Graphlogic.ai documentation

How to create and configure Slots

Entity Recognition | Synonym slot

Purpose and general information

The

@Synonym Slot⁠

allows you to perform a semantic search for an

@Entity⁠

using a table of a certain format with data — a

@Vocabulary⁠

— and save the found information in a

@Context Variable⁠

. Semantic search — recognition of various word forms of synonyms, and not just complete matches with them — is available for the languages listed in the table

Broken link⁠

. For other languages, a complete match search works.

When expressing an intention in natural language, the

@Bot User⁠

can use an

@Entity⁠

in the

@Request⁠

. For example, in the

@Request⁠

“How to get to New York by train” there are

@Entity⁠

“city” (New York) and

@Entity⁠

“transport" (train). In some cases, you need to allocate

@Entity⁠

values and write them to a

@Context Variable⁠

for later use in the

@Script⁠

. The essence can be expressed in different words — synonyms. For example, “New York” can be called “NY” and “train” — “electric rain" or “railway”.

Some scenarios require you to extract

@Entity⁠

values and write them to a

@Context Variable⁠

for later use in the

@Script⁠

. To search for an

@Entity⁠

, the

@Synonym Slot⁠

is used in combination with the

@Vocabulary⁠

functionality.

⁠

Examples:

It is necessary to determine the full official name of the city to transfer it to the external system. For this, a

@Vocabulary⁠

is created, consisting of a number of

@Entity⁠

es, each of which corresponds to a specific city and contains various variations of its name, for example: the name of the

@Entity⁠

is “St. Petersburg”, the contents are synonyms “Saint Petersburg, SPB, Leningrad”. The

@Bot user’s Message⁠

is run by the

@Synonym Slot⁠

through this

@Vocabulary⁠

and the name

@Entity⁠

found in it is used further in the

@Script⁠

It is necessary to divide the

@Script⁠

after determining the Intent depending on the

@Entity⁠

contained in the Replica.

Intent: “How to find <

@Entity⁠

>?”.

Set of

@Entity⁠

es: “boss’s office”, “dressing room”, “meeting room”, “cafeteria”, “doctor’s office”, “kitchen”, “toilet”

Using the

@Synonym Slot⁠

, the

@Entity⁠

is searched for in the

@Bot user’s Message⁠

, then, using the

@Transition Rule⁠

@Slot⁠

, the

@Script⁠

is divided into several

@Script Branch⁠

s depending on the

@Entity⁠

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.

Source — an

@Expression⁠

that gives the result — a string (

@Test String⁠

) in which the search for

@Synonym⁠

s from the

@Vocabulary⁠

will be performed. Either a full

@Expression⁠

with operators or an

@Expression⁠

from a single

@Context Variable⁠

can be specified, for example {{ variable }} or {{ places[0].name }}

RESULT VARIABLE — name of the

@Context Variable⁠

into which the information found by the

@Slot⁠

will be written. Format: string without parentheses, for example var1.

VOCABULARY —

@Vocabulary⁠

from

@Resourses Tab⁠

. Incremental case-insensitive search in the list works. There can be several RESULT VARIABLE + VOCABULARY pairs. To add a new pair, click on the Add search button.

⁠

As each additional RESULT VARIABLE + VOCABULARY pair is added to the VOCABULARY field, the drop-down list displays those

@Company⁠

@Vocabulary⁠

s that are not used in other VOCABULARY fields of this

@Slot⁠

. Thus, the RESULT VARIABLE + VOCABULARY pairs in one

@Synonym Slot⁠

can be a maximum of 20 (20 is the maximum number of

@Vocabulary⁠

in the

@Company⁠

Using syntax in Slot

In the

@Synonym Slot⁠

, it is possible to use

@Expression⁠

and

@Expression with Control Structure⁠

in the Source field. Read more:

Syntax⁠

Vocabularies

The

@Vocabulary⁠

is a

@Company Resources⁠

object, so it is accessible from all

@Agent⁠

s of this

@Company⁠

About creating a

@Vocabulary⁠

Vocabularies tab⁠

Slot operation

When going through a

@Synonym Slot⁠

, the following operations are performed in the order listed:

Removing from the

@Chat Context⁠

all

@Context Variable⁠

s specified in the RESULT VARIABLE field of this

@Slot⁠

;

