Semantic Similarity

The challenge of automatically measuring semantic similarity is to discern the degree of similarity between two different textual expressions by means of computer. Today, it is one of the most interesting research challenges that the community of researchers and artificial intelligence professionals have to face.

There are many methods, techniques and tools to undertake this task. From those based on the use of dictionaries that have been manually compiled like WordNet, to the latest vectorization techniques using deep neural networks like BERT, through Google Normalized Distance, word vectorization through word2vec, or the creation of semantic similarity controllers based on fuzzy logics to increase the interpretability of the system.

It is clear that in the near future we will see the emergence of many more computational methods, techniques and tools to meet this challenge. However, there are some problems that will have to be solved, such as increasing the interpretability of the models used to predict semantic similarity in order to that people can understand them.