In the data-driven work ows in the materials science domain, much of the data and knowledge is stored in di erent heterogeneous data sources maintained by di erent groups. This leads to a reduced availability of the data and poor interoperability between systems in this domain. Ontology-based techniques are an important way to reduce these problems and a number of e orts have started. In this paper, we use a phrase-based topic model approach and formal topical concept analysis on unstructured text in this domain to suggest additional concepts and axioms for the ontology that should be validated by a domain expert.
More and more researchers in materials science have realized that data-driven
techniques could accelerate the discovery and design of materials. Therefore, a
large number of research groups and communities have developed data-driven
work ows including data repositories (for an overview see [
However, developing ontologies is not an easy task and often the resulting ontologies are not complete. In addition to being problematic for the correct modelling of a domain, such incomplete ontologies also in uence the quality of semantically-enabled applications such as ontology-based search and data integration. Incomplete ontologies when used in semantically-enabled applications can lead to valid conclusions being missed. For instance, in ontology-based search, queries are re ned and expanded by moving up and down the hierarchy of concepts. Incomplete structure in ontologies in uences the quality of the search results. ?? Copyright c 2019 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
We propose a novel method for extending existing ontologies by detecting new concepts that should be included in the ontologies. We do this by presenting an approach, formal topical concept analysis, that integrates a variant of topic modeling and formal concept analysis. 2
Our approach for extending ontologies, shown in Fig. 1, contains the following
steps. In the rst step creation of a phrase-based topic model documents related
to the domain of interest are used to create topics (upper part in Fig. 1). We
use the phrases-based topic model in the ToPMine system [
As shown in Fig. 1, a domain expert is involved in the di erent steps in our approach to validate and interpret the results of the phrase-based topic model and the formal topical concept analysis in terms of interpreting all phrases appearing in all topics, interpreting topics using the representative phrases or a subset of representative phrases in a topic. The outcome can be divided into existing knowledge (EXIST(-m)), and new knowledge ADD(-m), where the `-m' quali er denotes that a modi ed version of the label exists or should be added (e.g., core-shell nanoparticle for core shell ). Further, the label could refer to a too general concept for the ontology (No-g) or not represent relevant knowledge (No). For the topic interpretation there is an additional outcome (Q) referring to too speci c concepts for the ontology, but such concepts could be de ned using concepts in the ontologies and OWL constructs. Finally, the domain expert interprets the formal concept analysis-based lattice and may nd new concepts and subsumption axioms. 3
Data and Experiments. The corpus that we use is based on reports on
nanoparticles from the Nanoparticle Information Library (http://nanoparticlelibrary.
net). For each nanoparticle report, we take the text in `Research Abstract'
as well as the abstracts (or only the titles if there is no abstract) from the
publications in `Related Publications'. The nal corpus contains 627 abstracts
(or titles). The ontologies that we extend are the Nanoparticle ontology [
Results and discussion of results. In Table 1 we show the results regarding the interpretation of phrases, topics and lattice nodes from the experiments.
We show the number of general concepts (No-g), existing concepts (EXIST(m)), new concepts (ADD(-m)), new axioms (ADD(-m)), queries (Q) and in uenced concepts by adding the new axioms. Our results showed that the approach generated many EXIST(-m) cases. This provides a sanity check for our approach as it shows that existing concepts can be found. Further, the approach found 35 and 32 new concepts for the NanoParticle ontology and the eNanoMapper ontology respectively, as well as 42 and 37 new axioms. In addition to the new concepts and new axioms, also other concepts are in uenced. Indeed, for a new axiom A is-a B, the sub-concepts of A receive B and all its super-concepts as its superconcepts (and thus inherit their properties), and all super-concepts of B receive A and its sub-concepts as sub-concepts (and thus all instances of these concepts are also instances of B and its super-concepts). In this experiment, 72 concepts from NanoParticle ontology were in uenced by the new axioms. Therefore, the quality of NanoParticle ontology-enabled applications is improved whenever one of the 35 new or 72 in uenced concepts is used. For the eNanoMapper ontology the number of in uenced existing concepts by adding new axioms is 37.
For the experiments we have currently used few resources, i.e. circa 600 abstracts and less than 10 hours for each of three experts (a domain expert and two knowledge engineering experts). As the quality of the ontologies and their use is raised, it is clear that the e ort for extending the ontologies was worth-while. 4
In this paper we have used a phrase-based topic model approach and our own variant of formal concept analysis for extending ontologies. A domain expert interprets the results which are phrases, topics and a lattice. This leads to the con rmation of ontological concepts (EXIST(-m)) or to the addition of new concepts and axioms (ADD(-m)). The latter is the actual extension of the ontologies. Also, concepts from more general or other domains may be found, as well as very speci c concepts in the domain that need not be added to the ontology. We have shown the usefulness of the approach by extending two ontologies in the nanotechnology domain using approximately 600 abstracts.
One issue that the domain expert noted was that it was not always easy to decide which level of granularity to use during the interpretation. In the future we will investigate how to help the domain expert dealing with this issue. In particular, the lattice appears to help re ning topics into concepts that are more general and meaningful in the domain. This may be a useful step forward towards a higher level of automation in the process of extracting ontology information out of unstructured text. Furthermore, we will investigate the scalability of our approach by experimenting with more documents. Another possible direction is to investigate synergy possibilities between the topics and the ontology concepts, e.g., by using the ontologies to generate the corpora, or by iterating between topic generation and interpretation.