Abstract—This study aimed to use a cognition-inspired method following Hawkins’s approach to optimize learning sequences for efficiency. The model for this learning approach is a new, flexible associative form of memory that can handle keys of different lengths to address all fitting sequences. Furthermore, it cannot only identify existing sequences but also learn new ones and ensure fault-tolerant operations. After introducing such memory hardware, its practicability is approved as a new kind of spelling checker. The evaluation uses the TREC-5 Confusion Track standard dataset to automatically correct incorrect words by comparing them with Levenshtein Distance, pyspellchecker, Long Short-Term Memory (LSTM), and Semantically Conditioned LSTM plus Elmo Transformer (Elmosclstm). In a small data set and at the word level, the processing time is only 0.001s, which is lower than other methods. At the sentence level, the cognition-inspired method can achieve 99.31% accuracy, better than Elmosclstm at 81.97% for training data. In a big data set and at the word level, the highest accuracy is 87.38% and 87.03%, beyond Elmosclstm at 77.44% and 74.41% for training data and testing data. At the sentence level, the cognition-inspired method can achieve 96.73% and 91.42%, better than Elmosclstm at 81.50% and 72.18% for training and testing data, respectively.
 
Keywords—Artificial Intelligence (AI), Hierarchical Temporal Memory (HTM), spelling check


Copyright © 2023 by the authors. This is an open access article distributed under the Creative Commons Attribution License (CC BY-NC-ND 4.0), which permits use, distribution and reproduction in any medium, provided that the article is properly cited, the use is non-commercial and no modifications or adaptations are made.