Rule syntax

• How do Deep categorization models work?
• Dashboard
  • Predefined models
  • Create a model
  • Add an example model
• Models
  • Settings
• Categories
  • Create and edit categories
• Rules
  • Rule syntax
  • Test parsing
• Macros
  • Create and edit macros
• Build/Test a model
• Import/Export a model
• Model view
• Using a model in your application

In this section, we will introduce the specific language to create rules, which will be divided into seven sections:

• Basic elements
• Morphological Aspects
• Semantic Aspects
• Logical Operators
• Context Operators
• Distance Operators
• Regular Expressions
• Other Elements

After we've seen all the possible operators that can be used with the elements we can define, we will see some examples that will illustrate how they can be combined.

Please note that rules always apply in the defined context for the model. This is determined by the split sentences parameter in the model settings. If split sentences is enabled, the engine will try to match each rule against each sentence in the text.

Basic elements

There are three types of elements:

• Simple term: The semantic unit represented by a single word, it is processed as one element by the system.
• Parsing multiword: The semantic unit made of two or more words joined by underscores. Two or more words will have to be defined as a parsing multiword if the system considers them as a single element. You can check the elements returned by the system using the test parsing console.
• Multiword or Literal expression (word-to-word): They convey exactly what is written between quotation marks, that is, those specific words in that same order.

The following table shows a few examples of each element:

Element	Rule	Example
Simple term	tree	"Trees have a trunk, with supporting branches and leaves in most species."
Simple term	butterfly	"Adult butterflies have large and brightly coloured wings."
Simple term	play	"My son usually plays football on Sunday." "I've never seen such a good play."
Parsing multiword	El_Salvador	"El Salvador is the smallest and the most densely populated country in Central America."
Parsing multiword	scuba_diver	"Scuba diving equipment allows you to visit the underwater world." "The scuba diver spent the whole day practicing the new technique."
Parsing multiword	vice_president	"The vice president is called the deputy president"
Literal expression (multiword)	"sense of direction"	"He has no sense, he gave you the wrong direction." "He has no sense of direction."
Literal expression (multiword)	"tech|technical|product support|assistance|service"	"The service support was fantastic." "The technical support was fantastic." "The product offers no support." "They offer great product support."
Literal expression (multiword)	"I get lost|confused"	"I often get lost." "You get confused easily." "I get confused easily."

Morphological aspects

1. Form and lemma

A lemma is the canonical form, dictionary form, or citation form of a set of words. In order to specify this feature in the model, use "L@" right before the term (L@term). In that way, the rule will take all possible options within that lemma into account. At the same time, homonymous words that come from different lemmas - which have different meanings – will not be considered (e.g. 'heading' includes both the gerund of the verb ‘to head” and the noun; L@heading includes the noun in its singular and plural forms, but not the verbal form).

Forms are different morphological variations of a given lemma. In order to specify this feature in the model, use "@F" right before the term (F@term). If a form is used, the rule will only work for that specific form.

Element	Example
play	"I usually play football on Sunday." "I will play football next weekend." "He is playing football right now." "I've never seen such a good play."
F@play	"I usually play football on Sunday." "I will play football next weekend." "He is playing football right now." "I've never seen such a good play."
double_tap	"She gave me a double tap." "I double tapped my new friend on Instagram."
F@double_tap	"She gave me a double tap." "I double tapped my new friend on Instagram."
"Good|bad|great promotion|F@deals"	"I always get good deals." "That was a great deal"
L@heading	"Please check the headings before publishing." "She's heading home right now."

2. Parts of speech

There are different grammar features that can be specified according to part of speech. This will help us disambiguate the context of the rule we are defining according to the grammatical function a specific word or words carry out in the text. To specify the grammar aspect, we just have to add right after the term @ and the letter that represents the aspect.

These are the possible features available:

Aspect	Description	Rule	Example
@V	Verb	smash@V	"He heard a smash of glass." "The thief smashed the window."
@V1	First person of the verb	eat@V1	"James will eat as much as possible." "I eat around noon."
@V2	Second person of the verb	eat@V2	"They eat together on Tuesdays." "You are eating too much."
@V3	Third person of the verb	eat@V3	"You can play after you eat." "His father is eating by himself."
@V-	Non-personal forms of the verb	eat@V-	"You can play after you eat." "Eating is necessary."
@N	Noun	advert@N	"I have already adverted to the solar revolution." "The adverts you see are brand new."
@A	Adjective	mean@A	"That doesn't mean the same thing." "He's a very mean person."
@E	Adverb	"worked hard@E"	"He worked hard jobs all his life." "He worked hard all his life."
@T	Article	A@T	"The next chord is A." "A guitar is an instrument."
@D	Demonstrative	"@D person|woman|man"	"The woman dropped the bag." "That man dropped the bag."
@M	Numeral	"@M euros|dollars"	"He has some euros left." "He owes me 10 euros."
@P	Personal pronoun	"@P can_withdraw|may_withdraw"	"He never will withdraw his support." "She can withdraw her consent."
@S	Possessive	"@S {family}"	"The family is coming." "His family is coming."
@Y	Preposition	"@Y people"	"The people he's sharing room with are strangers." "He feels comfortable among people of the same tastes."
@Q	Quantifier	"@Q car|bus|motorbike"	"Those cars are red." "Many cars are red."

