I am looking for a method that can realize if there is any semantic relation between two terms using ontologies. For example given two terms "kitchen" and "Chef" it can return s.th like chef works in a kitchen and a correlation value (for example 0.7). What about proper nouns or noun phrases? The relatedness of two phrases such as "the biggest defense against the US Navy's Laser Weapon System" and "China's crippling pollution problem". I would be appreciated if you could give me some references.

  • $\begingroup$ This is a tough problem. The closest thing I have personally heard of is the IBM Watson project. They digested Wikipedia and created such semantic relationships in order to answer Jeaopardy questions. Torsten Bittner gave a great talk on this in Seattle. It looks like another version is on YouTube. One can imagine a method that consists of 1) finding the wikipedia heading for each subject, 2) cross searching the subject for the other word, 3) downselecting cases with a clear object-subject pairing, 4) choosing the most common case. $\endgroup$ – AN6U5 Jul 24 '15 at 18:46
  • $\begingroup$ As @AN6U5 said, it is a really tough problem. From my experience, it is even more difficult to find an ontology that can cover everything. At the moment, ontologies are focused on a specific field. For example, you can find an ontology for Pathology or Algebra in Mathematics. Imagine that an ontology for Medicine is too broad based on the Pathology one. $\endgroup$ – Tasos Jul 27 '15 at 12:28

Other-than Ontologies do check word2vec, this tool provides an efficient implementation of the continuous bag-of-words and skip-gram architectures for computing vector representations of words and can be used to find the closest words for a user-specified word

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.