89 lines
15 KiB
Markdown
89 lines
15 KiB
Markdown
type:: [[REVIEWS]]
|
||
tags::
|
||
year:: 2024
|
||
venue:: [[RecSys]]
|
||
full-title:: Addressing Personalized Diversity in Eyewear Recommendation: a Lenskart Case Study
|
||
date-start:: [[17-08-2024]] - 08:12
|
||
date-submitted::
|
||
external-links::
|
||
status:: [[DONE]]
|
||
deadline-submission::
|
||
file::
|
||
parent::
|
||
todoist:: https://app.todoist.com/app/task/791-addressing-personalized-diversity-in-eyewear-recommendation-a-lenskart-case-6W224vqXxV47x3wc
|
||
|
||
- ### [[Highlights]]
|
||
- # Annotazioni
|
||
- (17/8/2024, 07:42:13)
|
||
- - “(1) integration of diverse environmental data” ([“RecSys_2024_paper_779”, p. 1](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=1&annotation=522NBYS3)) #5fb236
|
||
- - “(2) a dynamic weighting algorithm allowing fine-grained user control over routing parameters;” ([“RecSys_2024_paper_779”, p. 1](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=1&annotation=D77785LK)) #5fb236
|
||
- - “(3) transparent decision-making via an interactive interface displaying edge-level information;” ([“RecSys_2024_paper_779”, p. 1](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=1&annotation=BFVGR249)) #5fb236
|
||
- - “(4) a community-driven feedback mechanism that continuously refines recommendations.” ([“RecSys_2024_paper_779”, p. 1](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=1&annotation=5A7NPD5F)) #5fb236
|
||
- - “These improvements come with slight trade-offs in distance, especially in walking mode” ([“RecSys_2024_paper_779”, p. 1](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=1&annotation=KC94AZPC)) #5fb236
|
||
- - “comprehensive framework for multi-criteria route recommendation, balancing individual preferences, environmental factors, and community insights in urban navigation.” ([“RecSys_2024_paper_779”, p. 1](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=1&annotation=W4YVVJRJ)) #5fb236
|
||
- - “predominantly prioritize route efficiency” ([“RecSys_2024_paper_779”, p. 1](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=1&annotation=KYCRR2YK)) #5fb236
|
||
- - “typically measured by shortest travel time or distance” ([“RecSys_2024_paper_779”, p. 1](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=1&annotation=FWK3F7AM)) #ffd400
|
||
- *Google Maps, I'm not aware of other systems, give also information about Co2 emission for each alternative route.*
|
||
- - “multifaceted human-centric factors such as environmental quality, public health, and personal preferences.” ([“RecSys_2024_paper_779”, p. 1](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=1&annotation=U282ST54)) #5fb236
|
||
- - “user empowerment and environmental considerations.” ([“RecSys_2024_paper_779”, p. 1](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=1&annotation=4LCQBVNU)) #5fb236
|
||
- - “these systems frequently consider such factors in isolation” ([“RecSys_2024_paper_779”, p. 1](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=1&annotation=SWZHU5WJ)) #a28ae5
|
||
- - “without a holistic integration of multiple user-defined preferences” ([“RecSys_2024_paper_779”, p. 1](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=1&annotation=G3Q23FPM)) #a28ae5
|
||
- - “black-box nature of many optimization algorithms” ([“RecSys_2024_paper_779”, p. 1](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=1&annotation=YZ9VVVRS)) #a28ae5
|
||
- - “offering users little insight into how recommended routes align with their personal and environmental values,” ([“RecSys_2024_paper_779”, p. 1](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=1&annotation=D2AG7V3F)) #a28ae5
|
||
- - “Our research introduces Routify1 (Route intelligence for you), an open-source route recommendation system that fundamentally shifts the paradigm from the optimization to the democratization of route selection.” ([“RecSys_2024_paper_779”, p. 1](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=1&annotation=FLBBGMU4)) #ffd400
|
||
- - “air quality, noise levels, greenery, altitude, and real-time traffic conditions.” ([“RecSys_2024_paper_779”, p. 1](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=1&annotation=47KY76N9)) #a28ae5
|
||
- - “The problem with optimization in urban navigation often stems from its narrow focus on efficiency metrics (such as distance or travel time), neglecting qualitative and environmental aspects.” ([“RecSys_2024_paper_779”, p. 1](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=1&annotation=YE8RPI6D)) #f0ff00
