類神經網路已經被驗證可以應用於自動化的文件分類。然而,相較於其他的分類方法,要建構一個能夠自動分類的類神經網路,必須耗費龐大的空間以及冗長的訓練時間,這樣的限制導致類神經網路實用性大幅的降低。對此,本論文針對人類的神經系統與學習模式進行研究,並結合現有的類神經網路理論,提出了一個以人性為基礎的類神經網路概念,命名為人性化類神經網路。此網路概念主要包含了一個以擁有多種不同能力為導向而建立的類神經網路架構,與其在進行運作時可自動進行學習的非監督式學習方法。 本論文使用Reuters-21578進行文件分類之實證分析,觀察所提方法之成效。實驗結果顯示人性化類神經網路可以有效的將文件進行分類,其最佳的F1-measure可達92.5%;而所提的非監督式學習法則可有效且迅速的提升分類的準確性;且隨著架構的擴充,訓練與測試的時間是以線性成長,相較於傳統類神經網路的非線性成長,人性化類神經網路具有更良好的擴充性與實用性。
Artificial neural network (ANN) has been applied to text classification. However, comparing with other classification methods, ANN needs enormous memory space and training time to build the model. This makes ANN infeasible in practical applications. In this paper, we try to adopt the ideas of human neural system and learning style, and combine with the existing models of ANN. We propose a humanized neural network architecture that is based on human intelligence. In this architecture, a neural network with multiple abilities to solve numerous problems is presented, and with its unsupervised learning algorithm. In our experiment, Reuters-21578 was used as the dataset to show the effect of the proposed architecture on text classification. The experiment result showed that HNN can effectively classify texts with the best F1-measure of 92.5%. It also showed the learning algorithm can enhance the accuracy effectively and efficiently. With the expansion of the architecture, the training time and test time shows a linear growth. Comparing with ANN, HNN has better scalability and practicality.