Progress 12/01/23 to 11/30/24
Outputs
Target Audience:During this reporting period, we have engaged with community members in outreach efforts in preparation for the final intervention study and to execute a pilot intervention with a small community group. These are typically individuals from a low-socioeconomic background that arefood insecure. In addition to actual stakeholders that live in such an area, our project also engages community partners who are our liaisons to actual community members as well as community resources such as food pantries. We have also engaged with multiple undergraduate students who have served as researchers and/or developers on the project. Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided?The project has supported several graduate students (2at ND and 2and U.Michigan) who are being trained as Human-Computer Interaction researchers. How have the results been disseminated to communities of interest?
Nothing Reported
What do you plan to do during the next reporting period to accomplish the goals?The final tasks are to 1. Execute the intervention study and associate data collection 2. Analyze the intervention study data 3. Produce publications on the design of the appand results of the intervention as well as publicationson the findings from the Detroit studies.
Impacts
What was accomplished under these goals?
With regard to RQ6, we have completed the prototype are are currently beta-testing it in a small "pilot intervention". This pilot is being conducted with 5 participants and mirrors the full process of the large-scale intervention that we will carry out this spring. This intervention will allow us to collect data to answer RQ8. As you may recall from our 2023 annual progress report, our pilot study and ethnographic data suggest that physical access is actually much less of a problem than initially expected given the use (in South Bend) of local social networks to overcome the physical access barriers. In addition, food delivery to the home has been more normalized during the COVID19 pandemic and in some cases, even subsidized for some community members. In Detroit, the geographic properties are much different with neighborhoods being much more spread (distance-wise) and grocery stores being located much further on the margins. As a result, neighborhoods rely much more heavily on local food source providers that are typically not large commercial chains. These findings led us to believe that the planned "physical access" part of the intervention would not be successful in either South Bend or Detroit. Rather than moving forward with the hub delivery model, we pivoted o explore how to better integrate "local non-commercial" food sources into the "information access" part of the project. These local food sources included pantries, community gardens, and farmers' markets, for example, and they pose a challenge primarily around real-time tracking of inventory. This pivot direclty applies to RQ7 which we have replaced with the updated question: Updated RQ7: How can informal food sources be successfully integrated into technological food access solutions? To address this question, our team conducted an in-depth investigation into the role of technology in supporting food sovereignty initiatives, a key theme of Detroit urban farmers. This effort culminated in a late-breaking work submission to CHI 2025, Technological Tensions In Urban Food Sovereignty: Insights from Detroit's Urban Farmers. Over three months of fieldwork and three participatory design workshops, we explored how farmers perceive opportunities and challenges in using technology to address systemic food insecurity. Findings highlighted critical tensions between food sovereignty and technological innovation. For example, technology, when not thoughtfully designed, can inadvertently undermine community autonomy and self-determination. This work contributes empirical insights and actionable principles for designing sustainable food systems that align with the values of the food sovereignty movement. Our second goal aimed to develop a framework to address systemic challenges in food provision more broadly, with applications for future research and interventions. Using empirical data and case studies from South Bend, Indiana, and Detroit, Michigan, we identified key factors for sustainable food security efforts. These included understanding community food practices, addressing economic and environmental challenges through sustainable strategies, and considering the social and cultural dynamics of receiving assistance. We also identified research gaps, including addressing stigma around food aid and developing procurement methods that meet the diverse needs of communities. Additionally, we explored how digital technologies could better connect local growers, retailers, and consumers. We recommend that future research focus on addressing these gaps going forward.
