CN-121999983-A - Menu generation method and system based on deep learning model

CN121999983ACN 121999983 ACN121999983 ACN 121999983ACN-121999983-A

Abstract

The application belongs to the technical field of intelligent equipment, and discloses a menu generation method and a system based on a deep learning model, which are used for preprocessing the existing menu in the existing menu set, classifying the preprocessed existing menu by constructing a menu database, and guaranteeing the simplification of cooking steps and the reliability of data; the method comprises the steps of generating a menu database, generating a new menu, generating a deep learning model, carrying out expansion processing on the menu database, ensuring the data diversity and individuation of the menu database by generating the new menu, providing richer data support for the deep learning model when the deep learning model is trained by combining the menu database, identifying food material data and seasoning data from user information, obtaining a target menu with higher accuracy and individuation by inputting the food material data and the seasoning data into the deep learning model, and ensuring the rationality of each step in the target menu by utilizing the food material data and the seasoning data, thereby meeting the individuation cooking requirements of users.

Inventors

SHANG FEI
REN FUJIA

Assignees

杭州老板电器股份有限公司
成都老板创新科技有限公司

Dates

Publication Date: 20260508
Application Date: 20241101

Claims (10)

1. A menu generation method based on a deep learning model is characterized by comprising the following steps: Preprocessing each existing menu in an existing menu set, and constructing a menu database based on each preprocessed existing menu, wherein the existing menu set comprises at least two existing menus; Performing expansion processing on the menu database, and training a deep learning model based on the menu database after the expansion processing; And identifying food material data and seasoning data from the acquired user information, and inputting the food material data and the seasoning data into the trained deep learning model to obtain a target menu.
2. The method of claim 1, further comprising, prior to said preprocessing each existing recipe in the set of existing recipes: Collecting at least three menu data corresponding to three data types, wherein each data type corresponds to at least one menu data; when the data type of any one menu data is the first type, capturing a corresponding existing menu from the menu data based on a preset crawler algorithm; When the data type of any one menu data is the second type, extracting a corresponding existing menu from the menu data based on a preset interface script; When the data type of any one menu data is the third type, scanning out a corresponding existing menu from the menu data based on a preset character recognition engine; all the existing recipes are combined into an existing recipe set.
3. The method of claim 1, further comprising, prior to said preprocessing each existing recipe in the set of existing recipes: When the language type of any one existing menu is identified in the existing menu set and is inconsistent with the preset language type, determining characteristic data in the existing menu; Performing translation processing on the feature data, and inputting the feature data after the translation processing into a natural language processing model to obtain the feature data after the language processing; and generating a first menu based on the feature data after language processing, and updating the existing menu into the first menu.
4. The method of claim 1, wherein the preprocessing each existing recipe in the set of existing recipes comprises: identifying food material data and step data of each existing menu in the existing menu set, and calculating a corresponding hash value according to the food material data and the step data; When the fact that the name expressions of any two existing menus are consistent is detected in the existing menu set, calculating the similarity between hash values corresponding to the two existing menus respectively; And when the similarity exceeds a preset similarity threshold, eliminating any one of the two existing menus.
5. The method of claim 4, wherein the preprocessing each existing recipe in the set of existing recipes further comprises: When the absence of the numerical value of the food material data of any one of the existing recipes is detected in the existing recipe set after the elimination processing, determining at least two first similar recipes according to the name expression of the existing recipes; Performing a median calculation on all food material data corresponding to all the first similar menu, performing an interpolation process on the food material data of the existing menu according to a median result, and/or When step missing of step data of any one existing menu is detected in the existing menu set after the elimination processing, determining at least two second similar menus according to the name expression of the existing menu; identifying associated steps corresponding to the missing steps in each second similar menu, and carrying out integration treatment on all the associated steps to obtain filling steps; and filling the step data of the existing menu according to the filling step.
6. The method of claim 1 or 4, wherein said constructing a recipe database based on each of said existing recipes after preprocessing comprises: Identifying name expressions, seasoning data, food material data and step data from each pre-processed existing menu, and respectively classifying all the name expressions, all the seasoning data, all the food material data and all the step data; And constructing a menu database based on all the name expressions, all the seasoning data, all the food material data and all the step data after the classification processing.
7. The method of claim 6, wherein the expanding the recipe database comprises: Selecting at least two food material data, the seasoning data corresponding to each food material data and the step data from the menu database, and sorting all the step data according to a preset step rule; Inputting all the step data after the sorting treatment to a decision tree learning model, predicting a first reasonable value, and determining the name expression of a generated menu according to all the food material data when the first reasonable value is detected to exceed a preset first threshold value; And expanding the name expression, all the food material data, all the seasoning data and all the step data after the sorting processing corresponding to the generated menu in the menu database.
8. The method of claim 7, wherein the expanding the recipe database further comprises: Selecting any one of the step data corresponding to the name expression from the menu database, and determining a dependency relationship value between any two adjacent steps in the step data; When the dependency relationship value between any two adjacent steps is lower than a preset second threshold value, increasing the step data, and inputting the step data after the increasing treatment to the decision tree learning model to obtain a second reasonable value; And when the second reasonable value is detected to exceed the preset first threshold value, adding the processed step data, the name expression corresponding to the processed step data, the food material data and the seasoning data into the menu database.
9. The method of claim 1, wherein the training the deep learning model based on the augmented recipe database comprises: carrying out serialization processing on each text data in the menu database after expansion processing to obtain a corresponding text sequence; When the length of any one text sequence exceeds a preset third threshold value, cutting off the text sequence; when the length of any one text sequence is lower than the preset third threshold value, filling the text sequence; training the deep learning model based on the menu database after text data processing.
10. A deep learning model-based recipe generation system, comprising: the system comprises a data construction module, a data analysis module and a data analysis module, wherein the data construction module is used for preprocessing each existing menu in an existing menu set and constructing a menu database based on each preprocessed existing menu; The model training module is used for carrying out expansion processing on the menu database and training a deep learning model based on the menu database after expansion processing; And the menu generation module is used for identifying food material data and seasoning data from the acquired user information, and inputting the food material data and the seasoning data into the trained deep learning model to obtain a target menu.

