Readability of Spanish e-government information

This paper proposes the automated evaluation of readability, through the analysis of different linguistic characteristics associated with a better understanding of the websites of the Spanish Government's administrative procedures. To fulfil this task, a corpus made by web documents with distinct difficulty levels has been gathered. Then, these documents' difficulty is assessed through different classic readability metrics. By the use of machine learning methods, different algorithms are analyzed to measure their capability to predict text difficulty. The results obtained show that the official Spanish Government websites have a high difficulty level. The main contribution of this work is the combined application of a wide number of linguistic attributes and the construction of a new corpus addressed to official government texts.