1 ALIN is an ontology matching system specialized in the interactive ontology matching, and its main characteristic is the use of expert feedback to improve the set of selected mappings, using semantic and structural techniques to make this improvement. This paper describes its con guration for the OAEI 2019 competition and discusses its results.
1.1
Alin handles three sets of mappings: (i) Accepted, which is a set of mappings de nitely to be retained in the alignment; (ii) Selected, which is a set of mappings where each is yet to be decided if it will be included in the alignment; and (iii) Suspended, which is a set of mappings that have been previously selected, but (temporarily or permanently) ltered out of the alignment.
Given the previous de nitions, Alin procedure follows 5 Steps, described as follows: 1. Select mappings: select the rst mappings and automatically accepts some of them. We explain the selection and acceptance process below; 2. Filter mappings: suspend some selected mappings, using lexical criteria for that; 3. Ask expert: accepts or rejects selected mappings, according to expert feedback 4. Propagate: select new mappings, reject some selected mappings or unsuspend some suspended mappings (depending on newly accepted mappings) 5. Go back to 3 as long as there are undecided selected mappings
All versions of ALIN (since its very rst OAEI participation) follow this general procedure. In this 2019 version, however, we introduced modi cations in Step 1. In previous versions, ALIN automatically accepted only the entities with the same name. In this version, ALIN also automatically accepts the entities whose names are synonyms or with variations in name words. ALIN searches synonyms in the Wordnet. In the Anatomy track, ALIN uses the FMA ontology too.
ALIN applies the following techniques:
{ Line 1. ALIN selects mappings using linguistic similarities between entity
names. ALIN uses synonyms and variations in entity name words to
automatically accept mappings. At this time, ALIN automatically selects and
accepts only concept mappings. To do that, ALIN uses linguistic metrics.
ALIN uses the Wordnet and domain-speci c ontologies (the FMA Ontology
in the Anatomy track) to nd synonyms between entities.
{ Line 2. ALIN suspends the selected mappings whose entities have low lexical
similarity. We use the Jaccard, Jaro-Wrinkler, and n-gram lexical metrics to
calculate the lexical similarity of the selected mappings. We based the process
of choosing the similarity metrics used by ALIN on the result of these metrics
in assessments [
An ontology may have construction constraints, such as a concept cannot be equivalent to its superconcept. An alignment may have other constraints like, for example, an entity of ontology O cannot be equivalent to two entities of the ontology O0. A mapping anti-pattern is a combination of mappings that generates a problematic alignment, i.e., a logical inconsistency or a violated constraint. 1.3
ALIN is available 2 as a package to be run through the SEALS client. 2
Results Interactive ontology matching is the focus of the ALIN system. Comparing its results in the 2019 campaign to its previous participations (Table 5), ALIN improvements include an expressive reduction on the number of interactions with the expert and the increase of the quality of the generated alignment. 2.1
ALIN used new techniques to automatically accept mappings. These techniques led to an increase in the F-Measure of non-interactively generated alignment, which shows the e ectiveness of the techniques. (Table 1 and Table 2). Conference track, unlike the Anatomy track, has relationship mappings and attribute mappings that ALIN does not automatically accept, thus making the F-Measure on the Conference track, although higher than last year, still low. 2 https : ==drive:google:com=f ile=d=1SxJ L6f LRV qI84epm8DbAM lcscEoGbgZ=view?usp = sharing
In the Anatomy track, ALIN was tied for second in quality (F-Measure) with slightly lower total requests (Table 3). In the Conference track, ALIN was tied for rst in quality with a slightly higher total request (Table 4). Interactive Anatomy Track In this track, ALIN has had a decrease in the number of expert interactions and an increase in the quality of the generated alignment, showing that the new techniques used to automatically accept correct mappings are e ective (Table 5).
ALIN used the FMA ontology to help nd synonyms between the two ontologies of the Anatomy track. The Foundational Model of Anatomy Ontology (FMA) is a reference ontology for the domain of Human anatomy 3. 3 \Foundational Model of Anatomy Ontology". Available at http://sig.biostr.washington.edu/projects/fm/AboutFM.html Last accessed on Oct, 11, 2019. 365 236 388 228 221 82 Interactive Conference Track In this track, ALIN has had a decrease in the number of expert interactions keeping a good quality of the generated alignment (Table 7). 2.3
In this version, ALIN uses new techniques to automatically accept mappings. These techniques use synonyms and word variations to nd equal entities between the two ontologies. ALIN also started to use FMA ontology as an external resource.
The use of the new techniques proved to be e ective as it reduced the number of interactions while keeping a good level of quality. The new techniques also increased the quality of the alignment generated in Anatomy interactive tracking, where ALIN used the FMA ontology.
It is not always possible to use an external resource to nd synonyms between entities of two ontologies, but when this is possible, the results showed that it is worth it.
The quality of the alignment generated by ALIN is dependent on the correct expert feedback, as expert responses are used to select new mappings. When ALIN selects wrong mappings, the quality of the generated alignment tends to decrease. But if we compare this year's quality decline with last year's, we see that this fall is less sharp (Table 6 and Table 8). The less sharp decline in quality is because we need less user interaction as we are automatically accepting more mappings.
The organization of FMA ontology in memory and the search for synonyms and word variations led to longer run time (Table 9 and Table 10) { handling user error rate; Evaluating the results, we can see that the system has improved, although it can improve even further, towards: { generating a higher quality initial alignment in its non-interactive phase; { reducing the number of interactions with the expert;
And there was a worsening run time, where we could improve too.
ALIN used new techniques to automatically accept new mappings. They have been e ective in reducing the number of interactions, while also keeping good quality in the generated alignment. In the case of the Anatomy track, these new techniques both decreased the number of interactions and increased the quality of the generated alignment. We can explain this quality improvement in this track by the use of the FMA ontology as a new external resource. With the use of the new techniques in both Anatomy and Conference tracks, there has been a less sharp drop in quality as the expert makes mistakes. Nevertheless, ALIN had an increase in run time due to the use of the new techniques, which may be addressed in future work.