Closed christian-forta closed 7 months ago
@christian-forta Here's the finding for the first tx:
$ npm run tx 0xbb8fc2b258ef13b83f18f62e97ec24d584be822d923dc9c634f8c94b01d642e4
1 findings for transaction 0xbb8fc2b258ef13b83f18f62e97ec24d584be822d923dc9c634f8c94b01d642e4 {
"name": "Neutral Text Message",
"description": "Hi, I believe you have some DAIs to recover() from this contract 0xeBC29199C817Dc47BA12E3F86102564D640CBf99.",
"alertId": "NEUTRAL-NEUTRAL-TEXT-MESSAGE",
"protocol": "ethereum",
"severity": "Low",
"type": "Info",
"metadata": {
"sentiment_score": 0.7364263534545898,
"sentiment_time_secs": 1.2070223329999994,
"emotion_score": 0.6115671396255493,
"emotion_time_secs": 0.05640537499999887,
"anomaly_score": 1
},
"addresses": [],
"labels": []
}
The second tx message was in Chinese, so the bot couldn't parse the text correctly for analysis. We probably need a separate model to analyze Chinese text. Just not sure how often this happens. Maybe we just create a separate bot for chinese text.
大佬,哪怕打去黑洞也绝不要还回去啊,干死这些傻逼项目方 --> Boss, even if you hit the black hole, never return it, kill these idiots
We can potentially use this model https://huggingface.co/IDEA-CCNL/Erlangshen-Roberta-110M-Sentiment
https://etherscan.io/tx/0xbb8fc2b258ef13b83f18f62e97ec24d584be822d923dc9c634f8c94b01d642e4 https://etherscan.io/tx/0xbf295a011f0088d8a5cad4b9a15d3acebe3a4b5184f20af5ec69a07384b69a2c