TransOMCS: From Linguistic Graphs to Commonsense Knowledge
By Hongming Zhang, Daniel Khashabi, Yangqiu Song, Dan Roth. 2020
99 citations recorded by the source corpus.
Commonsense knowledge acquisition is a key problem for artificial intelligence. Conventional methods of acquiring commonsense knowledge generally require laborious and costly human annotations, which are not feasible on a large scale. In this paper, we explore a practical way of mining commonsense knowledge from linguistic graphs, with the goal of transferring cheap knowledge obtained with linguistic patterns into expensive commonsense knowledge. The result is a conversion of ASER [Zhang et al., 2020], a large-scale selectional preference knowledge resource, into TransOMCS, of the same representation as ConceptNet [Liu and Singh, 2004] but two orders of magnitude larger. Experimental results demonstrate the transferability of linguistic knowledge to commonsense knowledge and the effectiveness of the proposed approach in terms of quantity, novelty, and quality. TransOMCS is publicly available at: https://github.com/HKUST-KnowComp/TransOMCS.
Subfields: Artificial Intelligence, Natural Language Processing
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
- Yejin Choi
- Jianfeng Gao
- Zhe Gan
- Mohit Bansal
- Chunyuan Li
- Lijuan Wang
- Hannaneh Hajishirzi
- Jianwei Yang
In Natural Language Processing
- Yejin Choi
- Ronan Le Bras
- Mohit Bansal
- Jianfeng Gao
- Chandra Bhagavatula
- Hannaneh Hajishirzi
- Sean Welleck
- Ximing Lu
Citations in the corpus
- Symbolic Knowledge Distillation: from General Language Models to Commonsense Models (2021, methodology): “Among automatically constructed CKGs are TransOMCS (Zhang et al., 2020a) and CausalBank (Li et al.”
- COMET-ATOMIC 2020: On Symbolic and Neural Commonsense Knowledge Graphs (2020, methodology): “The TRANSOMCS (Zhang et al. 2020a) knowledge graph consists of 18.48M tuples that were automatically converted from syntactic parses of sentences from various web sources including Wikipedia, Yelp, and Reddit.”
- K-LITE: Learning Transferable Visual Models with External Knowledge (2022, background): “Knowledge-Intensive Models: In natural language processing (NLP), with the increase of model capacity via pre-trained language models [17], there emerges the need for more knowledgeable models [55] with advanced functionalities such as making use of encyclopedic [88, 4, 6] and commonsense knowledge ”
- Unifying Large Language Models and Knowledge Graphs: A Roadmap (2023, methodology): “Some other commonsense knowledge graphs, such as TransOMCS [75] and CausalBanK [76] are automatically constructed to provide commonsense knowledge.”