This paper considers episodic memory for companion–human interaction, aiming at improving user experience of interactions by endowing social companions with awareness of past experience. Due to noise and incomplete cues from natural language and speech in real-world interaction, accurate memory retrieval is very challenging and the noise resistance is important in practice. To improve the robustness of companion–human interaction, we propose a mixed-correlation analysis-based episodic memory (MCAEM) model, in which the correlations between memory elements are analyzed and then utilized for memory retrieval. In particular, the correlations are analyzed in three aspects: the relations between elements, importance of attributes and order of events. Based on the mixed-correlation analysis, a new similarity measure is constructed, which has substantially enhanced the noise resistance of memory retrieval. Experiments on a dataset collected from interaction in movies quantitatively evaluate the MCAEM model and compare it with prior work. Also, a user study is conducted to investigate the benefits of integrating the MCAEM model into social companions. The results demonstrate that the companions equipped with the MCAEM model not only have better retrieval performance, but also improve user experience in many aspects.
The Visual Computer – Springer Journals
Published: May 10, 2018
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
All the latest content is available, no embargo periods.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud