Abstract: is one of the key technologies for computer communication networks to provide intelligent online services for users. An understanding based Chinese automatic abstract system dedicated to a special field of neural networks learning algorithm （NNLA） is introduced in this paper, which is called LADIES （Literature Abstract and Digest Information Extract System）. Compared with other automatic abstract systems, LADIES uses the comprehensive information model based natural language processing theory for the first time. Comprehensive information is a name for the entirety of grammatical, semantic and pragmatic information of words. As a result, this model can describe the knowledge of words completely, which is very useful for sentence understanding. Semantic chunk assembling algorithm is another fine characteristic of LADIES, which is designed for grammatical and semantic sentence analyzing so as to solve the inherent difficulties of traditional sentence analyzing systems. Traditional systems need hard efforts to recognize the complicated grammatical structure of the whole sentence without understanding its semantic meaning, using word as the basic analyzing element. LADIES has also directly associated the sentence understanding with abstract information extraction with pragmatic information as the driver. Results of many experiments show that the whole set of method is very effective.