In the AI-Talian Application, we roam with you the hallways of AI in all industries; big and small. Because today, where you most and least expect it, is doing wonders.
Check the full series here for more!
Online food recipes are quick and yummy, but they’re not always accurate.
How many times have you followed an online recipe and found out a bit too late that the measurements weren’t really on point?
With AI breaking into every single industry, AI experts made sure to solve that problem for us like they solve oh-so-many others.
About the project
Using language models and recipe modifiers, RobotChef modifies recipes based on user reviews.
Imagine how when someone goes: “oh, that’s too much butter for this recipe”, the recipe automatically adjusts itself (literally!) to reduce the measurements for the butter.
How does it work?
As we mentioned, the code consists of two main components: the language model and the recipe modifier, and we’ll be discussing both in addition to the baseline.
1. The Language Models
The code in language_model/ can be used to train an RNN-based language model as well as to score review segments.
Running language_model/evaluate_lm.py will score a hand-labeled evaluation set, plot an ROC curve and display the best F1 score along with the threshold used to achieve it.
2. Recipe Modifier
The code in recipe-modifier/ can be used to train a network that will, given a recipe and a refinement, score each index in the recipe with how likely the refinement refers to said index.
The model can be found at recipe-modifier/new_recipe_net.py.
3. Baselines
baselines.py computes some baseline scores for the task of index prediction given a refinement and a recipe. The baseline models use a simple word-count-vector representation for segments and a few basic distance metrics to find the best scoring index. We show the Top-1 and Top-3 error rates for two tasks:
- identification of index for a modification refinement
- identification of index for an insertion refinement
The potential of AI is something we can’t deny, but the hiring process is always a pain. LyRise helps you hire your perfect AI talent in less than two weeks so nothing stands in the way of your big plans.
Get top-notch talent from our niche pool in less than 14 days; all you have to do is tell us what you need here.
