I am building an new action. I have multiple intents, and I added ~20 training phases for each of them. However the matching seems to be pretty terrible
I have webhooks configured for the scenes. I look at the raw query and the handler that was matched. I have the exact phrase in the training phrases, but a different intent gets matched.
It seems like old Dialogflow interface had a way to improve training (https://cloud.google.com/dialogflow/es/docs/training#tool) allowing you to match actual queries with the right intents. Is there an equivalent tool for the new actions builder?
There isn't (currently) such a tool.
However, it seems odd that a training phrase that was used wouldn't match if the Intent is one that is set to be recognized in a scene (or is a Global Intent).
The last point is the most important - make sure that you have the Intent set to a scene and that you're in that scene when you're attempting to trigger the Intent in question.
Related
Just in case I'm trying to solve the XY problem here, here's some context (domain is a role-playing game companion app). I have a document (campaign), which has a collection (characters), and I'm working with angular.io / angularfire.
The core problem here is that if I query the collection of characters on a campaign, I get back Observable<Character[]>. I can use that in an *ngFor let character of characters | async just fine, but this ends up being a little messy downstream - I really want to do something like have the attributes block as a standalone component (<character-attributes [character]="character">) and so on.
This ends up meaning down in the actual display components, I have a mixture of items that change via ngOnChanges (stuff that comes from the character) and items that are observable (things injected by global services like the User playing a particular Character).
I have a couple options for making this cleaner (the zeroth being: just ignore it).
One: I could flatten all the possible dependencies into scalars instead of observables (probably by treating things like the attributes as a real only-view component and injecting more data as a direct input - <character-attributes [character]="" [player]="" [gm]=""> etc. Displayable changes kind of take care of themselves.
Two: I could find some magical way to convert an Observable<Character[]> into an Observable<Observable<Character>[]> which is kind of what I want, and then pass the Character observable down into the various character display blocks (there's a few different display options, depending on whether you're a player (so you want much more details of your character, and small info on everything else) or a GM (so you want intermediate details on everything that can expand into details anywhere).
Three: Instead of passing a whole Character into my component, I could pass character.id and have child components construct an observable for it in ngOnInit. (or maybe switchMap in ngOnChanges, it's unclear if the angular runtime will reuse actual components for different items by changing out the arguments, but that's a different stack overflow question). In this case, I'd be doing multiple reads of the same document - once in a query to get all characters, and once in each view component that is given the characterId and needs to fetch an observable of the character in question.
So the question is: if I do firestore.collection('/foo/1/bars').valueChanges() and later do firestore.doc('/foo/1/bars/1').valueChanges() in three different locations in the code, does that call four firestore reads (for billing purposes), one read, or two (one for the query and one for the doc)?
I dug into the firebase javascript sdk, and it looks like it's possible that the eventmanager handles multiple queries for the same item by just maintaining an array of listeners, but I quite frankly am not confident in my code archaeology here yet.
There's probably an option four here somewhere too. I might be over-engineering this, but this particular toy project is primarily so I can wrestle with best-practices in firestore, so I want to figure out what the right approach is.
I looked at the code linked from the SDK and it might be the library is smart enough to optimize multiple observers of the same document to just read the document once. However this is an implementation detail that is dangerous to rely on, as it could change without notice because it's not part of the public API.
On one hand, if you have the danger above in mind and are still willing to investigate, then you may create some test program to discover how things work as of today, either by checking the reads usage from the Console UI or by temporarily modifying the SDK source adding some logging to help you understand what's happening under the hood.
On the other hand, I believe part of the question arises from a application state management perspective. In fact, both listening to the collection or listening to each individual document will notify the same changes to the app, IMO what differs here is how data will flow across the components and how these changes will be managed. In that aspect I would chose whatever approach feels better codewise.
Hope this helps somewhat.
How to I create an #Intent which looks something like this:
How much is a #ProductType?
Whereas the #ProductType is an simple Entity which consists of:
Soft Drinks: Coke, Pepsi, Sprite, Fanta
Fruits: Apple, Banana, Watermelon
I tried adding an Intent with above settings, but it doesn't seem to work. Is such ability natively supported in IBM Watson? Or otherwise, do I need to manually handle in the Dialog, using Conditions and stuffs? Please kindly advise.
The training is based on regular language and typical sentences or phrases. So #ProductType is not what you want in the phrase, but any of the fruits or drinks.
