Exposure ‐ models
The FoodEx2 smart coding application is equipped with a series of machine learning models that can suggest terms for each specific problem encountered during the coding in FoodEx2. In particular, three main problems are encounter as follows:
- Find the base term,
- Find the facet categories to enable,
- Find facets for each enabled category.
For each model has been defined the base dataset, on which to train the model, and the problem or objective to solve. Below additional information related with the models released:
- Machine Learning Framework: Spacy
- Model architecture: Simple CNN
| name | textcat type | train/val split | version | date of publication |
epochs | val score | top-3 test accuracy* |
|---|---|---|---|---|---|---|---|
| BT | single-label | 0.9/0.1 0.9/0.1 0.9/0.1 0.85/0.15 |
11.1.0 11.2.0 11.3.0 11.3.1 |
19-04-2021 26-04-2021 16-06-2021 ? |
30 30 30 30 |
80% 77% 78% 89% |
77.98% 78.30% 87.82% 93.30% |
| CAT | multi-label | 0.9/0.1 0.9/0.1 0.9/0.1 0.85/0.15 |
11.1.0 11.2.0 11.3.0 11.3.1 |
19-04-2021 26-04-2021 16-06-2021 ? |
30 30 30 30 |
71% 71% 74% 76% |
46.97% 48.19% 52.29% 57.11% |
| F01 | multi-label | 0.9/0.1 0.9/0.1 0.9/0.1 0.85/0.15 |
11.1.0 11.2.0 11.3.0 11.3.1 |
19-04-2021 26-04-2021 16-06-2021 ? |
21 16 24 16 |
41% 34% 40% 64% |
22.77% 16.64% 32.41% 56.50% |
| F02 | multi-label | 0.9/0.1 0.9/0.1 0.9/0.1 0.85/0.15 |
11.1.0 11.2.0 11.3.0 11.3.1 |
19-04-2021 26-04-2021 16-06-2021 ? |
19 30 20 19 |
58% 52% 51% 79% |
53.97% 50.72% 67.72% 86.39% |
| F03 | multi-label | 0.9/0.1 0.9/0.1 0.9/0.1 0.85/0.15 |
11.1.0 11.2.0 11.3.0 11.3.1 |
19-04-2021 26-04-2021 16-06-2021 ? |
28 30 30 28 |
85% 87% 88% 87% |
58.05% 62.15% 63.80% 73.87% |
| F04 | multi-label | 0.9/0.1 0.9/0.1 0.9/0.1 0.85/0.15 |
11.1.0 11.2.0 11.3.0 11.3.1 |
19-04-2021 26-04-2021 16-06-2021 ? |
30 30 30 30 |
58% 56% 57% 60% |
30.57% 30.57% 32.32% 36.41% |
| F06 | multi-label | 0.9/0.1 0.9/0.1 0.9/0.1 0.85/0.15 |
11.1.0 11.2.0 11.3.0 11.3.1 |
19-04-2021 26-04-2021 16-06-2021 ? |
21 21 23 18 |
67% 67% 69% 75% |
60.28% 61.23% 67.75% 77.86% |
| F07 | multi-label | 0.9/0.1 0.9/0.1 0.9/0.1 0.85/0.15 |
11.1.0 11.2.0 11.3.0 11.3.1 |
19-04-2021 26-04-2021 16-06-2021 ? |
27 27 27 19 |
57% 57% 61% 65% |
53.97% 51.53% 57.93% 69.77% |
| F08 | multi-label | 0.9/0.1 0.9/0.1 0.9/0.1 0.85/0.15 |
11.1.0 11.2.0 11.3.0 11.3.1 |
19-04-2021 26-04-2021 16-06-2021 ? |
30 30 27 25 |
54% 63% 63% 68% |
45.61% 48.50% 51.54% 76.51% |
| F09 | multi-label | 0.9/0.1 0.9/0.1 0.9/0.1 0.85/0.15 |
11.1.0 11.2.0 11.3.0 11.3.1 |
19-04-2021 26-04-2021 16-06-2021 ? |
27 27 26 21 |
65% 65% 64% 68% |
46.00% 46.14% 47.83% 56.23% |
| F10 | multi-label | 0.9/0.1 0.9/0.1 0.9/0.1 0.85/0.15 |
11.1.0 11.2.0 11.3.0 11.3.1 |
19-04-2021 26-04-2021 16-06-2021 ? |
30 30 29 29 |
70% 73% 73% 72% |
70.30% 70.21% 71.57% 76.70% |
| F11 | multi-label | 0.9/0.1 0.9/0.1 0.9/0.1 0.85/0.15 |
11.1.0 11.2.0 11.3.0 11.3.1 |
19-04-2021 26-04-2021 16-06-2021 ? |
1 1 27 1 |
28% 25% 64% 29% |
4.42% 11.24% 60.64% 7.63% |
| F12 | multi-label | 0.9/0.1 0.9/0.1 0.9/0.1 0.85/0.15 |
11.1.0 11.2.0 11.3.0 11.3.1 |
19-04-2021 26-04-2021 16-06-2021 ? |
1 1 30 1 |
24% 35% 84% |
17.36% 31.94% 91.67% 34.72% |
| F17 | multi-label | 0.9/0.1 0.9/0.1 0.9/0.1 0.85/0.15 |
11.1.0 11.2.0 11.3.0 11.3.1 |
19-04-2021 26-04-2021 16-06-2021 ? |
