Add tags, restructure info
Browse files
README.md
CHANGED
|
@@ -17,37 +17,6 @@ All the models (downstream tasks) are uncased and trained with whole word maskin
|
|
| 17 |
This task aims to extract named entities in the text, such as names and label with appropriate `NER` classes such as locations, organizations, etc. The datasets used for this task contain sentences that are marked with `IOB` format. In this format, tokens that are not part of an entity are tagged as `”O”` the `”B”`tag corresponds to the first word of an object, and the `”I”` tag corresponds to the rest of the terms of the same entity. Both `”B”` and `”I”` tags are followed by a hyphen (or underscore), followed by the entity category. Therefore, the NER task is a multi-class token classification problem that labels the tokens upon being fed a raw text. There are two primary datasets used in Persian NER, `ARMAN`, and `PEYMA`. In ParsBERT, we prepared ner for both datasets as well as a combination of both datasets.
|
| 18 |
|
| 19 |
|
| 20 |
-
|
| 21 |
-
### PEYMA
|
| 22 |
-
|
| 23 |
-
PEYMA dataset includes 7,145 sentences with a total of 302,530 tokens from which 41,148 tokens are tagged with seven different classes.
|
| 24 |
-
|
| 25 |
-
1. Organization
|
| 26 |
-
2. Money
|
| 27 |
-
3. Location
|
| 28 |
-
4. Date
|
| 29 |
-
5. Time
|
| 30 |
-
6. Person
|
| 31 |
-
7. Percent
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
| Label | # |
|
| 35 |
-
|:------------:|:-----:|
|
| 36 |
-
| Organization | 16964 |
|
| 37 |
-
| Money | 2037 |
|
| 38 |
-
| Location | 8782 |
|
| 39 |
-
| Date | 4259 |
|
| 40 |
-
| Time | 732 |
|
| 41 |
-
| Person | 7675 |
|
| 42 |
-
| Percent | 699 |
|
| 43 |
-
|
| 44 |
-
|
| 45 |
-
|
| 46 |
-
**Download**
|
| 47 |
-
You can download the dataset from [here](http://nsurl.org/tasks/task-7-named-entity-recognition-ner-for-farsi/)
|
| 48 |
-
|
| 49 |
-
---
|
| 50 |
-
|
| 51 |
### ARMAN
|
| 52 |
|
| 53 |
ARMAN dataset holds 7,682 sentences with 250,015 sentences tagged over six different classes.
|
|
@@ -80,11 +49,9 @@ You can download the dataset from [here](https://github.com/HaniehP/PersianNER)
|
|
| 80 |
|
| 81 |
The following table summarizes the F1 score obtained by ParsBERT as compared to other models and architectures.
|
| 82 |
|
| 83 |
-
| Dataset
|
| 84 |
-
|
| 85 |
-
|
|
| 86 |
-
| PEYMA | 98.79* | - | 90.59 | - | 84.00 | - |
|
| 87 |
-
| ARMAN | 93.10* | 89.9 | 84.03 | 86.55 | - | 77.45 |
|
| 88 |
|
| 89 |
|
| 90 |
## How to use :hugs:
|
| 17 |
This task aims to extract named entities in the text, such as names and label with appropriate `NER` classes such as locations, organizations, etc. The datasets used for this task contain sentences that are marked with `IOB` format. In this format, tokens that are not part of an entity are tagged as `”O”` the `”B”`tag corresponds to the first word of an object, and the `”I”` tag corresponds to the rest of the terms of the same entity. Both `”B”` and `”I”` tags are followed by a hyphen (or underscore), followed by the entity category. Therefore, the NER task is a multi-class token classification problem that labels the tokens upon being fed a raw text. There are two primary datasets used in Persian NER, `ARMAN`, and `PEYMA`. In ParsBERT, we prepared ner for both datasets as well as a combination of both datasets.
|
| 18 |
|
| 19 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 20 |
### ARMAN
|
| 21 |
|
| 22 |
ARMAN dataset holds 7,682 sentences with 250,015 sentences tagged over six different classes.
|
| 49 |
|
| 50 |
The following table summarizes the F1 score obtained by ParsBERT as compared to other models and architectures.
|
| 51 |
|
| 52 |
+
| Dataset | ParsBERT | MorphoBERT | Beheshti-NER | LSTM-CRF | Rule-Based CRF | BiLSTM-CRF |
|
| 53 |
+
|---------|----------|------------|--------------|----------|----------------|------------|
|
| 54 |
+
| ARMAN | 93.10* | 89.9 | 84.03 | 86.55 | - | 77.45 |
|
|
|
|
|
|
|
| 55 |
|
| 56 |
|
| 57 |
## How to use :hugs:
|