Active knowledge graph completion Omran, Pouya Ghiasnezhad; Taylor, Kerry; Mendez, Sergio Rodriguez ...
Information sciences,
August 2022, 2022-08-00, Volume:
604
Journal Article
Peer reviewed
Open access
Enterprise and public Knowledge Graphs (KGs) are known to be incomplete. Methods for automatic completion, sometimes by rule learning, scale well. While previous rule-based methods learn closed ...(non-existential) rules, we introduce Open Path (OP) rules that are constrained existential rules. We present a novel algorithm, OPRL, for learning OP rules.
Closed rules complete a KG by answering queries of unclear origin, usually derived from a holdback test set in experimental settings. However, OP rules can generate relevant queries for KG completion. OPRL generates queries even when there is no closed rule to answer the query, or when the correct answer is a missing entity that is not present in the KG.
For OPRL to scale well, we propose a novel embedding-based fitness function to efficiently estimate rule quality. Additionally, we introduce a novel, efficient vector computation to formally assess rule quality.
We evaluate OPRL using adaptations of Freebase, YAGO2, Wikidata, and a synthetic Poker KG. We find that OPRL mines hundreds of accurate rules from massive KGs with up to 8 M facts. The OP rules generate queries with precision as high as 98% and recall of 62% on a complete KG, demonstrating the first solution for active knowledge graph completion.
Probabilistic Rule Learning Systems Salam, Abdus; Schwitter, Rolf; Orgun, Mehmet A.
ACM computing surveys,
07/2021, Volume:
54, Issue:
4
Journal Article
Peer reviewed
This survey provides an overview of rule learning systems that can learn the structure of probabilistic rules for uncertain domains. These systems are very useful in such domains because they can be ...trained with a small amount of positive and negative examples, use declarative representations of background knowledge, and combine efficient high-level reasoning with the probability theory. The output of these systems are probabilistic rules that are easy to understand by humans, since the conditions for consequences lead to predictions that become transparent and interpretable. This survey focuses on representational approaches and system architectures, and suggests future research directions.
Knowledge graph completion is an advanced artificial intelligence (AI) methodology that enables the systematic organization and structuring of data. It can significantly enhance the digital economy ...by facilitating more accurate and appropriate decision-making processes. Logical rule algorithms, known for their interpretability, have attracted significant attention in the field of explainable artificial intelligence. To harness the interpretability benefits of logical rules, we propose a cross-level position constraint template based on Graph Path Feature Learning (GPFL) and introduce an optimized termination policy for rule generation. To improve prediction performance, we focus significant emphasis on the rule evaluation stage, specifically on how to effectively learn the rules. To this end, we propose a Multi-Dimensional Graph Rule Learning (MDGRL) that calculates features from different dimensions to represent the reasoning ability of the rule. Feature I presents metrics for assessing the similarity into rule’s structure. Feature II proposes calculating the path transfer probability to represent the rule reasoning ability from the path reasoning perspective. Moreover, Feature III incorporates embedding distance with the constraint template to represent the rule reasoning ability in a low-dimensional vector space. Lastly, we assign weights to different dimensional features based on their performance and integrate them into MDGRL. Our experiments demonstrate effectiveness across various datasets, showcasing time-saving benefits in rule generation and improved prediction performance in rule evaluation. The source code and dataset for the MDGRL algorithm are available at https://github.com/csjywu1/MDGRL.
Origins of Hierarchical Logical Reasoning Dedhe, Abhishek M.; Clatterbuck, Hayley; Piantadosi, Steven T. ...
Cognitive science,
February 2023, 2023-02-00, 20230201, Volume:
47, Issue:
2
Journal Article
Peer reviewed
Open access
Hierarchical cognitive mechanisms underlie sophisticated behaviors, including language, music, mathematics, tool‐use, and theory of mind. The origins of hierarchical logical reasoning have long been, ...and continue to be, an important puzzle for cognitive science. Prior approaches to hierarchical logical reasoning have often failed to distinguish between observable hierarchical behavior and unobservable hierarchical cognitive mechanisms. Furthermore, past research has been largely methodologically restricted to passive recognition tasks as compared to active generation tasks that are stronger tests of hierarchical rules. We argue that it is necessary to implement learning studies in humans, non‐human species, and machines that are analyzed with formal models comparing the contribution of different cognitive mechanisms implicated in the generation of hierarchical behavior. These studies are critical to advance theories in the domains of recursion, rule‐learning, symbolic reasoning, and the potentially uniquely human cognitive origins of hierarchical logical reasoning.
