Sensorimotor distance: A fully grounded measure of semantic similarity for 800 million concept pairs

Experimental design and computational modelling across the cognitive sciences often rely on measures of semantic similarity between concepts. Traditional measures of semantic similarity are typically derived from distance in taxonomic databases (e.g. WordNet), databases of participant-produced semantic features, or corpus-derived linguistic distributional similarity (e.g. LSA), all of which are theoretically problematic in their lack of grounding in sensorimotor experience. We present a new measure of sensorimotor distance between concepts, based on multidimensional comparisons of their experiential strength across 11 perceptual and action-effector dimensions in the Lancaster Sensorimotor Norms. We demonstrate that, in modelling human similarity judgements, sensorimotor distance has comparable explanatory power to other measures of semantic similarity, explains variance in human judgements which is missed by other measures, and does so with the advantages of remaining both fully grounded and computationally efficient. Moreover, sensorimotor distance is equally effective for both concrete and abstract concepts. We further introduce a web-based tool (https://lancaster.ac.uk/psychology/smdistance) for easily calculating and visualising sensorimotor distance between words, featuring coverage of nearly 800 million word-pairs. Supplementary materials are available at https://osf.io/d42q6/.

Detecting missing IS-A relations in the NCI Thesaurus using an enhanced hybrid approach

Background The National Cancer Institute (NCI) Thesaurus provides reference terminology for NCI and other systems. Previously, we proposed a hybrid prototype utilizing lexical features and role definitions of concepts in non-lattice subgraphs to identify missing IS-A relations in the NCI Thesaurus. However, no domain expert evaluation was provided in our previous work. In this paper, we further enhance the hybrid approach by leveraging a novel lexical feature—roots of noun chunks within concept names. Formal evaluation of our enhanced approach is also performed. Method We first compute all the non-lattice subgraphs in the NCI Thesaurus. We model each concept using its role definitions, words and roots of noun chunks within its concept name and its ancestor’s names. Then we perform subsumption testing for candidate concept pairs in the non-lattice subgraphs to automatically detect potentially missing IS-A relations. Domain experts evaluated the validity of these relations. Results We applied our approach to 19.08d version of the NCI Thesaurus. A total of 55 potentially missing IS-A relations were identified by our approach and reviewed by domain experts. 29 out of 55 were confirmed as valid by domain experts and have been incorporated in the newer versions of the NCI Thesaurus. 7 out of 55 further revealed incorrect existing IS-A relations in the NCI Thesaurus. Conclusions The results showed that our hybrid approach by leveraging lexical features and role definitions is effective in identifying potentially missing IS-A relations in the NCI Thesaurus.

Sensor Ontology Metamatching with Heterogeneity Measures

The heterogeneity problem among different sensor ontologies hinders the interaction of information. Ontology matching is an effective method to address this problem by determining the heterogeneous concept pairs. In the matching process, the similarity measure serves as the kernel technique, which calculates the similarity value of two concepts. Since none of the similarity measures can ensure its effectiveness in any context, usually, several measures are combined together to enhance the result’s confidence. How to find suitable aggregating weights for various similarity measures, i.e., ontology metamatching problem, is an open challenge. This paper proposes a novel ontology metamatching approach to improve the sensor ontology alignment’s quality, which utilizes the heterogeneity features on two ontologies to tune the aggregating weight set. In particular, three ontology heterogeneity measures are firstly proposed to, respectively, evaluate the heterogeneity values in terms of syntax, linguistics, and structure, and then, a semiautomatically learning approach is presented to construct the conversion functions that map any two ontologies’ heterogeneity values to the weights for aggregating the similarity measures. To the best of our knowledge, this is the first time that heterogeneity features are proposed and used to solve the sensor ontology metamatching problem. The effectiveness of the proposal is verified by comparing with using state-of-the-art ontology matching techniques on Ontology Alignment Evaluation Initiative (OAEI)’s testing cases and two pairs of real sensor ontologies.

A multi-strategy approach for the merging of multiple taxonomies

Taxonomy merging is an important work to provide a uniform schema for several heterogeneous taxonomies. Previous studies primarily focus on merging two taxonomies in a specific domain, while the merging of multiple taxonomies has been neglected. This article proposes a taxonomy merging approach to automatically merge multiple source taxonomies into a target taxonomy in an asymmetric manner. The approach adopts a strategy of breaking up the whole into parts to decrease the complexity of merging multiple taxonomies and employs a block-based method to reduce the scale of measuring semantic relations between concept pairs. In addition, for the problem of multiple inheritance, a method of topical coverage is proposed. Experiments conducted on synthetic and real-world scenarios indicate that the proposed merging approach is feasible and effective to merge multiple taxonomies. In particular, the proposed approach works well in the aspects of limiting the semantic redundancy and establishing high-quality hierarchical relations between concepts.