Publications
|
Progress 12/01/22 to 11/30/23
Outputs
Target Audience:The FINs project is aimed at supporting communities where access to healthy and affordable food is limited or where there are no grocery stores. The term "food desert" has often been used to describe such communities, however, recent efforts suggest this term is falling out of favor. Nonetheless, these communities are often underserved (low access) and low-income and disproportionately made up of people of color. In addition to actual stakeholders that live in such an area, our project also engages community partners who are our liaisons to actual community members as well as community resources such as food pantries. Changes/Problems:At the end of 2023 the Social Science Team convened to review the above challenges and develop a strategy to more effectively engage with the community in 2024. The first pivot will be to broaden the geographical scope of recruitment to include the neighborhoods south of census tract 27, and possibly west of there as well. Secondly, the Social Science Team will reach out to community partners and social service organizations-not only to reach their clientele who live in the area, but also to promote the app intervention to organizational staff and volunteers. The aim is to increase the likelihood of local user uptake, whether that be among residents of the target neighborhoods, or the health and social service workers who interact with residents and can speak to their needs. Thirdly, the team will revise the recruitment and intervention timeline to allow for multiple user cohorts (likely no more than ten people each) to have individual start dates, thereby increasing the likelihood of reaching our target numbers for recruitment and retention. Finally, the team is looking to hire a Community Engagement Specialist to help with outreach, recruitment, and participant retention. With the above measures in place, we hope to account for and mitigate some of the past year's recruitment challenges, and retain an effective sample size for the intervention this spring. On the software development aspects, the primary challenge has been successful integration of components. The addition of a professional developer to work with our undergraduate and graduate student developers has been very beneficial. We have also run into several technical challenges in accessing reliable data for our recommendation model. (e.g., nutritional informatin and realiable serving sizes). We believe our current solutions are effective. Finally, it was been incredibly challenging to "Categorize" food items from the walmar inventory into a reasonable categorization scheme -- this is necessary to reduce the space over which we search for items to recommend and/or subsitute for other items. Again, we are currently tuning our heuristics and believe they will be suffiicent to produce acceptable recommendations. What opportunities for training and professional development has the project provided?Graduate Student: Annalisa Szymanksi Annalisa was supported on this grant during the entire period. She has developed substantial skills in the application of linear programmign to optimization of a grocery list as well as exciting new skils in understanding how to use large language models to generate acceptable explanations for a food product recommendation. She has engaged deeply with dieticians on this aspect. She has also substantially improved her software development skills in her work integrating the recommendation engine with multiple data sources including the walmart infventor, food data central, and a home-grown hierarchical model of the food inventory space used to reduce our optimization search space. Graduate Student:Brianna Wimer Brianna developed extensive expertise in experiment design and quantitative data analyses. She also engaged deeply with large language models and software development as a contributor to several aspects of this grant work. Graduate Stduent: Oghenemaro Anuyah Oghenemaro "Maro" joined the project in September and has developed a background in linear programming for optimization as well as experience in the application of large language models. Graduate Student: Aarti Israni Aarti joined our U. Michigan team this fall. She is advised by Dr. Tawanna Dillahunt and has led the community engagement and social science components of the study in Detroit. She is mapping the Detroit food system and exploring the role technology might play in leveling theh playing field for these small, distributed food providers. Undergraduate students: Dikko, Heller, Lombardo, and Dvorak These four undergraduate students have spent the past 5 months in deep collaboration with a professional developer, further improving their own software development skills and professional processes. How have the results been disseminated to communities of interest?Four papers have been published in leading conferences and journals. What do you plan to do during the next reporting period to accomplish the goals?The final reporting period will include the major components: 1. Completion of the app RQ6 2. Execution of the intervention experiment -- including data analyses RQ1, RQ2, RQ3 3. Execution of the study in Detroit, MI to explore the integration of small food providers into the larger food data ecosystem. RQ1, RQ2 4. Exploration into the use of Large Language Models for processing ethnographic data and developing heterogenous network models. RQ5
Impacts
What was accomplished under these goals?
