With the advent of natural language processing and deep learning, the expectation on generating new approaches for solving issues related to Indian local language has increased. In this direction, Kannada and its literature poses many challenges. One among them is developing a solution for retrieval of Vachanas written between twelfth and nineteenth centuries encounters many issues. In this work, we present first of its kind work on Kannada Vachana (a poetic form) retrieval based on their philosophical interpretation (meaning) of Vachana using NLP and deep learning for retrieval purpose. The approach presents a challenging task of understanding philosophical thoughts of Vachanas and retrieving it based on user query text. We present two approaches based on Term Frequency-Inverse Document Frequency (TF-IDF) and Kannada sentence embedding model and compare both the results. The experiments are carried out for exact and similarity match retrieval of Vachanas based on user query text. The results are promising, for exact match retrieval we have achieved about 87 and 90% of retrieval accuracies and for similarity retrieval we have achieved about 51 and 57% of retrieval accuracies for TF-IDF and Kannada sentence embedding, respectively.