Botnation's NLP new version : Artificial Intelligence for everyone !

The Botnation team has been working for months to bring you a major evolution of the language processing algorithm of your chatbots.

We are very pleased to unveil the following updates :

There's no need to search variants to conjugate a verb, you only need to minimalise the NLP rule by simply indicating :

Examples :

On a user request : "I want a flat"

You needed to set up :

Before (v1)

[wants flat]

Now (v2)

[want flat]

For : "She wanted to know if you rent flats"

Before (v1) :

[wanted flat]

Now (v2)

[want flat]

>> As you see, the rule is strictly the same for those 3 examples on the v2 whereas you needed 3 rules on v1 for the very same result. It applies for every conjugation.

Need a clue to simplify your life ?

In the recommendation tool, after the user typed anything, an information is indicated in [...]. This will allow you to add relevant rules and expressions.

Example :

The user typed : "Do you sell bikes for long trips ?"

If no rule matched this request, then the tool suggest :

[sell bikes trips]

All you have to do is enter the expression "sell bike" or, more specifically, "sell bike trips" in your NLP rule to respond positively to this request. And the magic of the algorithms means that it will also respond positively to requests such as :

"Would you sell an affordable bike for a trip this week-end ?"

"Yesterday, you sold me 2 long trip bikes, I have a question on warranty please"

In those 3 cases, Botnation's NLP will take in real time the infinitive verb ("sell") and the singular words ("bike" and "trip"), despite the conjugation.

Set up v1 <> v2 Botnation NLP

In your chatbot's advanced settings, you can now find an option allowing you to use v1 or v2 for the NLP.

Every chatbot already made on Botnation are automatically set up in v1, including models. We then ensure a compatibility for every chatbot already existing on the platform.

The new chatbots and future templates are now set to NLP v2 by default.

Nothing prevents you from changing the configuration of an already existing chatbot from v1 to v2, but its operation may require many simplifications of your rules. We recommend you not to modify the configuration of an already operational chatbot but rather to duplicate this chatbot. On this duplicated version, change the configuration and then modify your NLP rules while testing in depth the new behaviour of these rules before publishing this duplicated version of your chatbot.

On the other hand, if you prefer the operation of v1, out of habit or curiosity, nothing prevents you from modifying the configuration of a new chatbot by switching it from v2 to v1 on this same settings screen.

Ah... "One more thing." :

Honestly, the libraries on Botnation are fantastic, but they used to be only usable as simple expressions. This could be problematic and often generated positive rules on a single word.

Nothing prevents you from using complex expressions, composed of several words, negative or outgoing words, in the libraries but we noticed that they are often used as simple lists of synonyms (auto, car, automobile, vehicle etc...).

Now you can place a library at any stage of an expression, such as :

"sell @car Paris"

The equivalent will be :

"sell vehicule Paris"

"sell SUV Paris" etc...

In the library menu, I generate a library "dog"

I can now integrate this library in a rule :

And it will answer, magically, positively to requests such as :

And Botnation keeps on correcting your users' mistakes !

A single constraint, we limit the use of libraries to only one library per expression. If you specify the use of a second library in the same expression, it will not be processed and will generate a skewed rule, based on the name of the library, which will certainly never be positive.

So, avoid :

"sell @car Paris @brand" or "sell @dog @bird"

As a reminder, here's our guide to libraries.

To put it simply, a tunnel on the NLP Botnation relaunches the entire search on a new context in a transparent way.

The exit trigger "_" allows priority redirection to a sequence, and therefore to a possible context. The tunnel operator "=" goes much further by allowing a search on the whole user's request in the affected context. The chatbot can then contextualise the discussion and continue it without interruption.

Example:

(general context)

Chatbot : What's your car brand ?

Utilisateur : I have Zenault

(context Zenault via "=zenault" but default message as there is no extra information in the user's request)

Chatbot : Zenault is excellent, it's French ! No second vehicule ?

User : Yes, I do have a Kiat but the gearbox is dead.

(context Kiat via "=kiat" which triggers the sequence linked to the expression "gearbox" in the context Kiat by relaunching the search through the trigger)

Chatbot : On Kiat, the gearbox in sensitive, especially for pre-2015 models, is it your case ?

In short, the tunnels allow second-generation relaunches without having to go through a precision request or foresee all the cases in the general context, you just have to isolate them well for each context. This is fantastic for cases like :

"I have a problem with my credit card" -> Context CB

"I have a problem with my credit card, I got it stolen" -> Context CB -> Stealing

"My cat is sich" -> Context Cat

"my cat is sick its thongue is green" -> Context Chat -> Contact veterinary

Contexts are not generalized on every Botnation's chatbot. We think it's an essential feature to create a performing chatbot and every user deserves it !

Here's a link to get to know more on contexts in your chatbot.

We gathered all the AI and NLP dedicated guides on Botnation in one and only highly structured document. To ease the process, so as you are able to read it at the office, in the subway, at the beach, we adapted the guide in a PDF version that be downloaded below.

All you have to do is to make your chatbot evolve and we are always available to answer your questions and assist you, so don't hesitate to contact us on the interface. Our team is already working on the next evolutions of Botnation... you'll love them !