Review 1 Strong Points 1. The work introduces an interesting approach to leveraging severe weather data to improve product recommendations in the home improvement domain. 2. The authors provide a detailed analysis of performance gains across different product categories, showing that the system is capturing the influence of severe weather events on user behavior well. 3. The paper is overall well written and has practical impact. It acknowledges the potential of the presented approach to be adapted to other domains where events directly influence user intent. Weak Points 1. The paper's idea and framework is very similar to a cited paper [1] with very limited novelty. It's a different application of the previously published work, just applied on a different dataset. 2. The authors don't provide a detailed comparison of MAWI Rec with other methods that incorporate external data or context into recommender systems. The baselines used are limited to SASRec and variants of the proposed approach. In the SASRec baseline, it's possible that user behavior before and after severe weather is completely different (and SASRec can't capture it as it doesn't process weather data) and hence the model performance is worse. Including more SOTA methods for comparison and a more thorough discussion of how MAWI Rec improves upon existing approaches would be helpful. 3. It's not clear why the approach compares models by randomly ordering the set of products for each in-store purchase and predicting the last product in the sequence. 4. The paper doesn't discuss potential limitations or biases that may arise from using severe weather data. For example, the system may be biased towards recommending weather-related products even when they are not necessary or it may not capture the influence of other external factors (e.g. economic conditions, seasonal trends) on user behavior. [1] Brendan Duncan, Surya Kallumadi, Berg-Kirkpatrick Taylor, and Julian McAuley. 2023. Jointly modeling products and resource pages for task-oriented recommendation. (2023). https://doi.org/10.1145/3543873.3584642 Overall Evaluation -2: (reject) Despite introducing an interesting approach to leverage severe weather data for product recommendations in the home improvement domain, the paper lacks novelty as it closely resembles a cited work and is another application of it on a different dataset. The work doesn't provide a detailed comparison with state-of-the-art methods as it only uses SASRec and variants of the proposed approach as baselines. This limits the understanding of how well the proposed method performs compared to other techniques and doesn't provide enough information on its generalizability and effectiveness. The paper also doesn't sufficiently address the model comparison methodology and potential biases from using severe weather data, which impacts the strength of its overall contribution. Review 1 Strong Points 1. The work introduces an interesting approach to leveraging severe weather data to improve product recommendations in the home improvement domain. 2. The authors provide a detailed analysis of performance gains across different product categories, showing that the system is capturing the influence of severe weather events on user behavior well. 3. The paper is overall well written and has practical impact. It acknowledges the potential of the presented approach to be adapted to other domains where events directly influence user intent. Weak Points 1. The paper's idea and framework is very similar to a cited paper [1] with very limited novelty. It's a different application of the previously published work, just applied on a different dataset. 2. The authors don't provide a detailed comparison of MAWI Rec with other methods that incorporate external data or context into recommender systems. The baselines used are limited to SASRec and variants of the proposed approach. In the SASRec baseline, it's possible that user behavior before and after severe weather is completely different (and SASRec can't capture it as it doesn't process weather data) and hence the model performance is worse. Including more SOTA methods for comparison and a more thorough discussion of how MAWI Rec improves upon existing approaches would be helpful. 3. It's not clear why the approach compares models by randomly ordering the set of products for each in-store purchase and predicting the last product in the sequence. 4. The paper doesn't discuss potential limitations or biases that may arise from using severe weather data. For example, the system may be biased towards recommending weather-related products even when they are not necessary or it may not capture the influence of other external factors (e.g. economic conditions, seasonal trends) on user behavior. [1] Brendan Duncan, Surya Kallumadi, Berg-Kirkpatrick Taylor, and Julian McAuley. 2023. Jointly modeling products and resource pages for task-oriented recommendation. (2023). https://doi.org/10.1145/3543873.3584642 Overall Evaluation -1: (weak reject) Despite introducing an interesting approach to leverage severe weather data for product recommendations in the home improvement domain, the paper lacks novelty as it closely resembles a cited work and is another application of it on a different dataset. The work doesn't provide a detailed comparison with state-of-the-art methods as it only uses SASRec and variants of the proposed approach as baselines. This limits the understanding of how well the proposed method performs compared to other techniques and doesn't provide enough information on its generalizability and effectiveness. The paper also doesn't sufficiently address the model comparison methodology and potential biases from using severe