wide-coverage unified sense inventory. Our approach for the construction and disambiguation of this corpus builds upon the structure of a large multilingual semantic network and a state-of-the-art disambiguation system: first, we gather complementary information of equivalent definitions across different languages to provide context for disambiguation; then we refine the disambiguation output with a distributional approach based on semantic similarity. As a result, we obtain a multilingual corpus of textual definitions featuring over 38 million definitions in 263 languages, and we publicly release it to the research community. We assess the quality of SENSEDEFS's sense annotations both intrinsically and extrinsically on Open Information Extraction and Sense Clustering tasks.

show abstract

“…Moreover, we will present the field of knowledge-based representations, in particular word sense embeddings (Chen et al 2014;Rothe and Schuetze, 2015;Camacho-Collados et al 2016;Pilehvar and Collier, 2016;Mancini et al 2017), as flexible techniques which act as a bridge between lexical resources and applications. Finally, we will briefly present some recent work on the integration of this encoded knowledge from lexical resources into neural architectures for improving downstream NLP applications (Flekova and Gurevych, 2016;Pilehvar et al 2017). …”

Section: Descriptionmentioning

confidence: 99%

Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Tutorial Abstracts

Bansal

Rush

2018

View full text Add to dashboard Cite

ii IntroductionWelcome to the Tutorials Session of NAACL HLT 2018 in New Orleans.The NAACL-HLT tutorials session is an opportunity for conference attendees to participate in a tutorial on a timely topic of importance to the field, and to hear from experts on that topic. This year, the tutorials committee comprised of tutorials chairs from four conferences: ACL, COLING, EMNLP and NAACL-HLT. A total of 51 tutorial submissions were received, of which 6 were selected for presentation at NAACL-HLT.We hope you find the six NAACL-HLT tutorials for this year to combine depth within each tutorial, and breadth across a set of topics that demonstrate the increasing relevance of NLP to a wide range of fields within and beyond computer science.We would like to thank Marilyn Walker (NAACL general chair), Ying Lin (NAACL website chair), Stephanie Lukin and Margaret Mitchell (NAACL publications chairs), and Priscilla Rasmussen (local arrangement chair) for their help during the whole process. We also want to extend our sincere gratitude to the other conferences' tutorial chairs who jointly helped with reviewing for all the tutorial submissions: Yoav Artzi, Jacob Eisenstein, Pascale Fung, Donia Scott, Marilyn Walker, Mausam, and Lu Wang. DescriptionJust like natural language is a tool that humans use to communicate with each-other, programming languages are tools that humans use to communicate with computers. Because of the increasing need for programs and programming in our working and everyday lives, there are now massive amounts of source code being produced every day. As a result, it is ever more important for an ever increasing segment of the populace to be able to understand and create programs to do what they would like to do. However, programming is a specialized skill, IT education is hard-pressed to make up for this demand.One key insight that can help us tackle this problem is that source code is bimodal . While one modality is targeted towards explicitly instructing the hardware on the actions to perform, the other is targeted towards the humans that need to read, understand, maintain and extend the code. Given that it is humans that are producing the software, the human-oriented modality is very strong and often takes the form of natural language: from natural language identifiers, such as variable and method names, to code comments and natural language documentation.As a result, there is recently a burgeoning interest in research that connects natural language with the programming language artifacts. This research area has the potential to improve the efficiency and ease of programming by making connections to natural language, which is (in general) easier for humans to understand and communicate with, particularly humans who are not yet well-versed in programming. Some examples of relevant tasks include:• Automatic explanation of programs in natural language (code-to-language): Highly connected with the task of grounded natural language generation in the NLP community, this is the task of generating natural lang...

show abstract

Towards a Seamless Integration of Word Senses into Downstream NLP Applications

Cited by 32 publications

References 58 publications

Using Multi-Sense Vector Embeddings for Reverse Dictionaries

Using Multi-Sense Vector Embeddings for Reverse Dictionaries

SenseDefs: a multilingual corpus of semantically annotated textual definitions

Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Tutorial Abstracts

Contact Info

Product

Resources

About