Framework Based Summarization approach for Generating Ideological and Political Education Texts

In recent days, text summarization in ideological and political education enhances learning efficiency by extracting key ideas while preserving core educational values. Traditional approaches for text summarization had faced several challenges which include loss of context and lack of semantic understanding. Therefore, this research proposes Framework-based Summarization (FrameSum) for automatic summarization generation of ideological and political education texts. Initially, data is collected from Cable News Network/Daily Mail dataset (CNN/Daily Mail Dataset) which contains 300,000 news articles from CNN and Daily mail where each paired with human-written summaries and also includes framework information which provides contextual knowledge about the underlying concepts. Then, this collected data is preprocessed by using stopword removal, tokenization and lemmatization which removes irrelevant words, breaks text into tokens, and reduces words to their base form for analysis. After that, similarity between sentences is determined by Similarity based on Overlap of Important Words (SOIW) which effectively captures the degree of resemblance between two sentences. Finally, proposed FrameSum generates the text summaries by incorporating framework information, capturing ideological and political attributes. The proposed FrameSum achieved better results in terms of accuracy (0.95), precision (0.89), recall (0.91) and F1-score (0.92) when compared with existing Neural Framework for Review Manipulation (NFRM).