You might be asking yourself, how do you use CNNs to classify text when they are most commonly used in image processing?
There are many
discussions in the literature, linked at the bottom of this tip, which have proven that CNNs are a generic feature extraction function that can compute
location invariance and
compositionality. The location invariance property helps the model to capture the context of words, irrespective of their occurrence in the corpus. Compositionality helps to derive higher-level representations using lower-level features:
- A CNN Based Scene Chinese Text Recognition Algorithm with Synthetic Data Engine ...