# many-sorted language

To formalize the notion of a many-sorted language, we start with a non-empty set $S$ whose elements we call sorts. Let $S^{+}$ be the set of all (finite) non-empty words over $S$. Elements of $S^{+}$ are called sort types and are written as tuples. So instead of writing $s_{1}s_{2}\cdots s_{n}\in S^{+}$, it is written $(s_{1},s_{2},\cdots,s_{n})\in S^{+}$.

The next item to be defined is the underlying signature   of a many-sorted language. A signature $\Sigma$ consists of

• a non-empty set $S$ of sorts,

• a set $F$ of function symbols, and

• a set $R$ of (non-logical) relation symbols,

such that each element in $F\cup R$ corresponds to a sort type. In other words, there is a function $t:F\cup R\to S^{+}$, and for every $r\in F\cup R$, $t(r)$ is its sort type. An element $c\in F$ such that $t(c)\in S$ is called a constant symbol (of sort $t(c)$).

In addition  to the signature $\Sigma$, we introduce additional symbols:

Using $\Sigma$ and the additional symbols above, we can build terms inductively as follows:

• each variable $x\in V$ is a term of sort $v(x)$

• if $f\in F$ is a function symbol of sort type $(s_{1},\ldots,s_{n})$, and for each $i=1,\ldots,n-1$, $t_{i}$ is a term of sort $s_{i}$, then $f(t_{1},\ldots,t_{n-1})$ is a term of sort $s_{n}$.

• all the terms are built this way.

Finally, from the terms, we inductively define formulas  :

• if $t_{1}$ and $t_{2}$ are terms of the same sort, then $(t_{1}=t_{2})$ is a formula

• if $r\in R$ is a relation symbol of sort type $(s_{1},\ldots,s_{n})$, and for each $i=1,\ldots,n$, $t_{i}$ is a term of sort $s_{i}$, then $r(t_{1},\ldots,t_{n})$ is a formula

• if $\alpha$ is a formula, then so is $(\neg\alpha)$

• if $\alpha,\beta$ are formulas, then so is $(\alpha\vee\beta)$

• if $\alpha$ is a formula and $x\in V$ is a variable, then so is $(\forall x(\alpha))$

• all the formulas are formed this way.

The signature $\Sigma$, additional symbols, and terms and formulas subsequently defined constitute what is known as the many-sorted language $L=L(\Sigma)$ on $\Sigma$.

As in first order language, the outer most parentheses may be eliminated without causing much harm, so that $(\neg\alpha)$ becomes $\neg\alpha$. In addition, we may introduce other familiar logical symbols $\wedge,\to,\leftrightarrow$, and $\exists$ in terms of $\vee,\neg$, and $\forall$. The specifics of how this is done can be found in the entry on first order language.

From this definition, we see at once that the classical first order language is a one-sorted language  ($S=1$). Sort types are identified with their lengths. Thus, the sort type of a function or a relation symbol is its arity.

Remark. It is not hard to show that a many-sorted language is not much different from a first-order language. Provided that $V$ is countably infinite, a many-sorted language $L$ can be “converted” into a first-ordered language $L_{1}$. Basically, all the function and relation symbols in $L$ are in $L^{\prime}$, as well as the additional symbols such as variables and logical symbols. For each sort $s\in S$, we introduce a new unary relation symbol $P_{s}$ in $L_{1}$. For any formula that contains a subformula of the form $\forall x\phi(x)$, we replace each occurrence of such a subformula by a formula of the form $\forall x(P_{s}(x)\to\phi(x))$, where $x$ is a variable of sort $s$ and $\phi$ is some formula in which $x$ occurs as a free variable   . The result is that $L_{1}$ becomes a one-sorted language.

 Title many-sorted language Canonical name ManysortedLanguage Date of creation 2013-03-22 17:44:40 Last modified on 2013-03-22 17:44:40 Owner CWoo (3771) Last modified by CWoo (3771) Numerical id 12 Author CWoo (3771) Entry type Definition Classification msc 03B70 Classification msc 03C07 Classification msc 03B10 Synonym many sorted language Synonym many-sorted logic Defines sort Defines sort type