From the social science perspective, particulalry with regard to RQ2, RQ6, and RQ7: Notre Dame Team - South Bend: The FINs team conducted a number of community engagement activities over the course of 2023 with the goal of gauging the level of public interest and participation in the app intervention, among other community-based programs. For example, early in 2023 the FINs team convened a public forum to solicit local residents' input on events and education opportunities that could alleviate food insecurity in the neighborhood of concern. Though interest levels at the forum were high, and numerous program ideas were proposed, the FINs Social Science team encountered various challenges in 2023 toward sustaining engagement over the long term. As outlined below, these challenges included resident transience, low levels of participation by the board members of local neighborhood associations, and competing demands on parents and heads of households from the target demographic in the recruitment area. Going into 2024, we will pivot our engagement strategy to better account for some of these challenges, with an aim toward ensuring recruitment and retention for the FINs app intervention. For most of 2023 the FINs Social Science Team worked mainly with one neighborhood near census tract 27 to plan community events, each of which was viewed as an opportunity to promote the app study and distribute flyers and interest forms for future follow-up. In addition to the neighborhood forum held in April, the team also participated in a back-to-school fair, a holiday-themed event, and an information session on affordable internet connections for low-income households. Whereas the back-to-school fair was very well attended (there were about 100 families that participated, 40 of whom gave their contact information to neighborhood association representatives), the remaining events did not garner as much interest. In between events, the team also worked with the neighborhood association to follow up with residents who had expressed interest in community programs. However, resident response rates to this outreach were inconsistent, mirroring our attempts to follow up with transient participants from previous rounds of data collection. Finally, several board members of the neighborhood association-some of whom had access to resident contact lists and the email listserv-stopped attending meetings, and we were unable to get their input on next steps. U. Michigan Team - Detroit: The Detroit-FINS Team reviewed scholarship spanning disciplines, including (but not limited to) community development, agriculture, nutrition, public health, and human-computer interaction, to investigate factors to consider when implementing a food hub in food-insecure areas. Our work, which we intend to submit for publication in the next month, addresses questions such as how research can be designed to support the development of a food hub, what existing barriers and constraints individuals face in food-insecure areas, and what are opportunities for technology to address open challenges. We identify several open research challenges based on an initial pilot in South Bend, Indiana. We contribute lessons learned from our pilot and consider Detroit as a future case study to assess and build upon our findings, given Detroit's food system has been cited as an economically promising industry. These insights include the importance of addressing accessibility and amplifying existing community values around food, addressing barriers to participation, and considering aspects of location that can influence food access and consumption including the availability of food sources, how and whether spaces are utilized and trusted by community members, and access to transportation. We contribute opportunities for technology to support the development of food hubs in food-insecure areas. In addition, we began mapping the food ecosystem in Detroit by identifying key stakeholders and relevant food programs and organizations. Our goal is to understand the network of individuals and organizations working towards food sovereignty, understand their previous work and current initiatives, and identify possible gaps that information and technology could help address. We will use this initial mapping to then guide further research inquiry around our motivating questions of "What does it mean to improve Detroit Food Systems (or a food system in general?)" and "How could a cooperative food delivery model benefit Detroit?". From the technical perspective, with regard to RQ4 and RQ6: The Technical Team continued to take the non-functional FIGMA prototype and develop it into a functional prototype application. The app is roughly 80% functional at this time. The team, consisting of 4 primary undergraduate student developers, was combined with a professional developer in the final 4 months of the reporting period to primarily integrate the various components of the app. This included development and integration of the front end application components, the back end database storage, the back end interaction analytics, components to integrate with and access food inventory from Walmart, components to integrate with the Walmart cart system for purchasing, and finally, the back end recommendation system components. The recommendation system is also approximately 80% complete. This aspect of the project is led by two graduate students and has also benefited from multiple undergraduate student input. The recommendation engine required significant software development to implement the linear programming optimization algorithm, multiple components to compute nutritional values and portion sizes, and a fairly complex component to create a hierarchical organization of the Walmart food inventory as well as an algorithm for mapping any food product to the hierarchy. Finally, the team developed an interaction approach to accept user input in the form of a generic grocery list item, map that item to the hierarchy, generate a primary recommendation, and generate multiple alternative recommendations based on the food hierarchy as well as several strategies for product substitution in support of user goals (e.g., financial goals, health goals) In addition to the app development, the technical team carried out several research projects to explore interactive digital food product labels, the use of large language models for generating food product recommendation explanations, and a preliminary optimization model to explore the efficacy of government assistance in support of healthy eating on a budget. Each of these is explained in much more detail in the publication products for this reporting period.
Publications
- Type:
Conference Papers and Presentations
Status:
Accepted
Year Published:
2024
Citation:
Szymanski, A., Eicher-Miller, H., Anuyah, O., Wimer, B., Metoyer, R., "Integrating Expertise in LLMs: Crafting a Customized Nutrition Assistant with Refined Template Instructions" To Appear in ACM SIGCHI 2024
- Type:
Conference Papers and Presentations
Status:
Accepted
Year Published:
2024
Citation:
Wimer, B., Szymanski, A., Metoyer, R., "Beyond Static Labels: Unpacking Nutrition Comprehension in the Digital Age", To Appear in ACM SIGCHI 2024
- Type:
Journal Articles
Status:
Published
Year Published:
2023
Citation:
Germino, J., Szymanski, A., Metoyer, R., & Chawla, N. V. (2023). A community focused approach toward making healthy and affordable daily diet recommendations. Frontiers in big Data, 6.