Calculation of the

@Expression⁠

specified in the Source field, obtaining the

@Test String⁠

If the

@Expression⁠

failed to evaluate and obtain the string (the

@Context Variable⁠

specified in the

@Expression⁠

does not exist or the

@Expression⁠

is incorrectly formed), then the

@Context Variable⁠

@error⁠

with the value Template rendering (source of synonym extraction) error occurred will be created in the

@Chat Context⁠

Converting the contents of the

@Test String⁠

to lower case;

Separating the contents the

@Test String⁠

by words using a special model (for example, by a space, by a comma, etc.) — obtaining an array of words;

For each completed RESULT VARIABLE - VOCABULARY pair: Search for each word from the

@Test String⁠

in the

@Vocabulary⁠

specified in the VOCABULARY field, and write the result to the

@Context Variable⁠

specified in the corresponding RESULT VARIABLE field:

If one or more words from the

@Test String⁠

match

with word forms of

@Synonym⁠

s for the languages listed in the table

@Vocabulary⁠

languages available for recognition

or with the exact meaning of the

@Synonym⁠

for any languages

Important: the case of
@Synonym⁠
s in the
@Vocabulary⁠
and the
@Test String⁠
does not affect the recognition result.

The value of the

@Context Variable⁠

specified in the RESULT VARIABLE field is an array of objects, where each element of the array is an object containing a positive word recognition result (positive result = the word was found in the

@Vocabulary⁠

The result format is JSON with information, in this example only the first element of the array of result objects is described: [ {'synonym': 'synonym found in variable', 'reference': 'first synonym in line', 'entity': 'entity name', 'start': serial number of the first character found in the synonym variable, 'end': serial number of the last character found in the synonym variable }, ... { <Nth element - search result> }, ]

Example: [ {'synonym': 'honda', 'reference': 'HONDA', 'entity': 'car_brand', 'start': 0, 'end': 5} ]

The array can have several elements, their number is equal to the number of matches found when checking in the

@Vocabulary⁠

, this is possible in the following cases:

one word from the

@Test String⁠

matches several in the given

@Vocabulary⁠

(found in several strings)

Example:
@Test String⁠
: I need two tickets to NY Recognition result: [ {'synonym': 'NY', 'reference': 'New York', 'entity': 'New York', 'start': 24, 'end': 25}, {'synonym': 'NY', 'reference': 'New York', 'entity': 'New York', 'start': 24, 'end': 25} ] The word NY was found in 2
@Entity⁠
in one
@Vocabulary⁠
.

several words from the

@Test String⁠

found in different

@Entity⁠

es in this

@Vocabulary⁠

Example:
@Test String⁠
: I want to buy tickets from LA to NY Recognition result: [ {'synonym': 'LA', 'reference': 'LA', 'entity': 'LA', 'start': 22, 'end': 23}, {'synonym': 'NY', 'reference': 'NY', 'entity': 'NY', 'start': 32, 'end': 33} ] The word LA was found in 1
@Entity⁠
in the
@Vocabulary⁠
, the word NY was found in 1
@Entity⁠
in the
@Vocabulary⁠
.

Several words from the

@Test String⁠

found in one

@Entity⁠

in this

@Vocabulary⁠

. For example, in a phrase one city is mentioned twice, each mention is a separate word that will give its own result object in the RESULT VARIABLE variable.

Example:
@Test String⁠
: buy tickets Moscow - St. Petersburg - Moscow [ {'synonym': 'Moscow', 'reference': 'Moscow', 'entity': 'Moscow', 'start': 14, 'end': 20}, {'synonym': 'Peter', 'reference': 'St. Petersburg', 'entity': 'St. Petersburg', 'start': 21, 'end': 26}, {'synonym': 'Moscow', 'reference': 'Moscow', 'entity': 'Moscow', 'start': 27, 'end': 33} ] The word Moscow matches the
@Entity⁠
'Moscow' in the
@Vocabulary⁠
and is present in the
@Test String⁠
twice.

If nothing is found, the

@Context Variable⁠

specified in the RESULT VARIABLE field will not be created in the

@Chat Context⁠

Purpose and general information

Slot creation and settings

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.