Traditional summarization evaluation metrics like ROUGE, BLEU, etc. have a serious limitation due to the fact that they essentially use word-matching based calculations while comparing a predicted summary against a reference summary, and ignore the semantic similarity completely. To deal with this issue, in this TextSumEval 2022 shared task submission, we propose a semantically rich summarization evaluation approach. The main idea is that both the summaries are firstly represented as a DistilRoBERTa based high-dimensional vector, and then cosine similarity based vector semantic calculation is carried out. The proposed approach achieves Pearson’s correlation values of 0.44 and 0.62 respectively w.r.t the BLEU and ROUGE metrics for the training set, and 0.6081 and 0.6635 for the test set. Also, the proposed approach obtains a mean semantic score of 0.6579 on the test set. With the proposed approach, we can generate positively correlated scores like traditional metrics, while capturing semantic similarity.