如何解决购物辅助数据集设计
我想使用 DL 或新的 Transformer 构建购物助手,但我不知道如何设计我的数据集??
我已经找到了一个数据集,其中包含机器人和客户之间的一些对话,例如:
在第一个原始数据中:
['Hello how may I help you?','i want to order something','I can help you with an order.','i want to buy some cat treats','Alright what kind are you interested in?','the fish flavor ones','Alright is there a specific brand?','the national brand with the cat on the package','I think thats purina?','i have a coupon to use. the code is meow15. can you place my order?',"That coupon code doesn't include fish flavor,it is only good for pastrami flavor.",'that will be fine']
... 第 n 个原始 :
['Hello how may I help you?',"Hi,I'd like to make an order please.",'Okay,what would you like to order?',"I'd like to order a 12 pack of hot pink toe socks,please.",'Okay. That is going to bring your cart total to $19.99',"OK thank you. Now I'd like to apply coupon code SUMMERSAVE50 as well,"I'm sorry that coupon doesn't apply to the item you want to purchase.",'What? Why not? It says 50% off anything on the website','Some exclusions apply',"Well ya'll need to make it clearer what exclusions there are then! Everything OK? I think I'll just cancel the order.",'You can file a complaint if you like. Is there anything else I can help you with?','No. Please cancel the order. Thank you for your time.','Have a good day.']
还有 我有另一个数据集,其中包含从亚马逊抓取的问题-答案,其中包含一个或多个答案,例如:
{'questionType': 'open-ended','asin': '1466736038','answerTime': 'Nov 3,2014','question': 'What is the warranty on this?','answer': 'No warranty'}
现在我的问题是:
- 哪个数据集更有用,或者我可以将它们用作两个来源?
- 如果使用转换器模型,是否需要在具有意图和槽值的数据集上训练模型?
- 我可以将第一个数据集(一般购物)用于产品的特定元素 e.x 吗
- 我是否应该为每一个回合的意图添加一列?
- 我是否应该为所有数据集手动添加槽值?有没有自动提取槽值的程序??
请客气,如果有人有数据集布局的示例可以帮助我。任何想法都会帮助我。
版权声明:本文内容由互联网用户自发贡献,该文观点与技术仅代表作者本人。本站仅提供信息存储空间服务,不拥有所有权,不承担相关法律责任。如发现本站有涉嫌侵权/违法违规的内容, 请发送邮件至 dio@foxmail.com 举报,一经查实,本站将立刻删除。