There're 3 main nodes in our watson conversation(assistant) dialog, we'd like to ask question between each nodes and each of them have a long child nodes, how can we keep the previous answer while we jump to ask another question ?for example, A qustion like: I'd like to book a flight, Answer: what'd date is your flight? Then Ask: My car is out of oder, Answer would be: please get to a repair store, now I will keep answering that flight date, a complex dialog diagram as the attached, any advice would be appreciated... another example as follows: ask a question on node1(not finished yet), ask another question on node2(finished), then keep answering question on node1
You can gather multiple pieces of information from a user within the same node as an alternative to create child nodes usings SLOTS. Slots collect information at the users' pace. Details the user provides upfront are saved, and the service asks only for the details they do not.
More details here: https://console.bluemix.net/docs/services/conversation/dialog-slots.html#dialog-slots
Also,you can Digress from the main topic by switching it ON
More details on Digressions: https://console.bluemix.net/docs/services/conversation/tutorial-digressions.html#tutorial-digressions
Related
I have a web background majorly with javascript, I have started learning Corda recently for project implementation and need guidance in this regard,
So our application is based on the web, the user signs up with different school name, create question papers, and then want to share either part of it or whole with teachers of some other school in our platform. they can make changes and assign it back to creator and the process goes back and forth, finally signing the paper to be finalized, once finalized it cannot be changed by anyone. I need to store these transactions in Corda application, not sure how to go about it, I did try replicating it using negotiation application in corda/kotlin/sample, but stuck in a bug as I was trying to send a list of objects.
I do have the following questions in mind
Should I use enterprise edition or go with open source as I think I need schema design for this. web db is in postgress
As far as I have seen each node is predefined in the config with username and password,is there a way to create the node while the user signs up.
I have schools and teachers inside the school, do I need a separate node for each school and then create states in each node(not sure if a node can be set up at run time). or do I use the account's library provided for creating the account of each teacher, if yes id there a way to use passwords in it, unable to find password field in it.
how do I send an array of objects to the state, or should I create a separate state for each question, as different questions can be assigned to different teachers, but again multiple questions can be assigned to the same teacher.
These are few questions on my mind any help is much appreciated, as most of the examples gave IOU samples or states with int and string, Please guide me in the right direction.
Alessandro has good advice here, definitely look at the samples repos for inspiration on how to build what you're looking for.
start with open source, it's easier to prototype and you can switch to enterprise later it won't be an issue for you
this depends on design, you wouldn't really want to create a new corda node per-person, you might want to have corda accounts that run on a single node instead. See accounts sdk here: https://github.com/corda/accounts
what you might do is make a corda node for each school and then accounts per teacher like you were already were thinking. That would mean only a couple of nodes based on the number of schools you have.
as long as your state is marked with #CordaSerializable you won't have problems sending arrays of data, I send an array in a state in this sample here: https://github.com/corda/samples-java/blob/master/Advanced/secretsanta-cordapp/contracts/src/main/java/net/corda/samples/secretsanta/states/SantaSessionState.java#L24
https://github.com/corda/samples-java
https://github.com/corda/samples-kotlin
I am building a quiz app and I am struggling about data structure in firebase.
Actually, I have two collections:
Questions collection, containing one document foreach question
Users Answers collection, containing one document foreach user. Inside user's document, there is one map per answered question where I store if question has been answered and if question has been right answered.
So, where is the problem ? For my app, I need two things:
Fetch list of a specific user answers. Actually there is no problem because I can query users answers collection for a specific user and retrieve all his answers within a single document.
Fetch list of a questions never answered by user. This is a quiz app, so it is mandatory to show questions never answered by user to progress in the game.
The current (not best) solution is this : when the user launch the app, I request all questions, get his answers and then I can know which question I should show to user and the ones I shouldn't. The problem of this solution is that I have to query all the questions documents so it represents a large number of reads (and firebase bill by number of reads 🙃).
I would like to paginate the number of questions to show to user (so questions never answered) to avoid extra-billing but I can't figure how to request them.
I tried to use not-in clause but it is limited to 10 items and I have a lot more than 10 questions.
I would like to know if you have any ideas to bypass this problem. Hope I'm clear enough.
Firestore doesn't support using the results of a query to exclude documents from another query. There are no SQL-like joins that query across documents in multiple collections.
If the provided inequality filter (10 items max) isn't sufficient for your use case, I'm afraid you're stuck with the solution you have now by filtering the results in the client app.
Some developers might choose to duplicate data into a SQL database to make these kinds of queries possible, but you'll have to decide for yourself if that's worthwhile.
We can consider this, I am wondering this could make your idea work.
This will only work when the users you know are constant.
Every questions docs need to have this property: not_answered_by
And set all users' ids list inside this property initially for all questions.
Once a user answered against one question then update certain question doc to remove that user id from not_answered_by So like this below, I remove user who has id:ccccc inside not_answered_by list from question :PXxzLaZp4kDSzewo7kht when user answered on that question.
