As an essential part of educational psychology, the propagated influence among pedagogical concepts (i.e., learning transfer) is important for optimizing Knowledge Tracing (KT) tasks. However, existing KT methods only consider the positive learning transfer and disregard the negative learning transfer. Thus, this paper proposes an innovative positive and negative Learning Transfer-based Knowledge Tracing model (LTKT), which makes the first attempt to concurrently utilize the positive and negative learning transfer relations among concepts to improve KT results. First, LTKT constructs a learning transfer graph. Then, a direct learning effect component and a learning transfer effect component are carefully designed in LTKT. The first component quantifies the impact of an exercise’s practice result on the concept examined in the exercise. The second component, in contrast, computes the impact of the result on the concept’s neighbouring concepts in the constructed learning transfer graph, considering the positive and negative learning transfer phenomena. Extensive experiments on publicity datasets demonstrate that LTKT outperforms all state-of-the-art KT methods.

Deep learning has found widespread application across diverse domains owing to its exceptional performance. Nevertheless, the lack of transparency in deep learning models’ decision-making processes undermines their usability, especially in critical contexts. While researchers have made noteworthy advancements in explaining these models, they have frequently overlooked the differences between static and temporal models during explanation generation. In temporal models, features change over time, posing new challenges in the generation of explanations. Though extensive research has been dedicated to surmounting these hurdles, a survey summarizing these contributions is currently absent. To bridge this gap, this paper endeavors to summarize existing methods and their contributions in terms of both static and temporal models, highlighting their disparities. Additionally, we propose an innovative classification approach based on the comprehensibility of explanations, demonstrating that different explanation methods vary in their understandability for users. Finally, to assess the limitations of the explanation capabilities of existing methods, we specifically choose knowledge tracing to analyze the evolution of explanation methods in this context of temporal modeling and interpretations.

Trip recommendation has become increasingly popular with the rapid growth of check-in data in location-based social networks. Most existing studies focused only on the popularity of trips. In this paper, we consider further the usability of trip recommendation results through spatial diversification. We thereby formulate a new type of queries named spatial diversified top-k routes (SDkR) query. This type of queries finds k trip routes with the highest popularity, each of which starts at a given starting point, consumes travel time within a given time budget, and passes through points of interest (POIs) of given categories. Any two trip routes returned are diversified to a certain degree defined by the spatial distance between the two routes. We show that the SDkR problem is NP-hard. We propose two precise algorithms to solve the problem. The first algorithm starts with identifying all candidate routes that satisfy the query constraints, and then searches for the k-route combination with the highest popularity. The second algorithm identifies the candidate routes and builds up the optimal k-route combination progressively at the same time. Further, we propose an approximate algorithm to obtain even higher query efficiency with precision bounds. We demonstrate the effectiveness and efficiency of the proposed algorithms on real datasets. Our experimental results show that our algorithms find popular routes with diversified POI locations. Our approximate algorithm saves up to 90% of query time compared with the baseline algorithms.

Abstract Roles, their emotion, and interactions between them are three key elements for semantic content understanding of movies. In this paper, we proposed a novel movie summarization method to capture the semantic content in movies based on a string of IE-RoleNets. An IE-RoleNet (interaction and emotion rolenet) models the emotion and interactions of roles in a shot of the movie. The whole movie is represented as a string of IE-RoleNets. Summarization of a movie is transformed into finding an optimal substring with user-specified summarization ratio. Hierarchical substring mining is conducted to find an optimal substring of the whole movie. We have conducted objective and subjective experiments on our method. Experimental results show the ability of our method to capture the semantic content of movies.