|
Progress 12/01/21 to 11/30/22
Outputs
Target Audience:The FINs project is aimed at supporting communities where access to healthy and affordable food is limited or where there are no grocery stores. The term "food desert" has often been used to describe such communities, however, recent efforts suggest this term is falling out of favor. Nonetheless, these communities are often underserved (low access) and low-income and disproportionately made up of people of color. In addition to actual stakeholders that live in such an area, our project also engages community partners who are our liaisons to actual community members as well as community resources such as food pantries. In year 2of the project, we engaged both community partners and community members directly through continued ethnographic research as well as usability studies with potential end-users of the app under development. We also ran a pilot experiment to better understand the feasibility of a hub model for physical access to food. During this experiment, we engaged with community members who served as participants and with a local community partner, Our Lady of Hungary Church, which served as the hub location. We also partnered with Cultivate Culinary, our local food rescue organization,to provide delivery to the hub location. In Detroit, MI, we engaged multiple community partner organizations to explore the potential of a hub delivery model. Key partnerships included Brilliant Detroit as well as engagement with local community food access expert, Alex Hill. Changes/Problems:The most significant change/pivot is with respect to our exploration of the physical access challenges. As mentioned previously, our pre-pilot study and ethnographic data suggest that COVID-19 has significantly changed the physical access landscape. In South Bend, communities have leveraged their local social networks to overcome physical access barriers and in addition, delivery to the home has been normalized and in some cases, even subsidized. In Detroit, the geographic properties are much different with neighborhoods being much more spread (distance-wise) and grocery stores being located much farther on the margins. As a result, neighborhoods rely much more heavily on local food source providers that are typically not large commercial chains. These findings lead us to believe that the planned "physical access" part of the intervention is likely inappropriate for either South Bend or Detroit. Rather than moving forward with an implementation of the hub delivery model, we propose a pivot to explore specific strategies in the two locations. In South Bend, we propose to explore the boundaries of a hub-delivery model to understand when such models might "work" vs. "not work" and for which population subgroups. This will involve a deeper literature review of hub delivery models and contexts in which they are successful and how technology use factors can factor into those models. In Detroit (and South Bend), we propose to pivot to an exploration of how to better integrate "local non-commercial" food providers/sources (e.g., pantries, community gardens, farmers' markets) into the information landscape so that they can be integrated into apps such as the one we are developing. More specifically, these sources pose a challenge primarily around real-time tracking of inventory. If we can develop approaches to accurately track inventory, we can explore how to include these sources as providers in the proposed app. What opportunities for training and professional development has the project provided?One Graduate student for the full year - this student led the modeling and machine learning effort and has been mentored/advised by Dr. Chawla and Dr. Metoyer Two Graduate students since Fall 2021 - these students have participated in design and development activities for the app. One student was responsible for leading the Mechanical Turk study to explore narrative visualization for food label comprehension. The second student was responsible for leading the effort to design and implement an optimization model for grocery recommendation. They have both been advised by Dr. Metoyer Three undergraduate summer interns - This team developed a survey instrument to confirm the initial findings of our ethnographic work regarding the 3 archetype shoppers that we found in our data (inventory shoppers, menu shoppers, and emergency shoppers). Four undergraduate students during the academic year (College of Engineering) - These students led the app design and development efforts . They were advised by Dr. Metoyer One undergraduate design student -- this student led the data collection and analysis efforts for the phase 2 observation data. One graduate student (Brianna Wimer), one research personnel (Michelle Sawaan), and one community partner (Jim Conklin of Cultivate Culinary) attended the National Science Foundation Smart and Connected Communities PI Meeting along with the PI - Ron Metoyer. Graduate students Szymanski and Wimer attended CHI 2022. How have the results been disseminated to communities of interest?Two papers on recipe modeling and recommendation efforts were published and disseminated through the IJCAI 2022 conference. Two papers on the ethnographic study and nutrition label comprehension are currently under review for CHI 2023. Another paper on the design and development of an optimization model in support of recommendations during grocery planning is under review bythe Frontiers of Big Data journal. What do you plan to do during the next reporting period to accomplish the goals?Over the next reporting period, we have two primary focuses. First, we will iterate over the app design and complete the implementation. We will conduct usability studies throughout with potential end users. These activities will round out the completion of Objective 3. Second, given our new understanding of the physical access challenges in our community (or lack thereof), we will pivot to a home delivery model for our intervention and explore sustainability models in South Bend. In Detroit, as mentioned above, we have uncovered a significant difference in the food landscape that leads us to believe that the app will likely not be effective given that it currently requires a grocery partner with an online purchasing API. The opportunity in Detroit is around local, small providers. In this next reporting period, we will better understand this local food landscape and develop requirements for integrating these more informal food sources into the app. Third, we will finalize our optimization-based recommendation algorithm and integrate it with the user app to complete a minimum viable implementation of the app for use in the intervention.