weather data, which impacts the strength of its overall contribution. Review 2 Strong Points - The method is novel in the e-commerce domain. - The method outperforms SASRec. - The code and the dataset are provided to ensure reproducibility. Weak Points - The method is compared against only one baseline. - Only a selected few months were chosen for the evaluation. - An additional analysis of the impact of weather events is missing. Overall Evaluation 1: (weak accept) The authors of the paper propose a novel method that incorporates weather into predicting the next item a customer will purchase. While weather is typically considered a contextual factor in tourism recommender systems, its application in the e-commerce domain is new. The paper is well-written, easy to follow, and understand. The availability of the code and dataset enhances the reproducibility of this research. The authors demonstrate that their method outperforms the baseline in terms of NDCG and HR. However, it is unclear why specific months were chosen to represent each season, rather than using data from the entire year. It would be interesting to identify which weather events most significantly influence purchases. In Table 3, the authors show that the most substantial improvement is for flashing. It would be interesting to analyze which weather events and months affect purchases in this category. Review 3 Strong Points 1. Well-written and Easy to Follow: The paper is clear and concise, making it easy for readers to understand the methodologies and findings. This is crucial for effectively communicating complex ideas and ensuring that the audience can follow the progression of the research. 2. Concise Experiments Demonstrating SOTA Improvements: The experiments are well-designed and demonstrate significant improvements over state-of-the-art (SOTA) sequential models. The use of real-world datasets and clear comparison metrics enhances the credibility and applicability of the results. 3. Innovative Approach to Leveraging External Data: The incorporation of severe weather events as a contextual factor in the recommendation process is novel. This approach opens up new avenues for enhancing recommender systems by integrating external data that directly influences user behavior. Weak Points 1. Insufficient Evidence for Broad Claims: Towards the end of the paper, the claim that MAWI Rec is "a novel approach to the more general problem of inferring high-level user intent" is not fully substantiated. The evidence provided does not sufficiently support this broad claim, which may be seen as an overreach given the scope of the experiments. 2. Limited Scope of Weather Event Impact: While the focus on severe weather events is interesting, the paper could be criticized for not exploring other types of events or contexts that might influence user intent. This limitation restricts the generalizability of the proposed method to other domains or types of external influences. 3. Potential Overfitting Concerns: The results indicate significant improvements, especially in weather-related product categories. However, there might be concerns about overfitting to specific signals from the GloVe embeddings and weather categories. The paper would benefit from additional validation to ensure that these improvements are robust and not overly tailored to the specific dataset used. Overall Evaluation 1: (weak accept) Overall, the paper presents a well-executed study with minor contributions to the field of recommender systems. Despite some limitations, the practical implications of leveraging severe weather data to enhance recommendations are compelling. The paper should be (weak) accepted for publication, with a recommendation for the authors to temper their broader claims and to consider additional validations to address potential concerns about overfitting and the generalizability of their approach. Review 4 Strong Points + It’s a quite new application, and it is interesting to use severe weather data in recommendation + Many methods were considered in the experiment, including a state-of-the-art baseline and three variant of the proposed method. + The method looks theoretical profound Weak Points - In terms of novelty, in this work, I can see that the application and the data used (weather data) are novel to the audience. What about the method - is the method used in the RS also novel? - In terms of generalizability, it looks like this work only uses home improvement as an example. Can the contributions of this work be used in general marketplaces? Overall Evaluation 1: (weak accept) This work proposes a recommender system using severe weather data as inputs. This work also uses real-world data to conduct experiments, comparing with a state-of-the-art baseline and multiple variants. Metareview Metareview for paper 18 Title MAWI Rec: Leveraging Severe Weather Data in Recommendation Authors Brendan Duncan, Surya Kallumadi, Taylor Berg-Kirkpatrick and Julian Mcauley Text This is the meta-review by the senior program committee member responsible for your paper, taking into account the reviews and later discussion, as well as my own reading of the work. Considering the scores received by the reviewers and the overall alignment with the RecSys short paper track, the paper seems suitable for publication at the conference. While most reviewers were positive in their evaluation of the paper, the authors are recommended to take into consideration comments made in all reviews, ranging from, e.g., perceived lack of novelty (R1, R4), selection of baselines (R1, R2), and selections of data (R2, R4) to name a few. Note that the rebuttal has been taken into consideration by all reviewers and this meta review.