Abstract The paper computed small yet representative sets of reasonable paths in a multi-modal transportation scenario involving cars, walking, and transit. Traditional route planning systems typically optimize for a single mode of transportation, but TNT aims to offer... The paper computed small yet representative sets of reasonable paths in a multi-modal transportation scenario involving cars, walking, and transit. Traditional route planning systems typically optimize for a single mode of transportation, but TNT aims to offer diverse and practical path combinations without preselecting a specific mode. Using Pareto sets with multiple optimization criteria like duration, transfer penalty, and car duration, the method captures diverse routes. However, not all Pareto-optimal paths are reasonable, so the authors define three types of reasonable paths based on relative durations: only car, mixed transit with limited car and walking, and much transit with walking. Thresholds filter unreasonable paths and extract concise results. Speed-up techniques include extending dominance by early results, rounding transfers to full minutes, and using implicit walking duration. Experimental results across cities like New York demonstrate that TNT effectively reduces query times to about one second on average while maintaining high result quality. The work emphasizes generating manageable, representative route options rather than strictly optimal ones, enhancing usability in real-world applications. Read More Read Less

Abstract The paper explores the identification of subjective language in personal narratives, focusing on distinguishing between two narrative levels: diegetic (events within the story) and extradiegetic (the narrator’s reflections). Subjective language, expressing emotions, opinions, and mental states, plays... The paper explores the identification of subjective language in personal narratives, focusing on distinguishing between two narrative levels: diegetic (events within the story) and extradiegetic (the narrator’s reflections). Subjective language, expressing emotions, opinions, and mental states, plays a crucial role in shaping the audience’s interpretation. The study uses a dataset of 40 annotated personal weblog narratives, employing text classification techniques to automatically identify subjectivity at both levels. A multiclass classification model is trained using features like bag-of-words and part-of-speech tags. Results show a 58% accuracy for six-way classification, outperforming a baseline. Binary classifications for subjectivity and narrative level achieve 78% and 81% accuracy, respectively. Despite limitations due to a small dataset, the findings highlight the feasibility of computational modeling for analyzing narrative subjectivity, with potential applications in sentiment analysis, information retrieval, and commonsense knowledge extraction. Read More Read Less