Commonsense mining as knowledge base completion? A study on the
            impact of novelty

Jastrzębski, Stanisław; Bahdanau, Dzmitry; Hosseini, S. Jafar; Noukhovitch, Michael; Bengio, Yoshua; Cheung, Jackie Chi Kit

doi:10.18653/v1/w18-1002

Cited by 19 publications

(24 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…While still in use today (Davis, 2017;Gordon and Hobbs, 2017), computational advances have allowed for more data-driven knowledge collection and representation (e.g., automatic extraction; Etzioni et al, 2008;Elazar et al, 2019). We will cover recent approaches that use natural language to represent commonsense (Speer et al, 2017;Sap et al, 2019a), and while noting the challenges that come with using datadriven methods (Gordon and Van Durme, 2013;Jastrzebski et al, 2018).…”

Section: Descriptionmentioning

confidence: 99%

Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics: Tutorial Abstracts

2020

View full text Add to dashboard Cite

While deep learning has transformed the natural language processing (NLP) field and impacted the larger computational linguistics community, the rise of neural networks is stained by their opaque nature: It is challenging to interpret the inner workings of neural network models, and explicate their behavior. Therefore, in the last few years, an increasingly large body of work has been devoted to the analysis and interpretation of neural network models in NLP.This body of work is so far lacking a common framework and methodology. Moreover, approaching the analysis of modern neural networks can be difficult for newcomers to the field. This tutorial aims to fill this gap and introduce the nascent field of interpretability and analysis of neural networks in NLP. This tutorial will cover cutting-edge research in interpretability and analysis of modern neural NLP models. The topic has not been previously covered in *CL tutorials.

show abstract

Section: Descriptionmentioning

confidence: 99%

Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics: Tutorial Abstracts

2020

View full text Add to dashboard Cite

show abstract

“…Numerous research efforts aim at extracting knowledge from text corpora but research on the exact purpose of commonsense knowledge (commonsense knowledge presents facts or individual features about the concept, such as 'Lemons are sour') which is machine-readable, is comparatively rare [33]. Automatically inferring missing facts from existing ones has thus become an increasingly important task.…”

Section: Related Workmentioning

confidence: 99%

Comprehensive structured knowledge base system construction with natural language presentation

Khanam

Liu

Chen

2019

Hum. Cent. Comput. Inf. Sci.

View full text Add to dashboard Cite

IntroductionMachine readable knowledge bases are used to store datasets so that these datasets can be accessible through systems. Machine-readable knowledge base construction (MRKBC) involves the automated extraction and integration of data from different sources and generating meaningful information with interoperable knowledge [1]. There is a large body of research on the automatic extraction of information for MRKBC. Initially, the research focused on syntactic information extraction [2, 3], but more recently, the extraction of lexical semantic information has received more interest from the research community [4,5]. Knowledge base systems (KBS) which use traditional databases are not effective due to the limited operational and analytical workload and AbstractConstructing an ontology-based machine-readable knowledge base system from different sources with minimum human intervention, also known as ontology-based machine-readable knowledge base construction (OMRKBC), has been a long-term outstanding problem. One of the issues is how to build a large-scale OMRKBC process with appropriate structural information. To address this issue, we propose Natural Language Independent Knowledge Representation (NLIKR), a method which regards each word as a concept which should be defined by its relations with other concepts. Using NLIKR, we propose a framework for the OMRKBC process to automatically develop a comprehensive ontology-based machine-readable knowledge base system (OMRKBS) using well-built structural information. Firstly, as part of this framework, we propose formulas to discover concepts and their relations in the OMRKBS. Secondly, the challenges in obtaining rich structured information are resolved through the development of algorithms and rules. Finally, rich structured information is built in the OMRKBS. OMRKBC allows the efficient search of words and supports word queries with a specific attribute. We conduct experiments and analyze the results of relational information extraction, with the results showing that OMRKBS had an accuracy of 84% which was higher than the other knowledge base systems, namely ConceptNet, DBpedia and WordNet.

show abstract

“…Several approaches have been proposed for training models for commonsense knowledge base completion (Li et al, 2016;Jastrzebski et al, 2018). Each of these approaches uses some sort of supervised training on a particular knowledge base, evaluating the model's performance on a held-out test set from the same database.…”

Section: Introductionmentioning

confidence: 99%

“…These works use relations from ConceptNet, a crowd-sourced database of structured commonsense knowledge, to train and validate their models (Liu and Singh, 2004). However, it has been shown that these methods generalize poorly to novel data (Li et al, 2016;Jastrzebski et al, 2018). Jastrzebski et al (2018) demonstrated that much of the data in the ConceptNet test set were simply rephrased relations from the training set, and that this train-test set leakage led to artificially inflated test performance metrics.…”

Section: Introductionmentioning

confidence: 99%

“…However, it has been shown that these methods generalize poorly to novel data (Li et al, 2016;Jastrzebski et al, 2018). Jastrzebski et al (2018) demonstrated that much of the data in the ConceptNet test set were simply rephrased relations from the training set, and that this train-test set leakage led to artificially inflated test performance metrics. This problem of traintest leakage is typical in knowledge base completion tasks (Toutanova et al, 2015;Dettmers et al, 2018).…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Commonsense Knowledge Mining from Pretrained Models

Davison¹,

Feldman

Rushton

2019

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

250

173

View full text Add to dashboard Cite

Inferring commonsense knowledge is a key challenge in natural language processing, but due to the sparsity of training data, previous work has shown that supervised methods for commonsense knowledge mining underperform when evaluated on novel data. In this work, we develop a method for generating commonsense knowledge using a large, pre-trained bidirectional language model. By transforming relational triples into masked sentences, we can use this model to rank a triple's validity by the estimated pointwise mutual information between the two entities. Since we do not update the weights of the bidirectional model, our approach is not biased by the coverage of any one commonsense knowledge base. Though this method performs worse on a test set than models explicitly trained on a corresponding training set, it outperforms these methods when mining commonsense knowledge from new sources, suggesting that unsupervised techniques may generalize better than current supervised approaches.

show abstract

Commonsense mining as knowledge base completion? A study on the impact of novelty

Cited by 19 publications

References 8 publications

Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics: Tutorial Abstracts

Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics: Tutorial Abstracts

Comprehensive structured knowledge base system construction with natural language presentation

Commonsense Knowledge Mining from Pretrained Models

Contact Info

Product

Resources

About