Supporting Information Recall for Elderly People in Hyper Aged Societies

As people become elderly, they often suffer from memory loss. This can present itself in a conversation where the participants cannot recall, for example, the name of a place they visited or a person they met. In this work, we present a support system for people in such situations that 1) recognizes utterances that trigger situations where information recall is required and then 2) retrieves the necessary information from data sources such as news articles. Our system is composed of two modules: a recognizer for utterances that trigger information recall and a search engine to retrieve necessary information. We describe the task of detecting utterances that trigger information recall and present a novel corpus. Then, we present the details of the search engine. The performances of baseline models for detecting utterances that trigger information recall suggest that these models achieve excellent performances in terms of recall but there is still have room for improvement in terms of precision. Moreover, the performance of the overall proposed system suggests that the retrieval task is somewhat difficult and merits further investigation.