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

**Authors:** Robyn Speer, Joanna Lowry-Duda  
**arXiv:** [1704.03560](https://arxiv.org/abs/1704.03560)  
**Date:** 2017-04-11  
**Summary page:** https://www.summarizepaper.com/en/paper/1704.03560v2/

## Key points

- Luminoso's system based on ConceptNet achieved first place in both subtasks of SemEval 2017 Task 2
- Their system utilized ConceptNet and distributional semantics to generate high-quality, multilingual word embeddings
- Luminoso's system ranked first in four out of five separate languages in language-specific evaluation
- It secured the top position in all ten cross-lingual language pairs
- The approach effectively captures semantic word similarity across different languages
- Authors Robyn Speer and Joanna Lowry-Duda present this research as a contribution to SemEval-2017 Proceedings
- The paper provides valuable insights into extending word embeddings using multilingual relational knowledge within the context of ConceptNet
- The approach showcases potential for future applications in natural language processing tasks.

## Summary

This paper, titled "ConceptNet at SemEval-2017 Task 2: Extending Word Embeddings with Multilingual Relational Knowledge," describes the participation of Luminoso in SemEval 2017 Task 2. Luminoso's system was based on ConceptNet, an open and multilingual knowledge graph that connects the meanings of words and phrases through general knowledge. The authors built upon their previous work and submitted an updated version to SemEval. Their system utilized a combination of ConceptNet and distributional semantics to generate high-quality, multilingual word embeddings. Impressively, their system achieved first place in both subtasks of the competition. In terms of language-specific evaluation, Luminoso's system ranked first in four out of five separate languages. Additionally, it secured the top position in all ten cross-lingual language pairs. This outstanding performance highlights the effectiveness of their approach in capturing semantic word similarity across different languages. The authors, Robyn Speer and Joanna Lowry-Duda, present this research as a contribution to the Proceedings of the 11th International Workshop on Semantic Evaluation (SemEval-2017). The paper provides valuable insights into extending word embeddings using multilingual relational knowledge within the context of ConceptNet. It demonstrates how effective such an approach can be for capturing semantic similarities between words across different languages and showcases its potential for future applications in natural language processing tasks.