The person teaching here (I don't know if she writes the scripts herself?) is very very good! She really ought to consider a career which involves teaching. Extremely clear, well organised and concise.
🎯 Key Takeaways for quick navigation: 00:00 🗣️ The most time-intensive part of creating a conversational AI is telling it what to do, accomplished through training data and rules. 01:11 📊 Training data includes user-generated text and patterns of conversation. The best data source is actual user conversations with the assistant. 02:35 📜 Training data is represented as "stories" in Rasa, teaching the assistant what to do next during conversations. 03:38 📝 Rules in Rasa describe short, fixed conversation sequences, helpful for handling small, predictable interactions. 05:12 💬 Intents are classes for multi-class classification, representing user input data for the assistant to recognize. 09:28 🧠 Intents should come from user data, not be entirely automated. Start with core intents, avoid having too many, and use entities for storing information. 17:06 🔄 End-to-end learning is used for ambiguous utterances, while intent classification is for unambiguous ones in Rasa.
Hey, amazing session. Just a quick question, I have been learning rasa for a chatbot that answers questions related to a few paragraphs of text. It seems like rasa can only be used to develop rules based chatbots wherein you have to train the model with question and answer for every possible question the user may ask. Am i right? I was actually looking for a framework that helps in learning machine learning. Is this not machine learning? Please Guide me!
Nice video, extraordinary made, congrats ! One question: how would you modelize a medical advice (actually pre-diagnosis and client file filling chatbot) app through Rasa ? You could have many many symptomatology and diagnosis "charts", would you modelize them with stories ? where would you use neuronal power, in just detecting the intents and this is all or it could also be used for decision-making in the dialog ?
Sure is! You can find all the videos in this playlist: th-cam.com/play/PL75e0qA87dlEjGAc9j9v3a5h1mxI2Z9fi.html (We're working on uploading the rest of them over the next couple weeks.) - Rachael
how can we handle if we have too many synonyms in our rasa model? is there any other way other than adding them in nlu.yml? in our case, we have wifi router name & it's mac address as synonyms. Like that we have many wifi routers in our rasa nlu
If you know the name/addresses ahead of time, I'd probably recommend a lookup table (rasa.com/docs/rasa/nlu-training-data#lookup-tables) to extract the names and a regex to extract the mac addresses, and then doing the mapping in a custom action using a set slot event (rasa.com/docs/action-server/events#slot) if it's important that you treat them exactly like synonyms. If you're using the info for database lookup I'd probably skip the synonym mapping and keep them seperate for lookup. -- Rachael
We make software tooling to help developers build chatbots. Rasa Open Source is our full framework & Rasa X is a free but closed source tool to help monitor and improve chatbots once they've been deployed. (Also we have an enterprise version with additional features if you're working at that scale.)
![[Live] : ONE 169 วันนี้!! "อนาโตลี vs อูมาร์"](http://i.ytimg.com/vi/IK3YZVWuFJg/mqdefault.jpg)
The person teaching here (I don't know if she writes the scripts herself?) is very very good! She really ought to consider a career which involves teaching. Extremely clear, well organised and concise.
🎯 Key Takeaways for quick navigation:
00:00 🗣️ The most time-intensive part of creating a conversational AI is telling it what to do, accomplished through training data and rules.
01:11 📊 Training data includes user-generated text and patterns of conversation. The best data source is actual user conversations with the assistant.
02:35 📜 Training data is represented as "stories" in Rasa, teaching the assistant what to do next during conversations.
03:38 📝 Rules in Rasa describe short, fixed conversation sequences, helpful for handling small, predictable interactions.
05:12 💬 Intents are classes for multi-class classification, representing user input data for the assistant to recognize.
09:28 🧠 Intents should come from user data, not be entirely automated. Start with core intents, avoid having too many, and use entities for storing information.
17:06 🔄 End-to-end learning is used for ambiguous utterances, while intent classification is for unambiguous ones in Rasa.
Very well summarised... Thankyou
Great video. Thanks 🙏🏿😌
i don't understand well the stories can you give more details on that
What does the rules and responses look like for the tickets example?
is Rasa trainable with your own data? Is it suitable for a Q&A conversational bot?
What is difference between rasa open source and rasa pro ? Can we do a bot with only rasa open source
where we find the source code
Hey, amazing session.
Just a quick question, I have been learning rasa for a chatbot that answers questions related to a few paragraphs of text. It seems like rasa can only be used to develop rules based chatbots wherein you have to train the model with question and answer for every possible question the user may ask.
Am i right?
I was actually looking for a framework that helps in learning machine learning.
Is this not machine learning?
Please Guide me!
RunsSlide clips
Nice video, extraordinary made, congrats ! One question: how would you modelize a medical advice (actually pre-diagnosis and client file filling chatbot) app through Rasa ? You could have many many symptomatology and diagnosis "charts", would you modelize them with stories ? where would you use neuronal power, in just detecting the intents and this is all or it could also be used for decision-making in the dialog ?
Thank you 🎉
Is this video part of a bigger playlist in which you explain all basics of Rasa?
Sure is! You can find all the videos in this playlist: th-cam.com/play/PL75e0qA87dlEjGAc9j9v3a5h1mxI2Z9fi.html (We're working on uploading the rest of them over the next couple weeks.) - Rachael
how can we handle if we have too many synonyms in our rasa model? is there any other way other than adding them in nlu.yml? in our case, we have wifi router name & it's mac address as synonyms. Like that we have many wifi routers in our rasa nlu
If you know the name/addresses ahead of time, I'd probably recommend a lookup table (rasa.com/docs/rasa/nlu-training-data#lookup-tables) to extract the names and a regex to extract the mac addresses, and then doing the mapping in a custom action using a set slot event (rasa.com/docs/action-server/events#slot) if it's important that you treat them exactly like synonyms. If you're using the info for database lookup I'd probably skip the synonym mapping and keep them seperate for lookup. -- Rachael
Can Rasa support Thai language?
What's her insta I wanna stalk and I don't even use Instagram damm
nice
What is Rasa about?
We make software tooling to help developers build chatbots. Rasa Open Source is our full framework & Rasa X is a free but closed source tool to help monitor and improve chatbots once they've been deployed. (Also we have an enterprise version with additional features if you're working at that scale.)
Rubén godoy islas