Non-structured text features, if vectorized via embedding models, can then be combined with regular structured/tabular features to enhance the performance of traditional predictive models. However, since text embedding vectors are high-dimensional numerical vectors, more storage and computational resources will be consumed if they are to be in use. To alleviate this issue, we may apply dimensionality reduction to the embedding vectors while keeping the performance of the corresponding predictive models largely un-affected.
Non-structured text features, if vectorized via embedding models, can then be combined with regular structured/tabular features to enhance the performance of traditional predictive models. However, since text embedding vectors are high-dimensional numerical vectors, more storage and computational resources will be consumed if they are to be in use. To alleviate this issue, we may apply dimensionality reduction to the embedding vectors while keeping the performance of the corresponding predictive models largely un-affected. Read More Technology Blogs by SAP articles
#SAP
#SAPTechnologyblog
