The following is a segmented description of a POV video: [Segmented_Description]

# Task Definition
The above text describes a first-person video segment. Based on the video description and the list of detected entities, extract the actions contained in the video.  

An action refers to a simple movement performed by a person in the video, such as "picking up," "stirring," "opening," etc. Actions are executed by a person and applied to an entity. An action may involve the following entities: **Agent**, **Patient**, **Instrument**, **Source**, **Target**.  
- Agent: The initiator of the action, usually a person (required).  
- Patient: The entity affected by the action.  
- Instrument: The tool used to perform the action.  
- Source: The initial location of the affected entity before the action.  
Target: The location of the affected entity after the action is completed.  

Only the Agent is mandatory; all other entities are optional. Source and Target are usually used in actions involving movement, and Instrument is used in actions involving tools. Each of the above items can only contain one entity.

An action includes the following attributes: action name, action description, and the time range of the action.  
The output list of actions must follow the chronological order of the original description.  

Examples:
1. Description: A man is doing yoga.  
{"action_name": "doing yoga", "agent": "man", "action_info": "The man is doing yoga.", "time_range": "00:00:00-00:00:30"}  
2. Description: A takes an apple out of the brown bag and places it on the table.  
{"action_name": "take out and place", "agent": "A", "patient": "apple", "source": "brown bag", "target": "table", "action_info": "A takes the apple out of the brown bag and places it on the table.", "time_range": "00:00:30-00:01:00"}  
3. Description: C cuts a potato with a knife.  
{"action_name": "cut", "agent": "C", "patient": "potato", "instrument": "knife", "action_info": "C cuts the potato with a knife.", "time_range": "00:01:00-00:01:30"}  

Make sure the entity name you fill in is exactly the same as that in the following list:
Characters: [chars]
Areas: [areas]
Objects: [objs]

# Output Format
{  
    "actions": [  // Must follow chronological order  
        {  
            "action_name": "",  
            "action_info": "",  
            "time_range": "hh:mm:ss-hh:mm:ss", 
            "agent": "",    // Required  
            "patient": "",  
            "instrument": "",  
            "source": "",  
            "target": ""  
        },  
    ]  
}
