An Improved Text Similarity Calculation Method Combining TF-IDF and LexRank

In text similarity calculation, the traditional TF-IDF (Term Frequency Across Document Frequency) method is widely used in text retrieval and information extraction tasks due to its simplicity, ease of understanding, and efficiency. However, the TF-IDF method has certain limitations in handling text semantics and contextual information. To compensate for this deficiency, this study proposes an improved text similarity calculation method combining TF-IDF and LexRank. We improved the TF-IDF weight calculation by introducing feature concentration, thereby enhancing the ability to distinguish keywords. The LexRank algorithm is based on sentence similarity maps and PageRank ideas, further enhancing the ability to capture semantic relationships between sentences. Combining these two methods, we have designed a comprehensive scoring model to evaluate text similarity. The experimental results show that the improved method combining TF-IDF and LexRank performs better in processing text similarity calculations than using TF-IDF or LexRank alone, with higher accuracy and stability. This method has demonstrated significant performance improvement in application fields such as text classification and information retrieval, and has high practical application value.