Impacts
What was accomplished under these goals?
Objective 1 is complete: We ran the phase 1 study of ethnographic interviews collecting notes and audio data that have been transcribed. We also completed Phase 2 observations. We have completed an initial analysis of the Phase 2 data, but will likely revisit it for further analyses. We have also largely answered RQ1 and are using the results to inform the design of the app as well as the primary focus on a new publication under development. RQ2 is currently under exploration in both South Bend and Detroit. RQ2 and RQ3 will continue to be addressed in the following years. Objective 2 is partially complete (75%). We aggregated the data necessary to construct a comprehensive network model of recipes, ingredients, and users and demonstrated the effectiveness of this model in recipe recommendation tasks (RQ4). This year, we will work on RQ5to explore how to integrate thick qualitative data from our interviews and observations more directly into our graph network models for recommendation purposes. Objective 3 is partially complete (50%). We have a fairly complete version of the app design -- RQ6(80%) and have implemented approximately 50% of that design. Objectives 4 and 5: We have been in communication with multiple community partners to identify the most appropriate hub location and have explored the logistics for how to operate the delivery hub. Our pilot study and ethnographic data suggest that physicalaccess is actually much less of a problem than initially expected given the use (in South Bend) of local social networks to overcome the physical access barriers. In Detroit, the geographic properties are much different with neighborhoods being much more spread (distance-wise) and grocery stores being located much further on the margins. As a result, neighborhoods rely much more heavily on local food source providers that are typically not large commercial chains. We will expand on this later in "changes/pivots". Finally, our external evaluator has produced a fairly comprehensive report on our project that also discusses our process, training/advising, as well as the proposed "pivot". You can find theexternal evaluation report here.
Publications
- Type:
Conference Papers and Presentations
Status:
Accepted
Year Published:
2022
Citation:
Tian, Y., Zhang, C., Guo, Z., Ma, Y., Metoyer, R., & Chawla, N. V. (2022). Recipe2Vec: Multi-modal Recipe Representation Learning with Graph Neural Networks. IJCAI 2022
- Type:
Conference Papers and Presentations
Status:
Accepted
Year Published:
2022
Citation:
Tian, Y., Zhang, C., Guo, Z., Huang, C., Metoyer, R., & Chawla, N. V. (2022). RecipeRec: A Heterogeneous Graph Learning Model for Recipe Recommendation. IJCAI 2022
- Type:
Conference Papers and Presentations
Status:
Under Review
Year Published:
2023
Citation:
Dillahunt, T., Sawwan, M., Wood, D., Wimer, B., Conrado, A., Eicher-Miller, H., Zornig, A., and Metoyer, R., Understanding Food Planning Strategies of Food Insecure Populations: Implications for Food Agentic Technologies, Submitted to CHI 2023
- Type:
Conference Papers and Presentations
Status:
Under Review
Year Published:
2023
Citation:
Wimer, B., Szymanski, A., Lu, Y., and Metoyer, R., Exploring the Use of Narrative Visualization for Nutrition Label Comprehension, Submitted to CHI 2023
- Type:
Journal Articles
Status:
Under Review
Year Published:
2023
Citation:
Joe Germino, Annalisa Szymanski, Ronald Metoyer and Nitesh V Chawla, A Community Focused Approach Towards Making Healthy and Affordable Daily Diet Recommendations, Submitted to Frontiers in Big Data
|
Progress 12/01/20 to 11/30/21
Outputs
Target Audience:The FINs project is aimed at supporting communities where where access to healthy and affordable food is limited or where there are no grocery stores. The term "food desert" has often been used to describe such communities, however, recent efforts suggest this term is falling out of favor. Nonetheless, these communities are often underserved (low access) and low-income and disproportionately made up of people of color. In addition to actual stakeholders that live in such an area, our project also engages community partners who are our liaisons to actual community members as well as community resources such as food pantries. In year 1 of the project, we engaged both community partners and community members directly through our ethnographic research. Changes/Problems:Pivot: Interview Process -- As one would expect, our Phase I early ethnographic interviews, while intended to be in person, were carried out remotely through zoom due to COVID19 precautions and restrictions. Pivot: Participant Recruiting -- We broadened the boundaries of participation in the study a bit beyond our original recruitment area (Census tract 27 - roughly the Rum Villiage neighborhood in South Bend) in order to allow for important demographic