Then you can run this query .where("not_answered_by", "array-contains", user_id) against questions collection to get all questions not answered by selected user.
If you have this structure of questions collection, then the powerful advantage is that you don't have to query double to get questions not answered by user, which does cost time and money. The remove action from not_answered_by doesn't affect speed low because you know which user id to be removed from which question id.
One tip: if you dont want to set not_answered_by to have all users' id list initially then just have another property: is_answer_started to set false for all questions. And whenever question is started to be answered then set not_answered_by to have all users' list.
not_answered_by is array of strings and it doesn't affect to your cost of firebase, important fact on your side is getting result in cheap and fast right?
Important to note: this is case of users' id list are constant, this doesn't work if users collection is dynamic which means user is added dynamically so please think this is a tricky method but I think it is best approach for now. PS: Firebase query doesn't support array_not_contains which I was about to find and use it. We can approach this in javascript via array includes function you know. :)
Vice versa, you can have answerd_by property inside question doc to get answered questions list against user.
I have a basic model set up but it appears that the Alexa beta builder expects the design of the model to be up front with explicit lists of possible answers to requests. However, the Jeopardy game on that platform appears to allow new answers from day to day and I am supposing they don't need to update their model each day. Can someone shed light on this process?
I would like to be able to:
Launch intent to start conversation
Handle back and forth within web service response and request
Handle basic stop/start from Alexa
How does the middle portion (response/request) get modeled up front?
Currently, you need to provide a list of expected values for each of your slot types (a slot type may be question_response in this case).
Not sure how Jeopardy works but I have a couple ideas:
They provide huge lists of possible answers (and they update them from time to time). I bet there is no one writing new questions every day, they have a pool of questions and they can anticipate the answers to those.
They have access to an API that is not open to the general public (as do people who are participating in the social bot challenge) or are allowed to use the deprecated LITERAL slot type.
For example if you had a question and script like
Are you over 13+
a. if yes go to #2
b. if no end
Please pick a chat room
Looking at hubot it looks like it looks for specific questions but doesn't necessarily store state or allow the progression of a longer interview by remembering previous answers and doing "if this then that".
Does hubot support this? If not is there another open source alternative that is a better fit for the use case?
Would I just store user information in the hubot brain for each user's answer, building up a user's state over time? I suppose with each user answer we can look up the user's state, and their current progression in the hubot brain and then handle it accordingly.
However, if it's more of a script rather than looking for any old random question via regex matching, does this take away from most of the benefit of using hubot in the first place?
I am building an application in ASP.NET, C#, MVC3 and SQL Server 2008.
A form is presented to a user to fill out (name, email, address, etc). I would like to allow the admin of the application to add extra, dynamic questions to this form.
The amount of extra questions and the type of data returned will vary. For instance, the admin could add 0, 1 or more of the following types of questions:
Have you a full, clean driving liscence?
Rate your drivings skills from 1 to 5.
Describe the last time you went on a long journey?
etc ...
Note, that the answers provided could be binary (Q.1), integer (Q.2) or free text (Q.3).
How would I allow the posting of 0, 1 or more random values from the form back to my code?
Any help would be greatly appriecated.
Thanks in advance.
PS. I asked a similar question regarding the storing of this data in sql, see here
You will need to use an EAV scheme for storage. For a discussion of the topic, along with common pitfalls and solutions, see Best Practices for Semantic Data Modeling for Performance and Scalability. Other alternatives are Sparse Columns and XML data type column, but I think EAV is superior most of the times.
You'll have to accomplish this using dynamic controls.
Basically you'll read the question list from your database, inject zero or more custom controls to emit the question(s) and accept the answer(s). Probably wrapped in a repeater.
On post back you'll have to rehydrate the dynamic controls so that you can get the values.
I have done this before when I worked on a survey tool.
So first of all you need to decide what questions type you'll have. I had SingleOption, MultiOption and FreeText type questions.
User logs into the admin screen and selects a type of question he wants to add. He then clicks on Add button, which submits a request to a controller/web service.
Controller looks at type of question and adds a record into the database.
Rough table structure below:
QuestionID | SurveyId | QuestionTypeId
2 4 1
This is your basic structure.
When you generate an interface for answering questions, you will retrieve all questions with a SurveyId = 4.
You will then write an HTML helper that looks at question type and generates an appropriate control.
When it comes to submitting a survey, or answers to your question, you will display all questions and answer fields in some div, say we call it "survey_questions". Inside this div will be a form which will contain all your questions and controls for answers (text inputs, checkboxes etc).
You will then submit this form through jQuery to a controller/web service. Controller/Web Service will read survey id and it will iterate through all submitted values.
You will then use these values to populate a table for storing survey results. I understand that this is very brief and abstract, but I have worked over a month on this, so if this is useful and you need further help, then please ask.
I found a great article that solves my problem.
You can find it here