Abstract
Named entity recognition (NER) has been seen as a fundamental component for various natural language processing (NLP) tasks, such as extracting information and answering questions. NER is used to comprehend the significance of information within a given context, and it also aids in retrieving and organizing data. The current study has introduced a new neural network, known as echo state network (ESN), to NER using CoNLL-2003. The characteristics were transformed into embeddings utilizing an embedding layer. These embeddings were then incorporated into the ESN. Additionally, the neural network was optimized utilizing the quantum-based sand cat swarm optimization algorithm. Eventually, CRF was employed to produce the predicted sequence of labels. Various assessment metrics, such as recall, precision, F1-score, MCC, and Cohen's Kappa, were used to evaluate the effectiveness of BiLSTM-MultiBERT6L, BiLSTM-CNNs-CRF, Bi-directional LSTM-CNNs, BiLSTM-ELMo, BERT, and the proposed ESN/quantum-based sand cat swarm optimization algorithm. Overall, it was demonstrated that the proposed model could achieve better results than the other proposed models. The main contribution of this study is the combination of QSCSO with ESN, which improves the model's capacity to comprehend long-term dependencies and effectively optimize hyperparameters. This research pushes forward the domain of NER and offers a scalable and efficacious architecture for related sequence labeling tasks. Recognizing entities such as person names, organizations, dates, and locations is essential as it allows machines to derive valuable insights from unstructured text. This aids in activities like information retrieval (for instance, locating pertinent documents), building knowledge graphs (such as connecting entities to establish relationships), and streamlining workflows (like summarizing news or extracting data for databases).