By defining the entities, Watson Assistant later learns the connection and to identify the entities and intents.
To get started, you define the intents and entities. Both can be imported from lists. Then you add the dialog which references the different types.
This blog should give insight to all the ways to train an entity and how it is used within intents.
https://medium.com/ibm-watson/all-about-entities-dictionaries-and-patterns-with-watson-assistant-part-1-5ef7254df76b
There are a number of possible pipelines you can choose from.
1. Indirect references: this is the preferred method.
Use natural language in your intent training data. "I want to buy a pear"
Watson will automatically see the other values you have related to pear and use those as intent training as well. This will be the fastest and simplest way to manage your data
2. Direct references: this should only be used if absolutely necessary
Directly reference the entity in your intent data. "I want to buy an #pear"
Nothing is done in the UI to tell you this works, but it does. This tells Watson the entity is a very important term and will increase the weight, as well as reference all synonyms with high weight. This is more effort for you to go through your entire workspace and relabel everything this way, hence why it is not recommended unless absolutely necessary. By doing this, you also tell watson that when the system sees various fruits without the # symbol, to ignore them as entities which is not ideal
3. Contextual entities. This is highlighting them like in your screenshot.
Note the UI has been updated so there is no an annotation mode instead of just highlighting. This builds a model around the entity, and is good for things like names or locations, but not necessary for a small list of items like crayons in a box, or fruit in a store. This will ignore all of the dictionary values youve created and only look at the model. It should be used according to the blog above when the use case is ideal.
What #data_henrik answered was partially correct. But it doesn't seem like Watson Assistant "automatically" learns the preferred #Entity just by simply inputting the pure (plain-text) Examples into the #Intent. In fact, that step was required. But we still need to do one more step.
After keying in the good plain-text Examples into the #Intent, we then still need to "right click" on the text-string of the possible #Entity entry, and then choose (teach Watson) the correct #Entity name from the dropdown list appeared.
Only then Watson starts to understand such; this #Intent uses that #Entity, I suppose.
Thank you #data_henrik, and appreciate your hint.
I have an Action where the user can set values of different parameters. Currently this is implemented something like this, and it works well:
Now I want to make the conversation less robot-like and more flexible, so I would like to allow users to set or change more than one value at a time. They should be able to say things like
Change the Interest Rate to 4% and the Term to 15 years.
or
Change the Interest Rate to 4%, the Term to 15 years, and the Years to Average Principal to 3.
There are a couple of ways to do this, but none of them are great, and all of them have issues of some sort when you try to scale them. (So they might work well for two or three parameters entered, but they probably won't work well for more than that.)
(It is worth noting, just for reference, that the Assistant itself has only recently started accepting more than one instruction at a time. But it only handles two, and this doesn't work for all commands.)
Add phrases with additional parameters
With this solution, you would supplement the phrases you have that collect one parameter with a similar set of phrases that collect two parameters. And then another set that also collect three parameters. You should be able to do these all as a single Intent and, in your fulfillment, determine which ones have been set.
It might look something like this:
That looks like it starts getting complicated, doesn't it? You need to list each combination of absolute values and percentages. If you have other types, you need to include each of those combinations as well. That starts getting unwieldy for 3 possible parameters, and certainly is above that. You also run the risk that it might get confused about which parameter should be set with which value (I haven't tested this - it is a theoretical concern).
Add an optional continuation phrase and handle that recursively
You can also treat this as the user saying "set a value, and then do something else" and treat the "do something else" part as another statement made to Dialogflow. The Intent might look something like this:
You can implement the "another statement made to Dialogflow" using the Dialogflow API. With Dialogflow V1, you'd use the Query endpoint. With Dialogflow V2, you'd use the detectIntent endpoint. In either case, you'd send the additional part of the query (if the user said something) and would get back the results from that. You'd add the resulting message from the call to the message from setting the current set of values and send the whole thing back.
As a recursive call, however, this does take up time. Since the initial call to Dialogflow really needs to be answered within 5 seconds, every additional call to Dialogflow (and then to your fulfillment) needs to be handled as quickly as possible. But even so, you probably won't be able to handle more than 2 or 3 of these before things time out on the front end.
It also runs the risk (or benefit) that other intents besides the edit.attribute Intent might be called in the "additional" portion. If you want to limit the risk of this, you could set a context to make sure that only Intents that have that incoming context would be called.
