I am working on a project with a huge number of big groups of sentences! I need to classify each sentence based on the sentence and other sentences on the group. Actually there is 2 level of classification, in the first level you should classify only based on the sentence, in the second level you should look at other sentences on the group and recognize level 2 class!
Each sentence has its own class and are too big (tens of thousands of words!) so we cannot simply put them together and classify whole group! And simple classifying sentence and others and then chose the final class accordingly do not have a good result(very low accuracy). Because we have some errors in each level of classification and in this way the errors would increase exponentially!!
The following picture will explain how we should classify sentence 4:
Which approache, architecture, algorithm, model or... do you suggest, any suggestions are appreciated.