Graphlogic.ai documentation

How to create and configure Slots

Intent Recognition | NLU slot + Intent and Fallback subslots

Purpose and general information

⁠

@NLU⁠

is a

@Slot⁠

designed to implement speech recognition by the

@Agent⁠

— processing messages in natural language in order to determine the expressed intention (

@Intent⁠

) of the

@Bot User⁠

and, depending on it, move to the appropriate

@Script Branch⁠

⁠

@NLU⁠

@Slot⁠

is a

@Complex Slot⁠

and cannot exist without its

@Child Slot⁠

@intent⁠

and

@fallback⁠

@Subslot⁠

s, which begin their own

@Script Branch⁠

es to which the transition will be made once the corresponding

@Intent⁠

is recognized.

The

@NLU⁠

@Slot⁠

processes the incoming

@Bot User⁠

‘s

@Request⁠

previously recorded in the

@client_message⁠

@Context Variable⁠

@NLU⁠

@Slot⁠

recognizes only those

@Intent⁠

s that are enclosed in its

@intent⁠

@Subslot⁠

s; the remaining

@Intent⁠

s existing in the

@Company⁠

do not participate in recognition in this

@NLU⁠

@Slot⁠

. For example, in the image above, the

@NLU⁠

@Slot⁠

will be trained to recognize only the two

@Intent⁠

s that are selected in its

@intent⁠

@Subslot⁠

s. If none of these

@Intent⁠

s are recognized, the

@Conversation⁠

will move to the

@fallback⁠

@Subslot⁠

Slot creation and settings

Slot attributes

⁠

Name — the name of the

@Slot⁠

, which will be displayed in the

@Script Tree⁠

. The maximum length of a field value is 40 characters.

Confidence Threshold — the minimum value of confidence in a recognized

@Intent⁠

, expressed as a percentage, at which the

@Intent⁠

is considered recognized and the

@Conversation⁠

can go to its

@Script Branch⁠

(the

@fallback⁠

@Subslot⁠

@Script Branch⁠

). The value can be an integer from 0 to 100.

It is recommended to set the

@Confidence Threshold⁠

value based on the statement that this value should be inversely proportional to the number of

@Intent⁠

s in the

@Slot⁠

: the more

@Intent⁠

s, the lower the

@Confidence Threshold⁠

should be. To begin with, it is recommended to set the value to 15 and change this value depending on the results of testing the

@Agent⁠

. Read more:

Testing and improving Agent recognition: selection of Confidence threshold⁠

Subslot intent attributes

⁠

Name — the name of the

@Slot⁠

, which will be displayed in the

@Script Tree⁠

. The maximum length of a field value is 40 characters. If a name is not entered, the

@Platform⁠

automatically copies the

@Intent⁠

name from the Intent field to the Name field.

Intent —

@Intent⁠

, upon recognition of which

@Conversation⁠

will go along the

@Script Branch⁠

following this

@Subslot⁠

SHOW INTENT IN A NEW TAB button — by clicking the button, the

@Intent⁠

selected in the field opens in the next browser tab.

Subslot fallback attributes

⁠

Name — the name of the

@Slot⁠

, which will be displayed in the

@Script Tree⁠

. The maximum length of a field value is 40 characters. When creating an

@NLU⁠

@Slot⁠

, the Name field of the

@fallback⁠

@Subslot⁠

is automatically filled with the contents of the Name field of the

@NLU⁠

@Slot⁠

. If the contents of the Name field are removed, the

@Subslot⁠

in the

@Script Tree⁠

will appear without a name.

Subslots intent and fallback

Automatic generation of subslots when creating an NLU slot

When creating an

@NLU⁠

@Slot⁠

, two

@intent⁠

@Subslot⁠

s and a

@fallback⁠

@Subslot⁠

are automatically generated. If there are no

@Intent⁠

s in the

@Company⁠

, then

@Subslot⁠

s with automatically generated

@Intent⁠

s will be created behind the

@NLU⁠

@Slot⁠

⁠

These

@Intent⁠

s will appear in

@Company Resources⁠

. About creating

@Intent⁠

Intents tab⁠

You can replace

@Intent⁠

s in generated

@Subslot⁠

s by expanding the

@Subslot⁠

by double-clicking the mouse, clicking on the name of the

@Intent⁠

and selecting the desired one from the drop-down list and clicking the Save button.

⁠

Adding Intent Subslots

To add a new

@Subslot⁠

, you need to click on the + button. The new

@Subslot⁠

will appear in the

@Script Tree⁠

above the

@Subslot⁠

, which is always located at the bottom.

⁠

Removing Intent and Fallback subslots

To remove an

@intent⁠

@Subslot⁠

, you need to right-click on it and click the Remove Branch button, then click OK in the pop-up dialog box.