Summary
This really isn't an easy problem to solve. On one hand, you have the problem of having to list out every combination. On the other hand, recursion takes time, and you don't have a lot of time to process everything. In both cases, there is a real possibility of the phrase being understood incorrectly and you'll need to figure out error handling in the case where some values have been changed and others haven't.
You may need to experiment a lot, and the results may still not be satisfactory.
You can implement the "another statement made to Dialogflow" using the
Dialogflow API. With Dialogflow V1, you'd use the Query endpoint.
With Dialogflow V2, you'd use the detectIntent endpoint. In either
case, you'd send the additional part of the query (if the user said
something) and would get back the results from that. You'd add the
resulting message from the call to the message from setting the
current set of values and send the whole thing back.
As a recursive call, however, this does take up time. Since the
initial call to Dialogflow really needs to be answered within 5
seconds, every additional call to Dialogflow (and then to your
fulfillment) needs to be handled as quickly as possible. But even so,
you probably won't be able to handle more than 2 or 3 of these before
things time out on the front end.
The first thing that came to mind after reading those two paragraphs was batch requests.
A batch request allows a client application to pack multiple API calls into a single HTTP request (this batching technique is also known as a multi-part request).
Many Google APIs support a batch endpoint and I was able to verify that DialogFlow has a batch endpoint by checking its API Discovery document. This batch endpoint is not formerly documented in DialogFlow's API reference but you can leverage the documentation of other APIs (like this one) to get a feel for how it works. This link should also be instructive now that the global batch endpoint is no longer supported.
Assuming your queries are independent (ie. they don't rely on the results of other queries) you should be able to use a batch request to fetch more data.
Looking at all the examples of Operational Transformation Frameworks out there, they all seem to resolve around the transformation of changes to plain text documents. How would an OT framework be used for more complex objects?
I'm wanting to dev a real-time sticky notes style app, where people can co-create sticky notes, change their positon and text value. Would I be right in assuming that the position values wouldn't be transformed? (I mean, how would they, you can't merge them right?). However, I would want to use an OT framework to resolve conflicts with the posit-its value, correct?
I do not see any problem to use Operational Transformation to work with Complex Objects, what you need is to define what operations your OT system support and how concurrency is solved for them
For instance, if you receive two Sticky notes "coordinates move operation" from two different users from same 'client state', you need to make both states to converge, probably cancelling out second operation.
This is exactly the same behaviour with text when two users generate two updates to delete a text range that overlaps completely, (or maybe partially), the second update processed must be transformed against the previous and the resultant operation will only effectively delete a portion of the original one, (or completely cancelled with a 'no-op')
You can take a look on this nice explanation about how Google Wave Operational Transformation works and guess from this point how it should work your own implementation
See the following paper for an approach to using OT with trees if you want to go down that route:
http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.100.74
However, in your particular case, I would use a separate plain text OT document for each stickynote and use an existing library, eg: etherPad, to do the heavy lifting. The positions of the notes could then be broadcast on a last-committer-wins basis.
Operation Transformation is a general technique, it works for any data type. The point is you need to define your transformation functions. Also, there are some atomic attributes that you cannot merge automatically like (position and background color) those will be mostly "last-update wins" or the user solves them manually when there is a conflict.
there are some nice libs and frameworks that provide OT for complex data already out there:
ShareJS : library for Node which provides all operations on JSON objects
DerbyJS: framework for NodeJS, it uses ShareJS for OT stuff.
Open Coweb framework : Dojo foundation project for cooperative web applications using OT
Today I've been presented with a fun challenge and I want your input on how you would deal with this situation.
So the problem is the following (I've converted it to demo data as the real problem wouldn't make much sense without knowing the company dictionary by heart).
We have a decision table that has a minimum of 16 conditions. Because it is an impossible feat to manage all of them (2^16 possibilities) we've decided to only list the exceptions. Like this:
As an example I've only added 10 conditions but in reality there are (for now) 16. The basic idea is that we have one baseline (the default) which is valid for everyone and all the exceptions to this default.
Example:
You have a foreigner who is also a pirate.
If you go through all the exceptions one by one, and condition by condition you remove the exceptions that have at least one condition that fails. In the end you'll end up with the following two exceptions that are valid for our case. The match is on the IsPirate and the IsForeigner condition. But as you can see there are 2 results here, well 3 actually if you count the default.
