I have seen some references to the Auto-learn function of Watson Dialog but I can't find coverage in any of the documentation. Can you point me to a source of information on how best to use Auto-Learn?
Thank you for your feedback, we are always working to improve our documentation.
For your immediate benefit, auto-learn is a bit of a misleading name for a feature t, but the name has stuck.
Autolearn has become the "did you mean..." with four bullet points that shows when a user sends an input that has no direct matches.
A little history...at one time, we thought that if a user typed something, saw the did you mean... and clicked a link, their intial input must have meant the same thing as the one they clicked, the system should automatically remember that.
Imagine this:
"What are your credit card fees?"
Did you mean... 1. apply for a credit card 2. cancel a credit card 3. pay your bill
click apply for a credit card
The user was simply interested in that, but obviously the two DO NOT have the same meaning. So we realized thats a bad idea, the system will learn incorrectly. However, we still call it auto learn.
Related
I want to create a skill that's a simple game, where first the user launches the skill with its invocation name and then Alexa asks a question, "Shall I roll the dice?" If the user answers "Yes," it rolls the dice, and says the result. Then Alexa asks again, "Shall I roll the dice?" If "Yes," do the same thing. This is the main loop I'm talking about, and it'll continue until the user answers "No" or "Quit" to this question.
I just can't figure out how to add the loop, or where it should go. I've looked at tutorials and videos and whatnot and just nothing I've found mentions a loop which I find really odd. But I'm a noob at this.
Any help would be awesome. I've been wanting to do this skill for so long but just am stuck on this loop thing.
I recommend you to take some time to understand how a skill work and then I recommend you to develop a quiz skill from this doc
You will then have a better understanding of how a request is made to Alexa service and how a response is returned. The logic behind Intent, how does a slot work, ...
An Alexa Skill is like a card game. The player can select any card at any time. Each card has its own function and is triggered by a voice.
So when the skill first asks the user for Shall I roll the dice?, the user will say either yes or no.
If the user says yes, it will then go to your AMAZON.YesIntent,
If the user says no, it will then go to your AMAZON.NoIntent.
But you also need to make sure that the user can also say:
Stop > Amazon.StopIntent
Anything else, such as, cheese > FallbackIntent
By doing the quiz skill cited above, you will understand how to build your interaction model effectively.
A loop is straightforward. If the user replies yes, then in your intent handler for AMAZON.YesIntent, you will need to trigger the same function that will inject, in the response builder the prompt: Shall I roll the dice ?.
Keep in mind that a user can also ask to repeat. Imagine a skill being a personal assistant. It's not a voice mail. There are many other ways to say Shall I roll the dice? to not sound like a robot. Try implementing different response values possible to have a great customer experience overall.
I have a skill that elicits a U.S. state and county from the user and then retrieves some data. The backend is working fine, but I am concerned about how to structure the conversation. So far, I have created an intent called GetInfoIntent, which has two custom slots, state_name, and county_name
There are about 3,000 U.S. counties with many duplicate names. It seems silly to me that I am asking for a county, without first "narrowing down", by states. Another way I can think of to do the conversation is to have 50 intents, "GetNewHampshireInfo, GetCaliforniaInfo, etc. If I did it this way, I'd need a custom slot type for each state, like nh_counties, ca_counties. etc.
This must be a pretty generic problem. Is there a standard approach, or best practice, I can use?
My (not necessarily best practice) practice tips:
Single slot for single data type. Meaning only have one slot for a four digit number even if you use it in more than one place for two different things in the skill.
As few intents as you need
no more no less. You certainly can and should break up the back end code with helper code, but try and not break the intents into too many smaller pieces. It can lead to difficulty when Alexa is trying to choose the intended intent.
Keep it voice focused. How would you ask in a
conversation. Voice first development is always the way to go.
For the slot filling I think it is fine to ask both state and county.
If the matching is not correct ask for confirmation.
Another option is to not use auto filling within the Alexa skill and use the dialog interface. Ask the county first and then only when it has more than one state option and is ambiguous continue the dialog to fill the state.
Even if you did have 50 separate intents you really never want to have two slots that can be filled by the same word. For example having a mo_counties and ky_counties that Clack satisfies both is ambiguous and can cause unneeded difficultly.
So for someone looking for the "best practice" I have learning that there isn't one yet (maybe never will be). Do what makes sense for the conversation and try and keep it as simple as it needs to be and no less on the back end.
I also find it helpful to find a non-developer to test your conversation flow.
This wasn't really technical and is all opinion, but that is a lot of what Alexa development is. I would suggest Tuesday Alexa office hours at https://www.twitch.tv/amazonalexa very helpful and you can ask questions about stuff like this.
So I'm working on something that's going to be a rather large undertaking. I've figured out how to do a "bare-bones" kind of dice hud that just rolls a basic 2-20. However now I need to go to the next step.
I want to make a roleplaying system dice hud for my sim. For this I want it so that when you click the HUD you get a menu, that lists all the skills in my system. When you click the skill it refrences a notecard in the Hud to do some minor math before displaying the result: IE.
