实现记忆检索和回忆中漫游的一种计算AM模型（论文撤稿）

为了捕捉自传体记忆，提出了一种针对自传体记忆的计算AM模型.模型为3层网络结构.底层由5W1H构成的特定事件知识进行编码，提供检索线索；中间层通过关联特定事件知识对事件进行编码；顶层通过关联相关事件对事件集进行编码.按照自下而上的记忆搜索过程，AM模型可以分别在中间层和顶层识别相应的事件和事件集，同时，还能通过规则记忆检索过程模仿人在回忆中漫游的现象.基于一组数据集的实验结果表明，模型不仅具有鲁棒的、灵活的记忆检索，而且相较于普遍采用的基于关键字的查询方法记忆检索模型，对含噪记忆检索线索有更好的响应性能，还能够模仿回忆中的漫游现象.

Computational AM Model for Memory Retrieval and Wander in Reminiscence

For capturing autobiographical memory, a computational AM model for autobiographical memory is proposed. It is a three-layer network structure. Its bottom layer encodes event-specific knowledge comprising 5W1H and provides retrieval clues. The middle layer encodes events by associating the event-specific knowledge, and the top layer encodes episodes by associating related events. By following the bottom-up memory search procedure, the corresponding event and episode can be identified in the middle and top layers respectively. At the same time, the AM model can simulate the phenomenon of wandering in reminiscence through the retrieval process of regular memory. Experimental results based on a group of data sets show that the proposed computational AM model not only has robust and flexible memory retrieval, but, compared with the commonly used memory retrieval model based on keyword-based query method, has better response performance as well to noisy memory retrieval cues; it can also simulate the phenomenon of wandering in reminiscence.