Enhancing the Gujarati corpus for effective article summarization using word co-occurrence patterns

The increasing volume of digital content produced in regional languages has now made text summation processes essential for information retrieval. The Indian dialect known as Gujarati has difficulties with article summaries because there are few articles and resources with high state-of-the-art linguistics. This paper presents a word co-occurrence information-based summarization approach that significantly improves quality of summary over LexRank method, Word Frequency method, TF-IDF Frequency method, which scored 78.26%, 49.72%, and 64.50% respectively and achieved the 92.22% similarity. By using semantic connections between words our approach improves text relevance and coherence, solving the complex language problems of the Gujarati language. To perform extractive and abstractive paragraph summarization, statistical co-occurrence methods are combined with natural language processing in the framework. The co-occurrence patterns-based methods gave valuable efficiency in controlling repetition of text: not only in Gujarati, but especially in other low-resourced languages. Enhancing the Gujarati corpus, along with the incorporation of transformers deep learning models and the hybrid model blending statistical and neural network approaches to summarization will improve the estimation accuracy. The development of efficient information processing systems for low resourced languages, especially for the cases of Gujarati language, increases the performance of AI news summary applications, education, and other modern technology.