The capacity to generate recursive sequences is a marker of rich, algorithmic cognition, and perhaps unique to humans. Yet, the precise processes driving recursive sequence generation remain ...mysterious. We investigated three potential cognitive mechanisms underlying recursive pattern processing: hierarchical reasoning, ordinal reasoning, and associative chaining. We developed a Bayesian mixture model to quantify the extent to which these three cognitive mechanisms contribute to adult humans’ performance in a sequence generation task. We further tested whether recursive rule discovery depends upon relational information, either perceptual or semantic. We found that the presence of relational information facilitates hierarchical reasoning and drives the generation of recursive sequences across novel depths of center embedding. In the absence of relational information, the use of ordinal reasoning predominates. Our results suggest that hierarchical reasoning is an important cognitive mechanism underlying recursive pattern processing and can be deployed across embedding depths and relational domains.
Compared to black-box neural networks, logic rules express explicit knowledge, can provide human-understandable explanations for reasoning processes, and have found their wide application in ...knowledge graphs and other downstream tasks. As extracting rules manually from large knowledge graphs is labour-intensive and often infeasible, automated rule learning has recently attracted significant interest, and a number of approaches to rule learning for knowledge graphs have been proposed. This survey aims to provide a review of approaches and a classification of state-of-the-art systems for learning first-order logic rules over knowledge graphs. A comparative analysis of various approaches to rule learning is conducted based on rule language biases, underlying methods, and evaluation metrics. The approaches we consider include inductive logic programming (ILP)-based, statistical path generalisation, and neuro-symbolic methods. Moreover, we highlight important and promising application scenarios of rule learning, such as rule-based knowledge graph completion, fact checking, and applications in other research areas.
A large literature has gauged the linguistic knowledge of signers by comparing sign-processing by signers and non-signers. Underlying this approach is the assumption that non-signers are devoid of ...any relevant linguistic knowledge, and as such, they present appropriate non-linguistic controls—a recent paper by Meade et al. (2022) articulates this view explicitly. Our commentary revisits this position. Informed by recent findings from adults and infants, we argue that the phonological system is partly amodal. We show that hearing infants use a shared brain network to extract phonological rules from speech and sign. Moreover, adult speakers who are sign-naïve demonstrably project knowledge of their spoken L1 to signs. So, when it comes to sign-language phonology, speakers are not linguistic blank slates. Disregarding this possibility could systematically underestimate the linguistic knowledge of signers and obscure the nature of the language faculty.
Action mining is a data mining method that aims to identify recommendations for changing attribute values that can lead to the classification of data instances as examples of another class. Action ...mining algorithms extract rules containing recommendations in the premises and class changes in the conclusion. To the best of the authors’ knowledge, no method has been proposed yet for generating action rules based on censored data. This study introduces the first method for survival action rule generation. The method stems from the covering rule induction algorithm but generates rules defining the actions required to change not the class but the survival curve of the covered examples. Thus, this study poses a new research problem: generating action rules for censored data and survival analysis. This study evaluated the proposed method using 22 data sets in which two application domains of survival analysis were distinguished: medicine and predictive maintenance. In addition, more detailed analyses of the generated survival action rules were presented in the form of case studies for the two selected data sets. The results show that the proposed method generates good-quality survival action rules and changes in the survival curves, resulting from the identified actions are significant.
What are the cognitive consequences of having a name for something? Having a word for a feature makes it easier to communicate about a set of exemplars belonging to the same category (e.g., “the red ...things”). But might it also make it easier to learn the category itself? Here, we provide evidence that the ease of learning category distinctions based on simple visual features is predicted from the ease of naming those features. Across seven experiments, participants learned categories composed of colors or shapes that were either easy or more difficult to name in English. Holding the category structure constant, when the underlying features of the category were easy to name, participants were faster and more accurate in learning the novel category. These results suggest that compact verbal labels may facilitate hypothesis formation during learning: it is easier to pose the hypothesis “it is about redness” than “it is about that pinkish-purplish color”. Our results have consequences for understanding how developmental and cross-linguistic differences in a language's vocabulary affect category learning and conceptual development.