Query Language

What queries consist of

The queries in the query language of Artefact are built from tokens, which may serve as operands, operators, parentheses, or modifiers. Parentheses are used for specifying the order in which the phrases appearing in the query are to be evaluated, whereas modifiers impose context restrictions on the documents that are searched for. It is also possible to specify document fields to be searched for the specified operands.

Operands

• A word may be an arbitrary word of the natural language in which the documents kept in the database have been written. It is a distinctive feature of Artefact that the words appearing in queries may be given in any grammatical form (in singular or plural, as well as in any case, gender, and declension/conjugation).

  For instance:

  taxes laws privileges computer video

• A word starting with a capital letter. The documents will be searched for the occurrences of that word starting with a capital letter.

  For instance:

  Clinton Yeltsin

• A right-truncated word is the beginning of a word followed by the character *. If a query consists of a single right-truncated word X*, it means that all documents in the database that contain words beginning with letters X are relevant to the query.

  For instance:

  diplomat* perform*

  If the query consists of the word diplomat*, then all documents containing the words diplomat, diplomats, diplomatic, etc. will be relevent to the query.

• A constrained right-truncated word is the beginning of a word followed by the characters !*, which are followed by a digit. If a query consists of a single constrained right-truncated word X!*D, it means that all documents in the database containing words of the form XY, where Y is any sequence containing no more that D letters, are relevant to the query.

  For instance, diplomat!*2 will be matched by diplomats, diplomatic, but not by diplomatist.

• A left-truncated word is the character * followed by the "rest" of the word to be searched for. *X will be matched by the words in the database that end in X.

• A left- and right-truncated word is the character * followed by the "middle" of the word, followed by either * or !* and a digit. *X* will be matched by the words in the database that contain X, and *X!*D by the words that contain X followed by no more that D letters.

• A word with a modifier is any natural language word followed by the character ! and one of the following options: s or p.

  Option s defines a "spelling search", i.e. a search for a misspelled word. The answers to this operand will show not only documents containing the required word but also documents containing related misprints.

  Option p defines a search for the "large paradigm" of the word for all words with the same root.

• A word in double quotes. A word, containing special symbols, such as parenthesis, must be enclosed in double quotes when written in queries. For example, the word not must be written as "not" in order to avoid confusion with the operator NOT.

• A date, which may be specified with the accuracy either up to a year, up to a month, or up to a day followed by !d or !д.
  • A date accurate up to a year, for instance 1990!d, will be matched by the documents containing any dates that belong to the year 1990.
  • A date accurate up to a month, for instance 02.1994!d will be matched by the documents containing dates that belong to February of the year 1994.
  • A date accurate up to a day, for instance 06.02.1994!d will be matched by the documents containing that date.
  • A date interval consists of two dates separated by -: first_date-last_date, for instance 22.07.1941-08.05.1945!d. The first and last dates may be specified up to any accuracy.specifying a search for all dates that are "less then" Artefact has no means for or "greater than" a given date. But, inpractice, the same effect can be achieved by specifying a suitable date interval.

• A template containing wildcards, which may specify a set of words to be searched for in the documents.

  A template may contain:

  • any characters that may appear in the documents, except for * and ?
  • the character *, which indicates that any character sequence may occupy that position in the word appearing in the document
  • the character ?, which indicates that any character may occupy that position in the word appearing in the document

  Wildcards are useful in cases where we either are not sure about the correct spelling of the word, or know only some part of the word.

  For example: shevar*nadze, *05-??, etc.

Operators

Operator AND

Aliases: and И и.

The AND operator combines its operands, which can be words and/or expressions enclosed in parentheses. If AND combines two words A and B, then the documents relevant to the phrase A AND B will be those containing both A and B. The documents relevent to the phrase A AND B AND C will be those containing all three words A, B, and C. In other words, AND requires that all its operands appear in a document, for it to be relevant to the query.

The AND operator can be omitted, so that the following phrases are equivalent:

Operator OR

Aliases: or ИЛИ или.

The OR operator combines its operands, which can be words and/or expressions enclosed in parentheses. If OR combines two words A and B, then the documents relevant to the phrase A OR B will be those containing either only A, or only B, or both A and B.

The operators AND and OR have the same(!) precedence, which means that the order in which they take effect is determined only by parentheses.

Sequencing operator :

The : operator requires that its operands appear in a sentence in a document in the specified order. The words that match the operands may be separated in the document by arbitrary sequences of words. It is possible to limit the number of separating words.

For instance:

tax :3 (export OR import)

Consider the query

(repairing OR selling) computers

It means that the document must contain a sentence that contains either repairing, or selling, or both repairing and selling, followed by computers, which may be separated by an arbitrary sequence of words. A similar query

(repairing OR selling) :2 computers

means that the document must contain a sentence that contains either repairing, or selling, or both repairing and selling, followed by computers, which may be separated by zero, one or two words.

Field search

For the search within a specified field (group of fields) in the document, the operand or query phrase in parenthesis must be preceeded with a "/" and field name (names).

For example:

/date 09.05.1996!d /title /text (Yeltsin : visit : USA)

Specifiers of the relevant context

Except for the queries containing sequencing operators, all queries we have so far considered specified conditions imposed on whole documents, i.e. the queries required that some words should appear in a document, without paying attention to the distances between those words in the document. Sometimes, if a database consists of large documents, this can result in too many documents being selected, a phenomenon that is often referred to as "search noise". This noise can be reduced by restricting the context in which the operands should appear in the document.

Modifiers

The context in which the search for words must be performed can be specified by modifiers. The search can be restricted to

• A specified number of adjacent sentences in the document: /s /sN /п /пN (the forms are equivalent), where N is the number of adjacent sentenses (the absence of N means 1 sentence).

  For instance:

  tax export /s
  tax (export OR import) /s2

• a specified number of adjacent words in a sentence in the document: /wN /WN /сN /СN (the forms w W с С are equivalent, where N is the number of adjacent words in the context, с and С are the Cyrillic "с").

  For instance:

  tax (export OR import) /w5

The scope of a modifier in the query

The scope of a modifier appearing at the top level is the whole query. Otherwise, if a modifier appears inside a pair of parentheses, its scope is the part of the query enclosed in the parentheses. This is illustrated by the following picture, where A, B, C, ... stand for words, and /a, /b, /c, for some modifiers

If several /wN or /sN modifiers have the same scope, only the lastof them takes effect.

If several /f modifiers have the same scope, all of them takeeffect: the context will include all the fields thus specified.

The character / in modifiers can always be replaced with \, without any change in their meaning.