Seismic attributes are critical in understanding geological factors, such as sand body configuration, lithology, and porosity. However, existing attributes typically reflect a combined response of multiple geological factors. The interplay between these factors can obscure the features of the target factor, posing a challenge to its direct seismic characterization, particularly when the factor is subtle. To address this, we develop an innovative neural network designed to disentangle and characterize the individual geological factor within seismic data. Our approach divides the geological information in the seismic data into two categories: the single geological factor of interest and an aggregate of all other information. A novel feature-swapping mechanism within our network facilitates the disentanglement of these two categories, providing an interpretable representation. We employ a triplet loss function to differentiate data samples with similar waveforms but varying subtle geological details, thus enhancing the extraction of distinct features. Additionally, our network employs a co-training strategy to integrate synthetic and actual field data during the training process. This strategy helps mitigate potential performance degradation arising from discrepancies between simulated and actual field data. We apply our method to synthetic data experiments and field data from two geologically distinct areas. Current results indicate that our method surpasses traditional approaches such as a deep autoencoder and a convolutional neural network classifier in extracting seismic attributes with more explicit geophysical implications.