Ontologies have been around in artificial intelligence (AI) research for the last 40 years. Just as trends come and go, ontologies too have had their ups and downs. Introduced in the 80s, ontologies became popular in the mid-90s. After machine learning (ML) came on the scene in 2000, the widespread opinion was that in the future every task performed with a computer (by means of AI and ML) could be solved with a smart algorithm. A lot of companies invested heavily in these algorithms hoping to have the next breakthrough in AI.
With the rapid development of AI and ML, especially after the emergence of the convolutional neural networks (CNN), the core technology – deep learning (DL) – has been growing rampantly in terms of parameter size and computing complexity. Today some of the most complex models have reached a scale of billions of parameters. 
Still, concerns have been raised regarding the current mainstream DL. Let us take supervised learning in image recognition as an example: the images used to train the AI models need to be manually identified in terms of the position and contour of the target objects, for the models to be able to find implicit pattern features between the data after comparing different labeling results. If we recall how we learn things as infants, us human beings can easily identify and classify different objects without needing this kind of instruction. 
DL methods have made tremendous progress and are now able to extract knowledge from the training data. However, this knowledge is not explicitly explainable, because the so-called “black box” training cannot reveal the complex relations hidden within the models. When facing new problems, the current DL models are unable to apply their acquired knowledge to solve new challenges in an effective way. 
There is another big concern regarding big data and the privacy issue related to it. What is more, current DL methods are based on big data which is not applicable to industries that generate small amounts of data such as certain fields in medicine and human resources. This case requires AI systems to have the ability to reason and judge, which can only be successful in specific domain areas. 
Many powerful ontologies already exist for specific domains, examples include the Financial Industry Business Ontology (FIBO) as well as numerous ontologies for healthcare, geography or occupations. It is widely believed that knowledge integration and DL are the important ideas for further amplifying the effectiveness of DL. For this reason, ontologies have made it back into the spotlight, along with many equivalents such as knowledge graphs or knowledge representations.
At JANZZ.technology, we started to build our ontology – JANZZon! – in 2008, before the tech giant Google invented and popularized the term “knowledge graph”. JANZZ.technology has been building its ontology using domain experts with various backgrounds (e.g. intellectual property law, fluid dynamics, car repair, open-heart surgery, or educational and vocational systems).
Today, JANZZon! is the largest multilingual encyclopedic knowledge representation in the field of occupation data. The main focus lies on jobs, job classifications, hard and soft skills, training/qualifications, etc. The number of stored nodes and relations comes to more than 350 million!
Integrated with both data-driven and expert consultation taxonomies, JANZZon! covers ESCO, O*Net, ISCO-08, GB/T 6556-2015, DISCO II and the UK skills taxonomy from Nesta just to name a few. Currently, 9 languages (German, English, French, Italian, Spanish, Portuguese, Dutch, Arabic and Norwegian) fully cover occupation, skills, specializations, function, education, etc., and we are working on achieving the same level in a total of 40 languages in 2020.
Being the backbone of our job and skills matching technology, JANZZon! represents the knowledge at the deepest level where all the entities have been encoded and vectorized in the semantic space. Therefore, when searching and matching, our technology can truly understand a concept and its semantic meaning and thus guarantee meaningful results.
If you are wondering how ontologies can help you empower your data in the fields of human resources and labor markets and aid you in realizing smart applications? Please write to email@example.com
 ODSC. 2018. Where Ontologies End and Knowledge Graphs Begin. URL: https://medium.com/predict/where-ontologies-end-and-knowledge-graphs-begin-6fe0cdede1ed [2019.11.20]
 Li Jun. 2019. Shen Du Xue Xi: Xin Shi Dai De Lian Jin Shu. URL : https://www.ftchinese.com/story/001084827?page=1&archive [2019.11.20]
 Cai Fangfang. 2019. Qin Hua Zi Ran Yu Yan Chu Li Ke Xue Jia Sun Maosong: Shen Du Xue Xi Peng Pi Zhi Hou, Wo Men Hai Neng Zuo Shen Me? URL: https://www.infoq.cn/article/OvhfhpPChTLpsMgrf43N [2019.11.20]