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

# Task Definition  
The above text describes a POV video segment, and there is also a question related to this video. You will need to further interact with the VLM to obtain more information to answer this question. But first, you must analyze the existing description and the question, then output the key entity names you believe are relevant to answering the question.  

You must select entity names from the following list:  
Characters: [chars]
Areas: [areas]
Objects: [objs]

# Output Format  
Question analysis: (your analysis)  
```json  
{  
    "key_entity_names": ["name1", "name2", ...]  
}  
```