⁠

It is possible to delete any

@intent⁠

@Subslot⁠

if, after deletion, the

@NLU⁠

@Slot⁠

remains at least two

@intent⁠

@Subslot⁠

Removing one of the last two

@intent⁠

@Subslot⁠

s is not possible. When you try to do this, the following error will appear at the bottom center:

⁠

Removing the

@fallback⁠

@Subslot⁠

is not possible. When you try to do this, the following error will appear at the bottom center:

⁠

Slot operation

The

@NLU⁠

@Slot⁠

is trained on

@Training Dataset⁠

@Intent⁠

s added as its

@Subslot⁠

@intent⁠

. About filling

@Training Dataset⁠

Natural language recognition⁠

The

@NLU⁠

@Slot⁠

receives the

@client_message⁠

@Context Variable⁠

as input, processes it, recognizes the

@Intent⁠

, and transfers the

@Conversation⁠

to the recognized

@Intent⁠

@Script Branch⁠

. If no

@Intent⁠

is recognized with the proper level of confidence, the

@Conversation⁠

goes to the

@fallback⁠

@Subslot⁠

Intent recognition process

Training: The

@NLU⁠

@Slot⁠

is trained on samples of

@Intent⁠

s added as its

@intent⁠

@Subslot⁠

Important: if the
@Agent⁠
was edited, but the changes affected only the contents of other
@Slot⁠
s, in addition to
@NLU⁠

@Slot⁠
, then when the
@Agent⁠
is retrained, the
@NLU⁠

@Slot⁠
will not be retrained.

Receiving a

@Request⁠

: the

@NLU⁠

@Slot⁠

receives the

@Bot user’s Message⁠

— the text in the

@client_message⁠

@Context Variable⁠

Classification of the

@Request⁠

: the message is classified from the

@client_message⁠

@Context Variable⁠

and arranges all

@Intent⁠

s of this

@NLU⁠

@Slot⁠

according to the level of confidence in them — confidence (expressed as a percentage, the same as CONFIDENCE THRESHOLD).

Creation of the

@nlu_result⁠

@Context Variable⁠

in the

@Chat Context⁠

: the

@nlu_result⁠

@Context Variable⁠

is created, the value of which becomes a json object containing information about the top 5 recognized

@Intent⁠

s in descending order of their confidence.

Example: [{'intent_name': 'size', 'confidence': 0.9993743300437927}, {'intent_name': 'product_in_stock', 'confidence': 0.00019508253899402916}, {'intent_name': 'model_price', 'confidence': 0.00012844 608863815665} , {'intent_name': 'address', 'confidence': 0.00010525238758418709}]

Transition to

@intent⁠

@Intent⁠

with maximum confidence, provided that this confidence >= CONFIDENCE THRESHOLD from the settings of this

@NLU⁠

@Slot⁠

, is the

@Intent⁠

by which the

@Request⁠

is recognized. Consequently:

The

@Conversation⁠

will go to the

@Script Branch⁠

of this

@intent⁠

@Subslot⁠

;

the

@intent_name⁠

@Context Variable⁠

appears in the

@Chat Context⁠

ー the name of the recognized

@Intent⁠

;

the

@intent⁠

@Context Variable⁠

appears in the

@Chat Context⁠

— the identifier of the recognized

@Intent⁠

;

the

@confidence⁠

@Context Variable⁠

appears in the

@Chat Context⁠

— the level of confidence that the

@Request⁠

belongs to a specific

@Intent⁠

, expressed as a percentage.

The

@intent⁠

and

@confidence⁠

@Context Variable⁠

s exist in the

@Chat Context⁠

until the next execution of the same or another

@NLU⁠

@Slot⁠

in this

@Agent⁠

Transition to

@fallback⁠

: If the confidence of all

@Intent⁠

s is below CONFIDENCE THRESHOLD from the settings of this

@NLU⁠

@Slot⁠

, then it is considered that not a single

@Intent⁠

was recognized, as a consequence:

The

@Conversation⁠

will go to the

@fallback⁠

@Subslot⁠

@Script Branch⁠

;

the

@error⁠

@Context Variable⁠

appears in the

@Chat Context⁠

— the error text. This

@Context Variable⁠

exists in the

@Chat Context⁠

until:

next execution of the same or another

@NLU⁠

@Slot⁠

;

⁠

@Regular Expression Slot⁠

execution;

⁠

@External Request⁠

@Slot⁠

execution.

Available languages to recognize by NLU Slot

Available languages to recognize by NLU Slot

Language

Abbreviation in the dropdown list

Russian

English

Portuguese

Ukrainian

Spanish

Vietnamese

Thai

Indonesian

Malay

Filipino

Serbian

German

Polish

French

Pakistani (urdu)

Kyrgyz

Arabian

There are no rows in this table

⁠

Purpose and general information

Slot creation and settings

Slot attributes

Subslot intent attributes

Subslot fallback attributes

Subslots intent and fallback

Automatic generation of subslots when creating an NLU slot

Adding Intent Subslots

Removing Intent and Fallback subslots

Slot operation

Intent recognition process

Available languages to recognize by NLU Slot

Want to print your doc?
This is not the way.

Try clicking the ⋯ next to your doc name or using a keyboard shortcut (

CtrlP

) instead.