Wanted: new ideas for the Rasa demo bot


(Akela Drissner) #23

@alucard001 no worries, we always welcome suggestions for improvements to our docs so feel free to create issues on our github repos if you spot anything. and @Juste, our developer advocate also regularly does some live streams/tutorial videos to address some things that can’t be captured in our docs :slight_smile: so there should be some more of those coming soon


(Weilingfeng1996) #24

I want an instance —》 ( " Slots Set By Clicking Buttons")


(Neerajb1) #25

We can implement auto suggestion text (intent) while user is typing. Please let me know if it is already is rasa


(Tarun Bhavnani) #26

The one thing I feel is missing in rasa or the documentation is how to seamlessly take user inputs. These are most required as they can be used directly to fill slots or rectify wrong filled slots. i.e. confirming for the name, taking inputs from lists, buttons also are cumbersome work with.

a dispatcher without closing the current action can also be a good alternative.


(siri) #27

i think we should develop feature such that RASA should build a functionality which facilitates the users to integrate libraries which help to interact with different platforms (like IOS/Android) thus making rasa available as a handy tool to built micro applications also.


(Gaurav) #28

I can understand Your pain, I also spent two whole days understanding Botpress, That’s why switched to Rasa, There are tons of Tutorial available for RASA but not officially on the main documentaion but as github repo. I understood more Building a chatbot with Rasa NLU and Rasa Core on Vimeo by seeing this instead of documentation


(Asmaier) #29

It would be nice, if the bot could do or try to do spelling correction. I would be very interested to see, how one can implement this with Rasa.


(siri) #30

new features added post that version currently 0.12 is the official latest rasa available and the features discussed there such as rasa_nlu_converters is no longer available in 0.12 version

i would request RASA dev team to also concentrate on proper documentation so that new features can be used


(Andreas Stuber) #31

It would be useful to have demo bots for various industries (Travel, Banking, Insurance,…). It would be good enough to have one happy path and 1-2 scenarios only (scripted just for a demo, so that a sales and not only an engineer can give a demo). Examples for demo setups are Microsoft with Contoso Resources - Health Bot | Microsoft Docs or Cisco with Cumulus Cisco Unified Contact Center Express 11.6 v3 | News | Cisco dCloud

In most customer cases it would be good enough to have a short, scripted setup. Customers understand that if they want to have something more elaborate there would be additional services. But to start with, they want to see “ok you can build a sample chatbot from a Facebook or Webpage”.

We are thinking about building (and sharing) a demo for banking, and one for travel. If anybody has already done this, let us know.


How do you measure the success (rate) of your bot?
(Christian Meyer) #32

Hello Andreas, we are also setting up a demo bot for the public sector showcasing public services by example. Maybe we can cooperate a little bit. Our main issue at the moment is developing a generic UI that can be adapted to the look&feel of several public sector authorities.

Greetings Christian Meyer


(Christian Meyer) #33

A good working fallback policy is mission critical for a reliable and believable chat bot.


(Al Mahdi Marhou) #34

i think improving the fallback policy with more features is an important point . it will considerably improve the learning curve


(salman) #35

I think to attract beginners or encourage more people to use rasa, these things will help.

-More components for nlu pipeline

-Video tutorials for easily achieving complex tasks in rasa core

-python, http api tutorials explaining different things in rasa

-In stories there should be an option to specify entity value based response

-Tips and tricks for rasa

-Strong community support for people issues


(WoWuQ) #36

The following are my recommendation:

  • the document should be more detail, and divided by versions
  • memory series policy should more, and it can support stories convert about user intent in the middle of current stories, i suggest that you can add redecide event
  • add more nlu component
  • after add more nlu component, the stories Configuration can not only by the intent

(Jwickers) #37

Implementing FAQ also seems to be a common use case for bots and there are a few threads in the forum suggesting using rasa addons, but it would great to have an example of best practice implementation in the demo using the rasa.com website FAQs for example.

PS: I know there is a faqs branch but it does not seem to what is usually suggested, so not sure if that is considered best practice.


(Psdpluto) #38

How about a demo Bot to guide users building their demo Rasa Bot? Not python,No coding, just chatting and the job done. Maybe more helpful for beginners


(pablo) #39

hi andreas, ihave made a demo with dialog flow and php conect with a webhook and a data base, my bot can answer questions, make sales, and generate information in the database also read information about it as it can be stock and discount a product’s stock for example, if you want we can get in touch to see if I can help you with your ideas, also the same bot can be integrated into the website, facebook and telegram, could be done with whatsapp but still not probe the api and would be only for whatsapp business


(Ievgen Naida) #40

Good fallback system.


(Serge B) #41

Support for multiple language - .e.g. French. I.e. how does the bot handle the entire conversation in 2 languages and how does it react to convo starting in one language and user switching to other? Maybe with a link to a blog that describes best practices for building bi-lingual chatbot.


(Ievgen Naida) #42

To support multiple languages you should: Train two NLU projects (can be the same NLU instance). Train two different core projects. Switch URL on a button click. or even pretrain stories in another language to switch URL on a client.