Our solution
Now what we came up with on how to solve this is that in the GUI where you are adding these exceptions, there should run an algorithm which checks for such cases and force you to define the exception more specifically. This is only still a theory and hasn't been tested out but we think it could work this way.
My Question
I'm looking for alternative solutions that make the rules manageable and prevent the problem I've shown in the example.
Your problem seem to be resolution of conflicting rules. When multiple rules match your input, (your foreigner and pirate) and they end up recommending different things (your cangetjob and cangetevicted), you need a strategy for resolution of this conflict.
What you mentioned is one way of resolution -- which is to remove the conflict in the first place. However, this may not always be possible, and not always desirable because when a user adds a new rule that conflicts with a set of old rules (which he/she did not write), the user may not know how to revise it to remove the conflict.
Another possible resolution method is prioritization. Mark a priority on each rule (based on things like the user's own authority etc.), sort the matching rules according to priority, and apply in ascending sequence of priority. This usually works and is much simpler to manage (e.g. everybody knows that the top boss's rules are final!)
Prioritization may also be used to mark a certain rule as "global override". In your example, you may want to make "IsPirate" as an override rule -- which means that it overrides settings for normal people. In other words, once you're a pirate, you're treated differently. This make it very easy to design a system in which you have a bunch of normal business rules governing 90% of the cases, then a set of "exceptions" that are treated differently, automatically overriding certain things. In this case, you should also consider making "?" available in the output columns as well.
One other possible resolution method is to include attributes in each of your conditions. For example, certain conditions must have no "zeros" in order to pass (? doesn't matter). Some conditions must have at least one "one" in order to pass. In other words, mark each condition as either "AND", "OR", or "XOR". Some popular file-system security uses this model. For example, CanGetJob may be AND (you want to be stringent on rights-to-work). CanBeEvicted may be OR -- you may want to evict even a foreigner if he is also a pirate.
An enhancement on the AND/OR method is to provide a threshold that the total result must exceed before passing that condition. For example, putting CanGetJob at a threshold of 2 then it must get at least two 1's in order to return 1. This is sometimes useful on conditions that are not clearly black-and-white.
You can mix resolution methods: e.g. first prioritize, then use AND/OR to resolve rules with similar priorities.
The possibilities are limitless and really depends on what your actual needs are.
To me this problem reminds business rules engine where there is no known algorithm to define outputs from inputs (e.g. using boolean logic) but the user (typically some sort of administrator) has to define all or some the logic itself.
This might sound a bit of an overkill but OTOH this provides virtually limit-less extension capabilities: you don't have to code any new business logic, just define a new rule set.
As I understand your problem, you are looking for a nice way to visualise the editing for these rules. But this all depends on your programming language and the tool you select for this. Java, for example, has JBoss Drools. Quoting their page:
Drools Guvnor provides a (logically
centralized) repository to store you
business knowledge, and a web-based
environment that allows business users
to view and (within certain
constraints) possibly update the
business logic directly.
You could possibly use this generic tool or write your own.
Everything depends on what your actual rules will look like. Rules like 'IF has an even number of these properties THEN' would be painful to represent in this format, whereas rules like 'IF pirate and not geek THEN' are easy.
You can 'avoid the ambiguity' by stating that you'll always be taking the first actual match, in other words your rules have a priority. You'd then want to flag rules which have no effect because they are 'shadowed' by rules higher up. They're not hard to find, so it's something your program should do.
Your interface could also indicate groups of rules where rules within the group can be in any order without changing the outcomes. This will add clarity to what the rules are really saying.
If some of your outputs are relatively independent of the others, you will also get a more compact and much clearer table by allowing question marks in the output. In that design the scan for first matching rule is done once for each output. Consider for example if 'HasChildren' is the only factor relevant to 'Can Be Evicted'. With question marks in the outputs (= no effect) you could be halving the number of exception rules.
My background for this is circuit logic design, not business logic. What you're designing is similar to, but not the same as, a PLA. As long as your actual rules are close to sum of products then it can work well. If your rules aren't, for example the 'even number of these properties' rule, then the grid like presentation will break down in a combinatorial explosion of cases. Your best hope if your rules are arbitrary is to get a clearer more compact presentation with either equations or with diagrams like a circuit diagram. To be avoided, if you can.
If you are looking for a Decision Engine with a GUI, than you can try this one: http://gandalf.nebo15.com/
We just released it, it's open source and production ready.
You probably need some kind of inference engine. Think about doing it in prolog.