There's a normal 2d6, 2d8, 2d10, 2d12, 2d16, 2d20 ((Whatever basic configuration that always rolls a standard die))
Though I want it to look into a note card to add in a character's "STATS" and "SKILL LEVEL"
So say they want to hit someone with a sword?
I want the Hud to generate a random value between 2 and 12, then add in the character's Strength, speed, perception stats as well as their sword skill level.
If I could see the basics of HOW to start this I can then move forward from there.
You cannot write into a note card using a LSL (or other) script.
If you want to roll a dice, simply use llRound( llFrand ); or integer x = (integer)llFrand(19)+1;
You could use a webserver to save information like that. Just google free web space, you're gonna find a lot. SL's HTTP communication is, let's say, ok.
This is kind of a big project. If you don't know where/how to start you should hire a professional. Just look for groups in-world. You're going to find a lot of people willing to help you :)
In LSL you cannot write in a notecard, however if you try OpenSim you can use that function:
osMakeNotecard(string notecardName, list contents);
But that is only available in OpenSim, see osMakeNotecard.
I have some sort of a heatmap and I can see how my users are using my website. Now I know where they are clicking, how much time it costs to complete a set of instructions, how much do they navigate between pages, etc.
So given that I have information with of this kind:
12/12/2014 12:45:00 - User pressed button 1 on page 1
12/12/2014 12:45:15 - User pressed button 2 on page 1
12/12/2014 13:00:00 (15 minutes delay) - User pressed button 3 on page 1
Now comes the hard part - how do I process this kind of information? For example how do I know that the user is lost on my website (if there is 15 minutes delay - does this means that his phone rang or my UI is bad?). And also - how can I find some patterns in large amount of data - say every third user spends 15 minutes after the second click to find what he has to click next.
What is the correct approach here? Thanks.
In order to derive useful information from raw data you need some context. You need to be clear about what expected user behaviour is and, where appropriate, what you are aiming for the user to do (eg. buy a product, register, make a comment etc.).
For example, if you have an event splash page with a big button to book a place, and you find that a lot of people click on that button very soon after they arrive, that's probably a good thing. If you have a page full of important information that you want people to read, and they click away just as quickly - that's really not a good thing.
It sounds obvious but so many people fall in to the trap of trying to evaluate user behaviour without being clear about the context - and without acknowledging that the very same number can mean very different things depending on that context.
Evaluate each page of major section of your site and outline what is there, and how you'd expect users to interact with it. How long would you expect a user to spend on that page? Where would be the logical place to go next? Is this a logical place to leave the site (I booked, I'm done), or is a user leaving the site here a failure? And so on. Then compare these expectations to the reality you see from your heat map.
Don't get too hung up on individual cases - if one person took 15 minutes on a page that should take 30 seconds, that was probably the phone ringing. If 90% of visitors take 15 minutes, then your page needs re-evaluating.
Lastly, pay as much attention to what people don't do as what they do. Everyone's eye is drawn to the bright spots on a heat map or the rows at the top of a chart with big numbers. With analytics, a lot of the most useful information is what you expected to see people doing, but they aren't. Again, to realise this information you need to have defined that expected behaviour.
I've got a WinForms client-server app that displays various offers in a list. Every user (client) has a "rating". An offer consists of various data including a minimum and maximum rating. If a user's rating does not fall in that interval, he should not be able to take the offer.
Of course I could just perform some server filtering and send a list of offers prefiltered for each user to the client application. But that would surely, and rightfully, lead to confused requests "Why isn't this offer showing up? I know it exists, it shows up on [other user]'s screen."
How should I handle this? My favorite solution so far is to grey out the offer and add a tooltip "You can't take this offer because your rating is too high/low" while displaying greyed-out offers at the bottom of the list to leave the actually valid offers easily visible on top of the list.
A disabled option tells the user:
The action is possible.
Just not right now.
But the user can make it possible.
Unless there is some simple action the user can do to change his or her rating (e.g., by selecting some other controls in the same window), do not use disabling and do not show the offers. Disabling may confuse some users who will then hunt around the window for something to do to enable those offers. It’s a great idea to use a tooltip to explain disabled objects, but that’s not a standard and not all users will think to hover the mouse over a disabled option (Why should they? It’s disabled).
Including offers users can’t have, even when disabled, clutters your display, forces more scrolling, and distracts the users from the offers you actually want them to consider. Furthermore, showing unavailable offers can come across as taunting (“ha, ha, your rating isn’t high enough”) and may diminished the perceived value of the available offers by comparison, resulting in lower user satisfaction.
It seems unlikely to me that users are going to go around comparing the offers on their windows, but maybe you have user research saying they do. In any case, you should label the list of offers to make the criteria clear (e.g., “Offers available at your rating” may be sufficient).
If you want to encourage users to increase their ratings, then maybe include something advertising the benefits of an improved rating. For example you could have a link "-Improve your rating- by four points and get -five additional offers-." The first -link- tells users how to improve their rating, while the second lists the offers as a motivator. The latter link should only be there if the offers will still be available if the user actually succeeds in getting four more points.
That sounds like a good way to do it.
As a slight improvement, if it makes sense for your application, you might consider including the actual numbers in the tooltip, e.g.:
This offer requires a rating between 5 and 8. Your rating is 4.