Machine learningDeep learning / NLP / CV
Multimodal Doc2Vec
Multimodal Doc2Vec extends the Doc2Vec paragraph-vector framework to incorporate information from more than one modality — typically text alongside images, audio, or structured metadata — producing a shared document-level embedding that captures semantics from multiple sources simultaneously. It is used for cross-modal retrieval, multi-source classification, and document representation where text alone is insufficient.
MethodMind'de açSoonVideoSoon
Tam yöntemi oku
Members only
Sign inSign in with a free account to read this section.
Sources
- Le, Q. V., & Mikolov, T. (2014). Distributed Representations of Sentences and Documents. Proceedings of the 31st International Conference on Machine Learning (ICML), PMLR 32(2), 1188–1196. link ↗
- Ngiam, J., Khosla, A., Kim, M., Nam, J., Lee, H., & Ng, A. Y. (2011). Multimodal Deep Learning. Proceedings of the 28th International Conference on Machine Learning (ICML), 689–696. link ↗