representation that we were not able to achieve in our initial recruitment. We engaged a small handful of participants outside the original neighborhood as a cross-check for any neighborhood biases. Pivot: Grocer Selection --One problem we have encountered is the varying policies with local grocers for how P-EBT (Food Stamps) can be used. Most stores will allow the cards to be used for payment online, however, they also require that the user be present, in person, to show the card and pick up the groceries. This does not work for our model since our intention was to allow our participants to purchase online for a third party to pick up for delivery to the hub. We believe we have found a solution in Walmart and are currently investigating the feasibility of using Walmart as our grocery delivery provider. What opportunities for training and professional development has the project provided? One Graduate student for the full year - this student has led the modeling and machine learning effort and been mentored/advised by Dr. Chawla and Dr. Metoyer Two Graduate students since Fall 2021 - these students have participated in design and development activities for the app. They have been advised by Dr. Metoyer Three undergraduate summer interns - This team conducted a sustainability study for the hub delivery model and developed initial design ideas for the app. They were advised by Dr. Metoyer and Alisa Gura. Four undergraduate students during the academic year (College of Engineering) - These two students led the initial development efforts to explore the technical stack options for the development of the app. They also began development of the app framework. They were advised by Dr. Metoyer Three undergraduate students (College of Arts and Letters) - These students participated in ethnography data collection as well as data transcription and analyses. How have the results been disseminated to communities of interest?Two papers have been published on the modeling efforts - one at ACM International Conference on Information & Knowledge Management and another in the Frontiers in Big Data journal. Tian, Y., Zhang, C., Metoyer, R., & Chawla, N. V. (2021, October). Recipe representation learning with networks. In Proceedings of the 30th ACM International Conference on Information & Knowledge Management (pp. 1824-1833). Tian, Y., Zhang, C., Metoyer, R., & Chawla, N. V. (2022). Recipe Recommendation with Hierarchical Graph Attention Network. Frontiers in Big Data, 123. The ethnographic research analyses have been synthesized into several artifacts including a slide deck presentation of the emerging themes, three personas of archetype app users, and three use-case scenarios. The slide deck presentation of the results/themes was shared with our community partners in a meeting of the advisory board. What do you plan to do during the next reporting period to accomplish the goals?Over the next reporting period, we will complete Objective 1 by wrapping up the phase 2 observations. We will likely aim to publish these results in a social science venue. We will also complete the design activities for the application. This design will be informed by ethnographic research and will result in a prototype application design. We are in the process of writing a paper for submission tothe CHI2023 conference. We also intend to have a first version partially functional prototype of the application interface by the end of the second reporting period. In addition, we will develop a working prototype of the recommendation algorithm that will serve as the primary back-end computation for the app. Finally, we plan to have the hub location finalized and the concept for the hub operation fully fleshed out and personnel identified for operating the hub.
Impacts
What was accomplished under these goals?
Objective 1 is nearly complete: We ran the phase 1 study of ethnographic interviews collecting notes and audio data that has been transcribed. We are currently finishing up Phase 2 observations. Objective 2 is partiallycomplete. We have aggregated much of the data necessary to construct a comprehensive network model of recipes, ingredients, and users. Objective 5: We have been in communication with multiple community partners to identify the most appropriate hub location and have explored the logistics for how to operate the delivery hub.
Publications
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2021
Citation:
Tian, Y., Zhang, C., Metoyer, R., & Chawla, N. V. (2021, October). Recipe representation learning with networks. In Proceedings of the 30th ACM International Conference on Information & Knowledge Management (pp. 1824-1833).
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2022
Citation:
Tian, Y., Zhang, C., Metoyer, R., & Chawla, N. V. (2022). Recipe Recommendation with Hierarchical Graph Attention Network. Frontiers in Big Data, 123.
|
|