|
||
- *Not fully true, systems like Google maps also suggest cow concerned recommendations*
|
||
- - “This can lead to suboptimal outcomes, for example, routes that save time but expose users to higher pollution levels or less appealing environments.” ([“RecSys_2024_paper_779”, p. 1](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=1&annotation=2BBXX5UW)) #a28ae5
|
||
- - “one-size-fits-all recommendations can create new congestion situations” ([“RecSys_2024_paper_779”, p. 1](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=1&annotation=3YMALBQQ)) #a28ae5
|
||
- - “multi-criteria recommendation system” ([“RecSys_2024_paper_779”, p. 1](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=1&annotation=IYNJ76QQ)) #a28ae5
|
||
- - “Providing custom parametrization that allows users to weigh different factors according to their preferences” ([“RecSys_2024_paper_779”, p. 1](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=1&annotation=S2XPPENE)) #a28ae5
|
||
- - “Promoting transparency and trust through its open-source nature and user-centric, value-sensitive design,” ([“RecSys_2024_paper_779”, p. 1](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=1&annotation=ZDFSUSK7)) #e56eee
|
||
- - “empowers users to make choices that reflect their personal values.” ([“RecSys_2024_paper_779”, p. 2](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=2&annotation=U5ARH9SG)) #e56eee
|
||
- - “prioritize routing for motor vehicles and public transit” ([“RecSys_2024_paper_779”, p. 2](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=2&annotation=PTV4ELF6)) #a28ae5
|
||
- - “The most accurate proxies of routing preferences [12, 13] also depend on subjective, cultural, and environmental factors [14].” ([“RecSys_2024_paper_779”, p. 2](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=2&annotation=PQGRTCMN)) #a28ae5
|
||
- - “the scarcity of data on relevant features” ([“RecSys_2024_paper_779”, p. 2](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=2&annotation=T9AYFT3M)) #f0ff00
|
||
- *How have you addressed that?*
|
||
- - “these services offer limited personalization, often extending the logic of vehicular routing to cycling and pedestrian mobility” ([“RecSys_2024_paper_779”, p. 2](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=2&annotation=GZVFBIYZ)) #a28ae5
|
||
- - “Moreover, despite their differences, the above services remain opaque regarding the parameters and factors used in route computation” ([“RecSys_2024_paper_779”, p. 2](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=2&annotation=MTQY5A9J)) #2ea8e5
|
||
- - “Personalized routing algorithms can provide alternatives that are more aligned to user preferences” ([“RecSys_2024_paper_779”, p. 2](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=2&annotation=2ENHKS56)) #a28ae5
|
||
- - “aggregated synthetic metrics” ([“RecSys_2024_paper_779”, p. 2](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=2&annotation=S85C95CU)) #5fb236
|
||
- - “. First, the integration of real-time, hyper-local environmental data (such as air quality and noise levels) into routing algorithms is still limited.” ([“RecSys_2024_paper_779”, p. 2](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=2&annotation=XFLICLNK)) #5fb236
|
||
- - “First, the integration of real-time, hyper-local environmental data (such as air quality and noise levels) into routing algorithms is still limited.” ([“RecSys_2024_paper_779”, p. 2](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=2&annotation=YN9GVTAW)) #f0ff00
|
||
- *Is it because of the lack of data or what?*
|
||
- - “systems lack transparency in their decision-making processes, operating as "black boxes" that users must trust without understanding them. Third,” ([“RecSys_2024_paper_779”, p. 2](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=2&annotation=SHR4TAM5)) #f0ff00
|
||
- *But this is not a scientific contribution, isn't it?*
|
||
- - “Routify addresses these gaps by providing a transparent, highly customizable system that integrates real-time environmental data and leverages community input.” ([“RecSys_2024_paper_779”, p. 2](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=2&annotation=W43M8Y3A)) #a28ae5
|
||
- - “goals of sustainable urban development and citizen empowerment in smart cities.” ([“RecSys_2024_paper_779”, p. 2](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=2&annotation=6IL6X364)) #5fb236
|
||
