The api is build to handle the massive query, which using multi-processing to process the query message to classify the emotion.
ISSUE: With lots of query, the waiting time will be long. Due to the first in first out process.
The scale up version will build in the future.
Input : json object
{
data : [
{
"message": "抽顯卡做研究 DeepLearning + 黑色沙漠",
...
},
{
"message": "送趙奕誠出國報conference",ds
...
}...
]
}
Output: json object
{
data : [
{
"message": "抽顯卡做研究 DeepLearning + 黑色沙漠",
"emotion1": "anger",
"emotion2": "haha",
"ambiguous" "True"
},
{
"message": "送趙奕誠出國報conference",
"emotion1": "Sad",
"emotion2": "anger",
"ambiguous" "False"
}...
]
}
The example is in the EmotionDetection_ch