Training data for City of Helsinki chatbots. Data includes training data for the maternity clinic chatbot NeRo for now. The chatbot is only available in Finnish for now.
The maternity clinic chatbot NeRo answers the clients of the Helsinki counselling centres’ questions without queuing at all times of the day. In addition to problems related to the child’s growth or development and pregnancy, the customers can also ask questions about dental care, speech development, and nutrition. Service can also connect the client to the nurse as needed, and the client can also tell the chatbot of their willingness to talk to the nurse.
The service is intended for customers of the City of Helsinki’s maternity and child health clinic. The service operates on a rule-based logic based on training data defined by the clinic’s experts. The training data consists of multiple parts, of which we publish the topics and their question variants (intents), dialogue variables/synonyms (entities) and answers.
No logs of customer discussions with the bot are published.
Attributes
Anwers
XLSX file.
- key = an identifying name (ID) unique to that response in the system. This is referred to in dialog definitions when assigning a response to a specific intent in a given situation
- value = the actual response text given to the client in the user interface. Occasionally includes so-called tags that provide clickable hyperlinks, selection buttons, livechat migration, or other functional elements. Texts separated by vertical line | are different variations of the same answer, so that there is a slight variation in the conversation.
Intents
XLSX file format: the first column contains example question, the second column ID for intent. That is, first the question method in which a particular thing can be expected to be asked, and then the Intent ID by which the system connects the question to the intent and performs a defined action for it.
Entities
XLSX file format: the first column contains entity ID, the following columns alternative forms for entity.
In the first column, the thing to which you want to be given a synonym or other thing that needs to be associated with that entity. Occasionally, bending forms are also added if the AI does not recognize them reliably enough without. In the following columns, synonyms/other words associated with the same thing. Note! The system from which exports are taken splits the same entity in exports over several lines for unknown reason.