Automatic Text Generation Via Text Extraction Based On Submodular

Automatic text generation is the generation of natural language texts by computer. It has many applications, including automatic report generation, online promotion, etc. However, the problem is still a challenged task due to the lack of readability and coherence even there are many existing works studied it. In this paper, we propose a two-phase algorithm, which consists of text cleanup and text extraction, to automatically generate text from multiple texts. In the first phase, we generate paragraphs based on the topic modeling and clustering analysis. In the second phase, we model the text extraction as a set covering problem after we find the keywords in terms of the scores of TF-IDF, and solve the problem via employing the tool of submodular. We conduct a set of experiments to evaluate our proposed method and experimental results demonstrate the effectiveness of our proposed method by comparing with some comparable baselines.