- - “diverse data sources to construct a comprehensive representation of the urban environment.” ([“RecSys_2024_paper_779”, p. 2](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=2&annotation=4CTM9MNB)) #5fb236
|
||
- - “his heterogeneous data is synthesized into a unified G” ([“RecSys_2024_paper_779”, p. 2](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=2&annotation=PGDECHPE)) #5fb236
|
||
- - “Graph” ([“RecSys_2024_paper_779”, p. 2](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=2&annotation=XBVGP87C)) #f0ff00
|
||
- *What are the nodes of the graph?*
|
||
- - “ges” ([“RecSys_2024_paper_779”, p. 3](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=3&annotation=F3S2A4BR)) #f0ff00
|
||
- *Are these defined on es for all? Some more details on the graph based representation would help understand better the approach.*
|
||
- - “The Application Layer serves as the interface between the system and the user. It presents intuitive visualizations and statistics derived from the routing results, allowing users to make informed decisions. Users interact with the system by inputting their preferences for various factors, represented by intuitive icons (e.g. green index, noise levels, etc.). Additionally, users can provide feedback on suggested routes, contributing” ([“RecSys_2024_paper_779”, p. 3](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=3&annotation=TJDFQFVU)) #f0ff00
|
||
- *The main focus of the paper is the algorithm, the encoding in a graph based representation of all data, the application users can exploit, all of them?*
|
||
- - “RESULTS” ([“RecSys_2024_paper_779”, p. 3](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=3&annotation=BT2IAANT)) #00b036
|
||
- *I'm not sure how to interpret the obtained results. Are they good, bad? What's the baseline to compare with?*
|
||
- - “control the family-wise error rate due to multiple comparisons, we applied the Bonferroni correction [30] to adjust the p-values.” ([“RecSys_2024_paper_779”, p. 3](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=3&annotation=X5BK4W8F)) #f0ff00
|
||
- - “Future work will focus on expanding to further environmental factors, developing advanced preference learning algorithms, and exploring multi-objective optimization techniques. We aim to integrate contextual factors like time-of-day and weather conditions, and investigate collaborative filtering approaches to enhance recommendations while maintaining privacy.” ([“RecSys_2024_paper_779”, p. 4](zotero://select/library/items/WKA5KN45)) ([pdf](zotero://open-pdf/library/items/7J9BPRVY?page=4&annotation=SZKM8JCD)) #f0ff00
|
||
- *Indeed it is interesting even though I'm missing missing a clear discussion about the sciebtific challenges author had while conceiving the proposed approach.*
|
||
- ### [[Comments]]
|
||
- The paper presents an investigation on enhancing the diversity of recommendations on the Lenskart e-commerce platform by integrating re-ranking algorithms like Maximal Marginal Relevance (MMR), Determinantal Point Processes (DPP), and Contextual Bandits (LinUCB) with Factorization Machines (FM). The study aims to balance personalization with diversity, which is crucial for user engagement.
|
||
- The paper is about a relevant and practical problem. However, the improvements in recommendation effectiveness appear marginal, and the paper needs a discussion on the generalizability of the findings, which is essential for scientific rigor.
|
||
- Strong points:
|
||
- The proposed solution is for a real and industrial problem
|
||
- The combination of MMR, DPP, and LinUCB with FM is a novel approach to improving recommendation diversity
|
||
- Weak points:
|
||
- The paper contains some repetitive sections, particularly explaining the algorithms, which could have been streamlined for clarity and conciseness.
|
||
- The reported improvements in recommendation effectiveness are marginal, which may limit the practical applicability of the proposed methods.
|
||
- The paper does not discuss the generalizability of the results or the potential threats to the validity of the findings, which is critical for the broader application of the study’s conclusions.
|
||
- ### [[REVIEWS/Notes]]
|
||
- ### YELLOW CONCERNS
|
||
background-color:: yellow
|
||
- {{query (and [[ffd400]] [[RecSys_2024_paper_791]] )}}
|
||
collapsed:: true
|
||
- ### ❓️Questions
|
||
- {{query (and [[question]] [[RecSys_2024_paper_791]] )[[question]]}}
|
||
query-table:: true
|
||
query-properties:: [:block] |