In natural language processing (NLP), machine learning models often encounter challenges due to the lexical ambiguity of languages, particularly in tasks such as machine translation, Text-to-Speech (TTS), and information retrieval. Homographs present significant challenges in this context. While many studies focus on homographs in resource-rich languages, such works are lacking for languages with limited resources. In this paper, we present a lexical resource for Somali homograph Disambiguation based on a dictionary. We employed various natural language processing techniques, including sentence embeddings and machine learning clustering algorithms, to analyze 4,809 homograph entries, representing 1,592 unique homographs. We created distributions of homographs based on their semantic relationships. Our research provides insights into the distribution, semantic relationships, and clustering patterns of Somali homographs, contributing to the understanding of lexical ambiguity in this understudied language. The data is available at https://github.com/Arralle21/homo .