HannaAbiAkl
commited on
Commit
•
1d1d41c
1
Parent(s):
135435b
Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,80 @@
|
|
1 |
-
---
|
2 |
-
license: mit
|
3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: mit
|
3 |
+
language:
|
4 |
+
- en
|
5 |
+
metrics:
|
6 |
+
- accuracy
|
7 |
+
- f1
|
8 |
+
- recall
|
9 |
+
- precision
|
10 |
+
library_name: transformers
|
11 |
+
pipeline_tag: text-generation
|
12 |
+
---
|
13 |
+
|
14 |
+
# FLAN-T5 small-WordNet
|
15 |
+
|
16 |
+
This model is a fine-tuned version of [flan-t5-small](https://huggingface.co/google/flan-t5-small) on the WordNet dataset.
|
17 |
+
|
18 |
+
## Model description
|
19 |
+
|
20 |
+
The model is trained to classify terms into one of four term types: noun, verb, adjective or adverb. The types themselves are learned and then generated by the model with no more than one type associated with a specific term.
|
21 |
+
|
22 |
+
The model also works well as part of a Retrieval-and-Generation (RAG) pipeline by leveraging an external knowledge source, specifically [Wordnet Semantic Primes](https://huggingface.co/datasets/HannaAbiAkl/wordnet-semantic-primes).
|
23 |
+
|
24 |
+
## Intended uses and limitations
|
25 |
+
|
26 |
+
This model is intended to be used to generate a type (class) for an input term.
|
27 |
+
|
28 |
+
# Training and evaluation data
|
29 |
+
|
30 |
+
The training and evaluation data can be found [here](https://github.com/HamedBabaei/LLMs4OL-Challenge-ISWC2024/tree/main/TaskA-Term%20Typing/SubTask%20A.1(FS)%20-%20WordNet).
|
31 |
+
|
32 |
+
The train size is 40559.
|
33 |
+
|
34 |
+
The test size is 9470.
|
35 |
+
|
36 |
+
## Example
|
37 |
+
|
38 |
+
Here's an example of the model capabilities:
|
39 |
+
|
40 |
+
- **input:**
|
41 |
+
- *Lexical Term L:* question
|
42 |
+
- *Sentence Containing L (Optional):* there was a question about my training
|
43 |
+
|
44 |
+
- **output:**
|
45 |
+
- *Type:* noun
|
46 |
+
|
47 |
+
- **input:**
|
48 |
+
- *Lexical Term L:* lodge
|
49 |
+
- *Sentence Containing L (Optional):* Where are you lodging in Paris?
|
50 |
+
|
51 |
+
- **output:**
|
52 |
+
- *Type:* verb
|
53 |
+
|
54 |
+
- **input:**
|
55 |
+
- *Lexical Term L:* genus equisetum
|
56 |
+
- *Sentence Containing L (Optional):*
|
57 |
+
|
58 |
+
- **output:**
|
59 |
+
- *Type:* noun
|
60 |
+
|
61 |
+
## Training procedure
|
62 |
+
|
63 |
+
### Training hyperparameters
|
64 |
+
|
65 |
+
The following hyperparameters were used during training:
|
66 |
+
- learning_rate: 1e-05
|
67 |
+
- train_batch_size: 4
|
68 |
+
- eval_batch_size: 4
|
69 |
+
- seed: 42
|
70 |
+
- num_epochs: 5
|
71 |
+
|
72 |
+
### Training results
|
73 |
+
|
74 |
+
| Training Loss | Epoch | Step | Validation Loss |
|
75 |
+
|:-------------:|:-----:|:----:|:---------------:|
|
76 |
+
| 0.1725 | 1.0 | 1000 | 0.0640 |
|
77 |
+
| 0.1250 | 2.0 | 2000 | 0.0535 |
|
78 |
+
| 0.1040 | 3.0 | 3000 | 0.0469 |
|
79 |
+
| 0.0917 | 4.0 | 4000 | 0.0421 |
|
80 |
+
| 0.0830 | 5.0 | 5000 | 0.0384 |
|