The invention relates to a method and a system for automatic analysis of a text entered by a user into a computing device, in order to offer suggestions for the text grammatical correction as well as to morphologically, syntactically and semantically improve the writing style. According to the invention, the method comprises: a training phase in which two types of corpora are used to create specific models, namely collections of reference documents for a certain field, to establish admissible values for various text complexity indices and a collection of wrong phrases and correction suggestions, to which, in a first stage, a corpus augmentation algorithm is applied, followed, in a second stage, by training a model based on neural networks, focused on the phrase automatic correction at the morpho-syntactic level, and a testing phase in which, in a first stage, a text is entered into the user's device, in the second stage the text is pre-processed by using natural language processing techniques, in a third stage, text complexity indices are calculated, in order for them to be used for the described style characterization and possibly grouped by decomposition means, such as PCA (Principal Component Analysis), into components which characterize various writing dimensions, in a fourth stage, rules adapted to each field are applied to verify if the values related to the indices/components are admissible, in relation to the collections of reference documents used in the training phase, in a fifth stage the correction model trained on the collection of wrong phrases is applied, and, in a sixth stage, a feedback concerning the entered text is generated while taking into consideration the recommendations resulted from the fourth and fifth stages. The claimed system is integrated in a computing device in which the user writes a text or uploads a document to be analyzed.
展开▼