30 30 30 16 |
84% 77% 69% |
39.17% 38.63% 40.03% 64.37% |
| F18 | multi-label | 0.9/0.1 0.9/0.1 0.9/0.1 0.85/0.15 |
11.1.0 11.2.0 11.3.0 11.3.1 |
19-04-2021 26-04-2021 16-06-2021 ? |
30 30 28 26 |
94% 91% 88% |
53.54% 53.39% 58.50% 81.18% |
| F19 | multi-label | 0.9/0.1 0.9/0.1 0.9/0.1 0.85/0.15 |
11.1.0 11.2.0 11.3.0 11.3.1 |
19-04-2021 26-04-2021 16-06-2021 ? |
30 29 29 30 |
76% 71% 87% |
59.19% 60.39% 64.13% 74.04% |
| F20 | multi-label | 0.9/0.1 0.9/0.1 0.9/0.1 0.85/0.15 |
11.1.0 11.2.0 11.3.0 11.3.1 |
19-04-2021 26-04-2021 16-06-2021 ? |
27 27 23 29 |
59% 50% 60% |
50.94% 47.92% 58.23% 79.21% |
| F21 | multi-label | 0.9/0.1 0.9/0.1 0.9/0.1 0.85/0.15 |
11.1.0 11.2.0 11.3.0 11.3.1 |
19-04-2021 26-04-2021 16-06-2021 ? |
22 22 30 28 |
65% 78% 72% |
52.24% 49.81% 71.64% 92.46% |
| F22 | multi-label | 0.9/0.1 0.9/0.1 0.9/0.1 0.85/0.15 |
11.1.0 11.2.0 11.3.0 11.3.1 |
19-04-2021 26-04-2021 16-06-2021 ? |
29 29 30 26 |
83% 76% 66% |
49.70% 51.58% 53.00% 61.17% |
| F23 | multi-label | 0.9/0.1 0.9/0.1 0.9/0.1 0.85/0.15 |
11.1.0 11.2.0 11.3.0 11.3.1 |
19-04-2021 26-04-2021 16-06-2021 ? |
24 24 29 25 |
87% 84% 79% |
94.60% 94.87% 97.12% 98.93% |
| F24 | multi-label | 0.9/0.1 0.9/0.1 0.9/0.1 0.85/0.15 |
11.1.0 11.2.0 11.3.0 11.3.1 |
19-04-2021 26-04-2021 16-06-2021 ? |
30 30 29 28 |
97% 90% 92% |
87.89% 83.24% 98.41% 99.38% |
| F25 | multi-label | 0.9/0.1 0.9/0.1 0.9/0.1 0.85/0.15 |
11.1.0 11.2.0 11.3.0 11.3.1 |
19-04-2021 26-04-2021 16-06-2021 ? |
18 18 18 23 |
100% 100% 100% |
96.20% 95.32% 96.20% 99.12% |
| F26 | multi-label | 0.9/0.1 0.9/0.1 0.9/0.1 0.85/0.15 |
11.1.0 11.2.0 11.3.0 11.3.1 |
19-04-2021 26-04-2021 16-06-2021 ? |
30 30 30 28 |
99% 99% 99% |
89.56% 88.87% 95.42% 96.15% |
| F27 | multi-label | 0.9/0.1 0.9/0.1 0.9/0.1 0.85/0.15 |
11.1.0 11.2.0 11.3.0 11.3.1 |
19-04-2021 26-04-2021 16-06-2021 ? |
29 29 27 24 |
49% 47% 51% |
50.52% 50.09% 57.84% 59.66% |
| F28 | multi-label | 0.9/0.1 0.9/0.1 0.9/0.1 0.85/0.15 |
11.1.0 11.2.0 11.3.0 11.3.1 |
19-04-2021 26-04-2021 16-06-2021 ? |
30 30 30 30 |
72% 71% 72% |
36.62% 37.14% 43.87% 51.62% |
| F29 | multi-label | 0.9/0.1 0.9/0.1 0.9/0.1 0.85/0.15 |
11.1.0 11.2.0 11.3.0 11.3.1 |
19-04-2021 26-04-2021 16-06-2021 ? |
n.a. n.a. 1 1 |
n.a. n.a. 52% |
n.a. n.a. 86.67% 100.0% |
| F30 | multi-label | 0.9/0.1 0.9/0.1 0.9/0.1 0.85/0.15 |
11.1.0 11.2.0 11.3.0 11.3.1 |
19-04-2021 26-04-2021 16-06-2021 ? |
n.a. n.a. 1 1 |
n.a. n.a. 94% |
n.a. n.a. 66.67% 75.00% |
| F31 | multi-label | 0.9/0.1 0.9/0.1 0.9/0.1 0.85/0.15 |
11.1.0 11.2.0 11.3.0 11.3.1 |
19-04-2021 26-04-2021 16-06-2021 ? |
30 22 21 22 |
98% 95% 75% |
61.98% 59.38% 89.84% 97.14% |
| F32 | multi-label | 0.9/0.1 0.9/0.1 0.9/0.1 0.85/0.15 |
11.1.0 11.2.0 11.3.0 11.3.1 |
19-04-2021 26-04-2021 16-06-2021 ? |
28 21 16 17 |
100% 99% 82% |
92.42% 93.27% 99.66% 99.83% |
| F33 | multi-label | 0.9/0.1 0.9/0.1 0.9/0.1 0.85/0.15 |
11.1.0 11.2.0 11.3.0 11.3.1 |
19-04-2021 26-04-2021 16-06-2021 ? |
18 17 30 25 |
52% 60% 64% |
21.96% 23.70% 79.32% 90.01% |
* top-3 test accuracy refers to the accuracy value returned by a custom evaluation which attribute a score based on the position of the gold value among the top 3 predictions returned by the model. This value is currently calculated using a non rapresentative test dataset (not all list of classes have examples).