A Word Embedding Model for Topic Recommendation

The proposed work aims to recommend relevant hash-tags from social media posts and blog posts related to food, recipes and other popular domains using topic word-embeddings. These hash-tags can be used in turn to give food item/recipe as the user requires. For topic modeling a machine learning approach called LDA(Latent Dirichlet Allocation) is used. The method treats the social media posts as a corpus of words and applies a topic modeling to generate the relevant hash-tags. The hash-tags recommended are cross validated with a word embedding model developed from the Wikipedia corpus to check the accuracy of the recommendation system. This algorithm is further tested on posts that do not contain hash-tags to recommend new relevant hash-tags for the same.