An Artificial Intelligence System for Computer-Assisted Menu Planning

https://doi.org/10.1016/S0002-8223(98)00231-4Get rights and content

Abstract

Planning nutritious and appetizing menus is a complex task that researchers have tried to computerize since the early 1960s. We have attempted to facilitate computer-assisted menu planning by modeling the reasoning an expert dietitian uses to plan menus. Two independent expert systems were built, each designed to plan a daily menu meeting the nutrition needs and personal preferences of an individual client. One system modeled rule-based, or logical, reasoning, whereas the other modeled case-based, or experiential, reasoning. The 2 systems were evaluated and their strengths and weaknesses identified. A hybrid system was built, combining the best of both systems. The hybrid system represents an important step forward because it plans daily menus in accordance with a person's needs and preferences; the Reference Daily Intakes; the Dietary Guidelines for Americans; and accepted aesthetic standards for color, texture, temperature, taste, and variety. Additional work to expand the system's scope and to enhance the user interface will be needed to make it a practical tool. Our system framework could be applied to special-purpose menu planning for patients in medical settings or adapted for institutional use. We conclude that an artificial intelligence approach has practical use for computer-assisted menu planning. J Am Diet Assoc. 1998;98:1009-1014.

Section snippets

Menu Planning Systems

Building any expert system is a collaborative effort between an expert and a “knowledge engineer.” The expert explains the thought processes used to perform some task, and the knowledge engineer models these processes in a computer program. Together, they review the program, refining and validating it, iterating as they learn more about the nature of the task.

We built 2 expert systems, the CAse-based Menu Planner (CAMP)1

System Evaluation

CAMP and PRISM were evaluated by running them on a wide variety of test cases designed to produce different kinds of menus. Over repeated trials, outputs were reviewed and problems were corrected as they were identified. For CAMP, this meant adding new cases to include more types of menus. For PRISM, this meant adding or refining rules. Next, feedback was solicited from practicing dietitians and nutrition students. Both systems were judged to produce useful menus, but they were found to have

The Hybrid System

The hybrid system CAMPER combines CAMP's ability to satisfy nutrition constraints with PRISM's creativity. The system was built by expanding CAMP with rule-based enhancements from PRISM. These enhancements augment CAMP with a “what if” analysis module and a larger, more versatile, database.

CAMPER's database includes more food items than CAMP's and allows the food items to be used in different ways. With CAMP, a food item is viewed only in context within a case. Its database is never queried to

Menu Planning Challenges

We have been investigating the use of artificial intelligence for menu planning since 1988. Our first system, the Expert System on Menu Planning, planned menus for patients on a severely restricted low-protein diet (7). Since then, we have gained insight into the nature of the task of computerized menu planning and what makes it so challenging. Part of the problem appears to be the common sense involved. There's a sense that meals appeal to people, and computers do not share this sense.

Applications

CAMPER is a tool for planning daily menus in accordance with the nutrition needs and personal preferences of individual clients. Although the present version of CAMPER plans menus for healthy adults, the framework and methodology could also apply to planning special-purpose menus for use in many different settings. For example, preplanned menus for metabolic diets in a clinical research center can become a case base, which can then be accessed for menus. Menus revised to meet specific research

References (22)

  • E.F. Eckstein

    Menu planning by computerthe random approach

    J Am Diet Assoc.

    (1967)
  • R.T. Macura et al.

    Case-based reasoningopportunities and applications in health care

    Artif Intell Med.

    (1997)
  • J.L. Balintfy

    Menu planning by computer

    Communications of the ACM.

    (1964)
  • M.C. Spears

    Foodservice OrganizationsA Managerial and Systems Approach

    (1995)
  • E.F. Eckstein

    Menu Planning

    (1978)
  • E. Rich et al.

    Artificial Intelligence

    (1991)
  • J. Kolodner

    Case-Based Reasoning

    (1993)
  • N. Yang

    An Expert System on Menu Planning [master's thesis]

    (1989)
  • R.R. Hinrichs

    Problem Solving in Open WorldsA Case Study in Design

    (1992)
  • Ganeshan K, Farmer J. Menu planning system for a large catering corporation. In: Proceedings of the Third International...
  • C.R. Marling

    Integrating Case-Based and Rule-Based Reasoning in Knowledge-Based Systems Development [dissertation]

    (1996)
  • Cited by (33)

    • Multimodal hybrid reasoning methodology for personalized wellbeing services

      2016, Computers in Biology and Medicine
      Citation Excerpt :

      If RBR fails, CBR is used [15]. In the second strategy, CBR is used for the master reasoning process and RBR is used to refine the decision [16]. An example of this strategy is reasoning system for diabetes management [17].

    • New product development for the healthcare industry: A case study of diet software

      2012, Health Policy and Technology
      Citation Excerpt :

      There are several nutrition software programs that have different capabilities that vary from basic features to complex applications. As indicated by Petot et al. [30] managing nutrition and energy intake planning have been a focal research area in the literature since early 1960s. Accordingly, Balintfy [31] has employed linear programming for optimization of diets with respect to variables which are namely nutritional balance, cost and satisfaction.

    • Food Composition Data: The Foundation of Dietetic Practice and Research

      2007, Journal of the American Dietetic Association
      Citation Excerpt :

      A variety of multiuse software applications are available for personal computers (and personal digital assistants) as are programs tailored to the needs of specific practice areas such as school nutrition, clinical research, institutional and restaurant foodservice, food labeling, and education of practitioners (16,64). Novel software applications, including smart cards, personal digital assistants, and artificial intelligence (65-72) are being developed for a variety of practice settings. Despite this potential availability of data and software, not everyone has access to or uses the available technology (73-75).

    View all citing articles on Scopus
    View full text