Analogies are a comparison between one word and another. Humans compare these words based on many factors like gender, is it a place, can we eat it, what is it's use and so on. When it comes to machines they use similar procedure where each word has a set of attributes and it is rated within range -1 to 1. Based on the relation between the first 2 words the algorithm will find the difference of attributes and their values and use them as reference to find the word that matches the third word from the list. I have used a pre-trained Global vector for word representation that has word and its embeddings. Each word has 50 values and there are a total of 4,00,000 words in it's vocabulary.

The Input can be given in form of

After going through the entire dictionary The answer is displayed as

Must download glove 50d embedding matrix along with this repo from here and rename the file as "glove.txt" after extraction.

Place all the downloaded files in same folder.