# Error: embedding_intent_classifier as far as I can remember you probably have #intent which have only one item … #intent: abc - abcd should have more than one : #intent: abc - abcd - defg # Feedback on ConveRT Model + Rasa NLU # pip install rasa==1.5.0 ERROR: Could not find a version that satisfies the requirement rasa==1.5.0 do I have to do something special ? # How do you upgrade RASA I have ver. 1.3.9, how do I upgrade to the latest version ? I cant find this in the Docs ! # Wanted: new ideas for the Rasa demo bot I would like very VERY extensive QA/FAQ … may be mined from the forum and stackoverflow and probably carbon-copy sentences as answers from the docs … plus quick examples of doing stuff. May be as a second stage short guided Recipes … # Prediction: Sentence or words? As I’m reading the papers a question comes to mind … How would I go about if I want to featurize along with the intent and entities also POS tags. In my case Subject-Verb-Object. Let say I have a way to programmatically extract them via spacy OR pattern.en library # Prediction: Sentence or words? thanks thats what i was looking for… i.e. how does Rasa work … if you have more details on that let me know # Prediction: Sentence or words? Given the INTENT and all other STATE information how does RASA predict the response … Does it predict full sentence as a single unit OR as separate words that combined create a sentence ? Can you elaborate ? # Dependencies? Does RASA depend on Spacy ? If yes how to find the installed&required versions ? I’m asking because it seems rasa downgraded! my spacy version from 2.1 to 2.0 and this messes other stuff … Can I upgrade to the latest spacy version ? will this break Rasa ? BTW I dont use Spacy pipeline … # [Anon:DSL] Bot language What is this ? In short this is Domain Specific Language (DSL) for Rasa and can be used to generate all the required Rasa-files that you use to train your chatbot. # Split : domain.yml I just also figured out I can use C/C++ preprocessor c cpp -P included.inc > output.file # Split : domain.yml thanks … hiyapyco - cool… i searched for yaml&markdown lib and there was none that you can use to manipulate the structure in python … it seems also the markdown parser the rasa uses, doesn’t support multi-lines like this : # Leading questions? can you do that ? i mean story starting with bot action ! how would then be this triggered !?? how does the bot knows to use this !! # Split : domain.yml Is there a plans to be able to divide the domain.yml the same way we can split the intent and stories in multiple .md files ? What is this good for ? the file will be less cluttered it will be easy to switch on/off big portions of train data you can do incremental development of the bot # Greedy intent? hmm … thanks will try it … i just happen to be reading the tutorial on FORMS ! # Greedy intent? Lets say I have the following story : ## story1 * something - ask_question * answer - process_answer # Rasa Masterclass: Episode List & Resources Is there a plans for episodes on the architecture : like the ones you have on earlier versions, but abit more detailed, so we can make a connection with the API for extentding RASA, such as Actions,Custom NLU components, the pipeline, feature generation&merge, vectors creation and NN modules …etc # Can I use a database for training bot? I dont think so … nobody but you knows the schema of your data. F.e. i wrote some data in non-standard .txt file and then I wrote a script to generate rasa .md files from it. # Can I use a database for training bot? One option is to write script to generate nlu.md, stories.md and domain.yml from the DB. If you mainly have the replies in the DB you can create Actions to fetch the data and decide based on intent and slot data…what to do… # Novel idea cool thanks, will try it # Leading questions? Let say I want to code a dialog along the following lines : - do you want to try an exercise ? > ok - do you want to try random example ? > yes - Ex: here goes : x = 5; print x + 7, Q: what do you expect the result to be ? > 12 - Correct :) .. do you want to try another one? > yes - Ex: ..... # Multi line and bullets So inside domain.yml I should use two slashes to be able to write multiline inside the yaml file ex: utter_abc: - text: blah...blah............ \ blah ...blah # Retrevial-response : multiple intents match? but doesn’t then I lose the Respnse-pick-logic that handles the responses i.e. I have to code that myself ? # Novel idea hi, have you guys thought about creating a Chatbot to accompany the Rasa docs i.e. rasa-bot to teach and answer questions about RASA ? # RASA == Production system? yes … the IF-THEN rules match 90% Stories (except that stories take context into account in PS you do this via Working-memory) . Also normally it have a score attached to every rule which is updated in a similar manner to Reinforcement learning Q-value. AFAIK you do similar thing during interact… # Retrevial-response : multiple intents match? i see … i have in domain.yml - faq: triggers: respond_faq any way to attach some action code to respond_faq ? thanks # Retrevial-response : multiple intents match? When using retrevial-response mode it seems only the basic intent is matched ? Is this correct ? f.e. let say i have this intent : ## intent: faq/strings+strip when i --debug i see only “faq” intent… # Action POST request i see … it seems too complicated to be able to generate it on the fly … hmm. # Extracting : verbs, nouns, adj, phrases .etc? from another post I got this : msg = tracker.latest_message.get('text') I can spacy-nlp-pos it to get those … so I’m ok for now, thanks # intent: math_expr? how do I do that ? # intent: math_expr? Is there a way to recognize Regex as intent i.e. something like this : ## intent: math_expr - [+-*/()0-9 ](expr) Also I can’t seem to be able to capture the regex-match into a SLOT ??? # Error: embedding_intent_classifier i made it work # Extracting : verbs, nouns, adj, phrases .etc? Is there a way to extract verbs, nouns, adj, phrases .etc ( in supervised_embedding case) It will go into slots, right ? If possible will it influence ‘intents’ ? # Rasa X : [Correct this] jumping around? Can you guys fix the [Correct this] box. When it show up or when you click on it the whole chat jumps up and down … it is very confusing… # Error: embedding_intent_classifier i have responses.md file how do i make it work with ResponseSelector ? # Error: embedding_intent_classifier that was it !, btw i would need it for FAQ handling quick question : if I have multiple *.md files in /data, does rasa train pick them up … or there is some way to point them # Error: embedding_intent_classifier [nlu.md](https://forum.rasa.com/uploads/short-url/8ECFKNBTxw9bJ9NSUzCIanZ4ULz.md) (2.6 KB) # RASA == Production system? As I’m experimenting with RASA … something struck me … It very much resembles Production systems. What is a production system ? In general it is a system where all the processing happen via IF-THEN rules f.e. IF wm.intent == greet THEN utter_greet # pipeline: # "supervised_embeddings" - name: "WhitespaceTokenizer" intent_tokenization_flag: true intent_split_symbol: "+" - name: "RegexFeaturizer" - name: "CRFEntityExtractor" - name: "EntitySynonymMapper" - name: "CountVectorsFeaturizer" - name: "CountVectorsFeaturizer" # Inspecting data from the shell! here is quick example : (I’m slurping the files into arrays) [shell_cmds.py](https://forum.rasa.com/uploads/short-url/y1L5sJJnj8GxUas55nD8bdDSlAt.py) (2.3 KB) # Inspecting data from the shell! just view it … it will need retraining anyway if you change it ! will there be benefit of editing ? # 'Parsing' intent and filling slot? yes… i tested it … and works ! thanks… Is there a way to set the slot with specific value every time … f.e. let say i have : [conditional](etype) [if](etype) [if-then](etype) can I force it to set etype=conditional in all 3 cases ? # Action POST request how do you call actions. I tried the following : curl -d "etype=sum" -X POST http://localhost:5055/webhook curl -d '{"etype":"sum"}' -X POST http://localhost:5055/webhook # Inspecting data from the shell! you don’t leave the shell. when the files become big it is harder to find what you are looking for. # 'Parsing' intent and filling slot? let say i have the following story : ## code example * example - utter_example * explain - utter_explain * gotit - utter_gotit ## intent: example - [assign](etype) example - [sum](etype) example - [multiplication](etype) example - [conditional](etype) - [loop](etype) example - [eqvseq2](etype) … # RasaX: create intent when annotating OR creating new entries in a Chat often the new utterances can’t be matched to the available intents. # Inspecting data from the shell! Would it be very hard to inspect original data from the shell. Let me give example : > /show intent greet hi hello howdy .... # pipe is grep > /show i greet | he hello # show story .. >/s s sa… # RasaX: create intent Is there a plans to add capability to create new-intent in Interactive mode AND/OR RasaX. # Using slots? that’s the problem the documentation does not explain that !! # Using slots? I’m unsure how it works ? Does it autovivify implicitly if I just use it OR I have to explicitly have to declare it in domain.yaml so I can use it in stories.md # Using slots? if I have entities like this : * substring of a [string](type) but nothing is declared in domain.yaml is the slot used ? # Hierarchial dialog with repeatable parts? First question : Is there some formatting markup for displaying code ?? Here is an example dialog for which most of the responses are like a QA, but it is also very hierarchical and have many repetitions. How do you handle that ? # Extending FAQ scenarios? Response retrieval model allow simple QA scenarios. The problem seems to be from what I tested so far two fold. First the annoying one … you have to sync the editing of 4 different files … it is really hard keep track.