Extraction of cause-effect-concept pair series from web documents

<span>This research aims to extract a cause-effect-concept pair series of consequent event occurrences in health information of hospital web-boards. The extracted cause-effect-concept pair series representing a disease causation pathway benefits for the automatic diagnosis and solving system. Where each causative/effect event concept is expressed by an elementary discourse unit (EDU which is a simple sentence). The research has three problems; how to determine causative/effect concept EDUs from the documents containing some EDU occurrences with both causative concepts and effect concepts, how to determine the cause-effect relation between two adjacent EDUs having the discourse cue ambiguity, and how to extract cause-effect-concept pair series mingled with either a stimulation relation EDU or other non-cause-effect relation EDUs from the documents. Therefore, we apply annotated NWordCo pairs with causative-effect concepts to represent EDU pairs with causative-effect concept where the NWordCo size solved by Naïve Bayes. We also apply Naïve Bayes to solve NWordCo-concept pairs having the cause-effect relation from the adjacent EDU pairs. We then propose using cue words and the collected NWordCo-concept pairs with the cause-effect relation to extract the cause-effect-concept pair series. The research results provide the high precision of the cause-effect-concept pair series determination from the documents. </span>

Knowledge Aware Semantic Concept Expansion for Image-Text Matching

Image-text matching is a vital cross-modality task in artificial intelligence and has attracted increasing attention in recent years. Existing works have shown that learning semantic concepts is useful to enhance image representation and can significantly improve the performance of both image-to-text and text-to-image retrieval. However, existing models simply detect semantic concepts from a given image, which are less likely to deal with long-tail and occlusion concepts. Frequently co-occurred concepts in the same scene, e.g. bedroom and bed, can provide common-sense knowledge to discover other semantic-related concepts. In this paper, we develop a Scene Concept Graph (SCG) by aggregating image scene graphs and extracting frequently co-occurred concept pairs as scene common-sense knowledge. Moreover, we propose a novel model to incorporate this knowledge to improve image-text matching. Specifically, semantic concepts are detected from images and then expanded by the SCG. After learning to select relevant contextual concepts, we fuse their representations with the image embedding feature to feed into the matching module. Extensive experiments are conducted on Flickr30K and MSCOCO datasets, and prove that our model achieves state-of-the-art results due to the effectiveness of incorporating the external SCG.

Towards Semantic Sensor Data: An Ontology Approach

In order to optimize intelligent applications driven by various sensors, it is vital to properly interpret and reuse sensor data from different domains. The construction of semantic maps which illustrate the relationship between heterogeneous domain ontologies plays an important role in knowledge reuse. However, most mapping methods in the literature use the literal meaning of each concept and instance in the ontology to obtain semantic similarity. This is especially the case for domain ontologies which are built for applications with sensor data. At the instance level, there is seldom work to utilize data of the sensor instances when constructing the ontologies’ mapping relationship. To alleviate this problem, in this paper, we propose a novel mechanism to achieve the association between sensor data and domain ontology. In our approach, we first classify the sensor data by making them as SSN (Semantic Sensor Network) ontology instances, and map the corresponding instances to the concepts in the domain ontology. Secondly, a multi-strategy similarity calculation method is used to evaluate the similarity of the concept pairs between the domain ontologies at multiple levels. Finally, the set of concept pairs with a high similarity is selected by the analytic hierarchy process to construct the mapping relationship between the domain ontologies, and then the correlation between sensor data and domain ontologies are constructed. Using the method presented in this paper, we perform sensor data correlation experiments with a simulator for a real world scenario. By comparison to other methods, the experimental results confirm the effectiveness of the proposed approach.

Predicting Context-Dependent Cross-Modal Associations with Dimension-Specific Polarity Attributions Part 1 – Brightness and Aggression

Although researchers have repeatedly shown that the meaning of the same concept can vary across different contexts, it has proven difficult to predict when people will assign which meaning to a concept, and which associations will be activated by a concept. Building on the affective theory of meaning (Osgood, Suci, & Tannenbaum, 1957) and the polarity correspondence principle (Proctor & Cho, 2006), we propose the dimension-specificity hypothesis with the aim to understand and predict the context-dependency of cross-modal associations. We present three sets of experiments in which we use the dimension-specificity hypothesis to predict the cross-modal associations that should emerge between aggression-related concepts and saturation and brightness. The dimension-specificity hypothesis predicts that cross-modal associations emerge depending upon which affective dimension of meaning (i.e., the evaluation, activity, or potency dimension) is most salient in a specific context. The salience of dimensions of meaning is assumed to depend upon the relative conceptual distances between bipolar opposed concept pairs (e.g., good vs. bad). The dimension-specificity hypothesis proposes that plus and minus polarities will be attributed to the bipolar concepts, and associations between concrete and affective abstract concepts that share plus or minus polarities will become activated. Our data support the emergence of dimension-specific polarity attributions. We discuss the potential of dimension-specific polarity attributions to understand and predict how the context influences the emergence of cross-modal associations.