Judicial Judgments (Findings of Fact).
The judicial judgments (fact-finding) feature extracts fact-finding entities from judicial judgments and is a model training algorithm service on the NLP Self-Learning Platform. This service is currently in a trial phase. To use it, you must submit a request to be added to the whitelist. For more information about submitting a request, join the NLP Self-Learning Platform Q&A Group 2 (DingTalk group ID: 44619071).
Before you use this service, ensure that you have activated the NLP Self-Learning Platform. After activation, you can purchase a resource plan.
Activate NLP Self-Learning Platform: Activation page
Purchase an NLP Self-Learning Platform resource plan: Purchase page
1. Create a project
The NLP Self-Learning Platform (Click to enter the self-learning console) supports multiple basic projects and application algorithms. This tutorial guides you through creating a "Judicial judgments (fact-finding)" project on the platform.
Go to My Projects or Create Project. Select the Judicial judgments (fact-finding) algorithm and click Create. On the page that appears, enter a project name and description.
2. Prepare data
In My Projects, you can manage your data in the data center. You can create data in two ways:
1. Create an annotation task. 2. Upload a dataset.
2.1 Create an annotation task
Step 1: Upload documents to annotate and add annotators
The project creator and project administrators are annotators by default. You can also assign annotation tasks to your Alibaba Cloud RAM users. These users can then log on to the platform with their RAM user credentials to perform data annotation.
RAM user logon instructions:
1. Go to the RAM user logon page: https://signin.aliyun.com/login.htm
2. After you log on, go to My Projects and click Create Annotation Task.
Note: Only data files with UTF-8 encoding are supported.
Step 2: Add custom tags
The judicial judgment model includes more than 10 built-in entity labels that do not require annotation. This reduces the data annotation cost for model training by more than 50%.
Select the preset fields for this annotation task, and then add the custom tag field names one by one.
Step 3: Annotate data
2.2 Upload a dataset
The model learns patterns from annotated data. You must first prepare a labeled dataset.
If you have existing annotated data, you must organize it into the format that the NLP Self-Learning Platform supports and then upload it.
The annotated data that you upload must be in JSON format and comply with the following structure. Place the text content in the `content` field and the label content in the `records` field. In the `records` field, the key is the entity name, and the value is the offset of the annotated content in the original text.
{
"id_0001": { // This is the unique identifier for the document in your dataset.
"content": "At about 15:00 on September 4, 2021, the defendant Liu * Li, while intoxicated, was driving a small car with the license plate number Zhe B**208 near the intersection of Batou Road and Zhenda Road in Daqi Street of this district when his car scraped against a small sedan driven by the victim Han * Hong, causing a traffic accident with vehicle damage. It was later determined by the Traffic ** Brigade of the Beilun Branch of the Ningbo Municipal Public Security Bureau that the defendant Liu * Li was fully responsible for this accident. When police officers from the Traffic ** Brigade of the Beilun Branch of the Ningbo Municipal Public Security Bureau arrived to handle the traffic accident, they suspected that the defendant Liu * Li was driving under the influence, and the responding officers then conducted an on-site test on Liu * Li. The test showed that the defendant Liu * Li's breath alcohol content was 212 mg/100 ml, and a subsequent blood sample test showed that his blood ethanol concentration was 219 mg/100 ml. After the incident, the defendant Liu * Li and Han * Hong reached an agreement on civil compensation, which has been fulfilled.",
"records": {
"crime_scene_location": [{
"span": "driving from east to west on Jiefang Street",
"offset": [ // The offset of the actual bytes of the annotated field. The offset of the first byte is 0.
127,
139
]
}],
"breath_alcohol_content": [{
"span": "106mg/100ml",
"offset": [
212,
223
]
},
}]
}
}
}3. Create a model
In the Model Hub, click Create Model.
To view the model details, click View on the right side of the model.
Model details page
Test the model
After the model training is complete, you can test the performance of the trained model online.
Test the platform model
You can test text using the models that are trained on the platform. You can also train a model separately to improve the accuracy of specific labels.