Description

Menu generation method and system based on deep learning model Technical Field The application belongs to the technical field of intelligent equipment, and particularly relates to a menu generation method and system based on a deep learning model. Background With the acceleration of the pace of social life and the increase of the demand for healthy diet, the health degree and the manufacturing mode of the existing takeaway cannot meet the health demands of people, and more consumers begin to pay attention to their own eating habits, namely, make healthier foods by themselves. However, for many novice cooking, even if the novice cooking has food materials, creative and time are lacking to make a diet plan for achieving a good health, common processing methods include menu review and model prediction, wherein a manual menu review or web-search menu can provide a certain menu option, but the selection is difficult and the cooking steps are complicated, and the traditional model prediction method can avoid the selection difficulty, but the predicted menu still has the problems of unreasonable steps and the like, so that accurate diet selection cannot be provided. Disclosure of Invention The application provides a menu generation method and a system based on a deep learning model, which aims to solve the technical problems that a menu selection can be provided by manually consulting the menu or searching the menu through a network, but the selection is difficult, the cooking steps are complicated and the like, the selection is difficult can be avoided by a traditional model prediction mode, but the predicted menu still has the problems of unreasonable steps and the like, and further the accurate diet selection cannot be provided, and the like, and the technical scheme is as follows: In a first aspect, an embodiment of the present application provides a method for generating a menu based on a deep learning model, including: Preprocessing each existing menu in the existing menu set, and constructing a menu database based on each preprocessed existing menu, wherein the existing menu set comprises at least two existing menus; performing expansion processing on the menu database, and training the deep learning model based on the menu database after the expansion processing; and identifying food material data and seasoning data from the acquired user information, and inputting the food material data and the seasoning data into the trained deep learning model to obtain a target menu. In an alternative aspect of the first aspect, before preprocessing each existing recipe in the set of existing recipes, the method further comprises: Collecting at least three menu data corresponding to three data types, wherein each data type corresponds to at least one menu data; when the data type of any one menu data is the first type, capturing a corresponding existing menu from the menu data based on a preset crawler algorithm; When the data type of any one menu data is the second type, extracting a corresponding existing menu from the menu data based on a preset interface script; When the data type of any one menu data is the third type, scanning out a corresponding existing menu from the menu data based on a preset character recognition engine; All existing recipes are merged into an existing recipe set. In a further alternative of the first aspect, prior to preprocessing each existing recipe in the set of existing recipes, further comprising: When the language type of any one existing menu is identified in the existing menu set and is inconsistent with the preset language type, determining characteristic data in the existing menu; performing translation processing on the feature data, and inputting the feature data after the translation processing into a natural language processing model to obtain feature data after language processing; and generating a first menu based on the language processed characteristic data, and updating the existing menu into the first menu. In yet another alternative of the first aspect, preprocessing each existing recipe in the set of existing recipes includes: Identifying food material data and step data of each existing menu in the existing menu set, and calculating a corresponding hash value according to the food material data and the step data; when the fact that the name expressions of any two existing recipes are consistent is detected in the existing recipe set, calculating the similarity between hash values corresponding to the two existing recipes respectively; when the similarity exceeds a preset similarity threshold, any one of the two existing recipes is removed. In a further alternative of the first aspect, the preprocessing of each existing recipe in the set of existing recipes further comprises: When the numerical value of the food material data of any one existing menu is detected to be missing in the existing menu set after the elimination processing, determining at least two first simi