All of these features can also be used by themselves to refer to any word that works as the chosen part of speech in the text. This is an example of how this works:

3. Negation feature

The negation feature + or - identifies if a term is affected by negation or not.

Rule	Example
+enthusiastic	"I am enthusiastic about going to the bowling alley tonight." "I am not enthusiastic about going to the bowling alley tonight."
-enthusiastic	"I am enthusiastic about going to the bowling alley tonight." "I am not enthusiastic about going to the bowling alley tonight."

Semantic aspects

Semantic information can be used to define rules, enabling you to group words by their meaning. Currently only the semantic information defined in MeaningCloud's default resources is supported. These are the possible values:

• Ontology entity type: it's accessed using S@[ontology type] and searches the internal semantic attribute sementity.
• Theme information: it's accessed using T@[theme type] and searches the internal semantic attribute semtheme.
• Geographic information: it's accessed using G@[semgeo] and searches the internal semantic attribute semgeo.

Semantic aspects are explained in more detail here.

Logical operators

Logical operators are used to associate the appearances of different terms using Boolean logic. There are three operators available: AND, OR, and AND NOT, (brackets should be used to define their precedence in a rule). Please note that these operators apply only on the pre-established context.

Context operators

Context operators define the context where the rules apply.

• Exclusion operator: The :: operator omits the following term from its preceding broader range. In other words, this operator excludes some terms from a larger list.

"S@Top>LivingThing>Animal::cat|dog|hamster"

It will detect all the words that fall into the "Animal" category except those that are specified afterwards.

"Her turtle's name is Vaca." "Her cat's name is Tortuga"

• Number of words operator: The WORDS operator delimits the number of words that the context must have in order for the rule to apply.

"woman AND WORDS<7"	The rule will apply if the context has less than seven words.	"Pretty woman walking down the street." "Pretty woman the kind I'd like to meet."
"woman AND WORDS>7"	The rule will apply if the context has more than seven words.	"Pretty woman I don't believe you, you're not the truth." "Oh, pretty woman."
"woman AND WORDS=7"	The rule will apply if the context has exactly seven words.	"Pretty woman stop awhile." "Pretty woman talk awhile." "Pretty woman give your smile to me."

Distance operators

Distances operators allow you to define co-appearances of several words in a specific range. They can include as many terms as you need, but the distance between the first and the last term must be specified. There are two options to be considered when incorporating distance operators:

• LENIENT: [termA termB]~number: Terms can appear in any order with the maximum distance (ascertained in the rule) between those defined that appear first and last.
• STRICT NEAR: [termA termB]-number. Terms must appear as dictated by the rule; in regard to, the order and the maximum distance between the first and last of the defined terms (order matters).

So, let's see how terms' count works with some examples:

Multiwords can also be used within square brackets to include a word-to-word expression:

Let's see some examples of these operators in action:

Rule	Text	Detects
[zone|space|section|area parking|child|children's|smoking|non-smoking|play|reserved]~5	The zone has a parking nearby.
[zone|space|section|area parking|child|children's|smoking|non-smoking|play|reserved]~5	The hotel parking is a space nearby.
[zone|space|section|area parking|child|children's|smoking|non-smoking|play|reserved]~5	The assigned parking is around the corner, in a space behind the hotel.
[speak fast|rapid|quick|quickly|slow]-2	He speaks fast.
[speak fast|rapid|quick|quickly|slow]-2	He speaks English very fast.
[speak fast|rapid|quick|quickly|slow]-2	He speaks very slowly.
[give|have "what I asked|ordered"]-3	They don't have what I ordered.
[give|have "what I asked|ordered"]-3	They don't have your purchase or what I ordered.

Regular expressions

Regular expressions are allowed. However, you will have to escape certain characters such as +, ( ), *, and % by using "\".

Below you can find some useful regular expressions:

Regex	Description	Rule	Detects
.*_term	Matches the chosen term inside a parsing multiword.	.*_president	The former president will visit us tomorrow.
[0-9.,]*	Matches any number, including decimal and whole numbers. A space after the number may appear or not.	"[0-9.,]* ?€"	The bill was 30€. The bill was 30.5€. The bill was 30,5€. The bill was 30,5 €.
[list of letters]	Matches any of the letters inside the square brackets.	apologi[sz]e	Please, apologize Please, apologise
?	The letter may appear or not.	behaviou?r	Good behavior Bad behaviour

Other Elements

There are two external elements that can be used in the rules:

• Macros: macros are referred to by their label between curly brackets, and can be used the same way a simple term would.

  {my_macro_label} AND words

• Categories: categories can also be referred to within the rules by using their code preceded by #. This is specially useful to deal with coappearances of several categories and disambiguate depending on the context.

  #catA AND context

Now that we've seen all the different operators and expressions that can be used in a rule, let's see some examples of how they can be combined.