<!DOCTYPE article PUBLIC "-//NLM//DTD JATS (Z39.96) Journal Archiving and Interchange DTD v1.0 20120330//EN" "JATS-archivearticle1.dtd">
<article xmlns:xlink="http://www.w3.org/1999/xlink">
  <front>
    <journal-meta />
    <article-meta>
      <title-group>
        <article-title>Forward Countermodel Construction in Modal Logic K?</article-title>
      </title-group>
      <contrib-group>
        <contrib contrib-type="author">
          <string-name>Mauro Ferrari</string-name>
          <xref ref-type="aff" rid="aff2">2</xref>
        </contrib>
        <contrib contrib-type="author">
          <string-name>Camillo Fiorentini</string-name>
          <xref ref-type="aff" rid="aff0">0</xref>
        </contrib>
        <contrib contrib-type="author">
          <string-name>Guido Fiorino</string-name>
          <xref ref-type="aff" rid="aff1">1</xref>
        </contrib>
        <aff id="aff0">
          <label>0</label>
          <institution>DI, Univ. degli Studi di Milano</institution>
          ,
          <addr-line>Via Comelico, 39, 20135 Milano</addr-line>
          ,
          <country country="IT">Italy</country>
        </aff>
        <aff id="aff1">
          <label>1</label>
          <institution>DISCO, Univ. degli Studi di Milano-Bicocca</institution>
          ,
          <addr-line>Viale Sarca, 336, 20126, Milano</addr-line>
          ,
          <country country="IT">Italy</country>
        </aff>
        <aff id="aff2">
          <label>2</label>
          <institution>DiSTA</institution>
          ,
          <addr-line>Univ. degli Studi dell'Insubria, Via Mazzini, 5, 21100, Varese</addr-line>
          ,
          <country country="IT">Italy</country>
        </aff>
      </contrib-group>
      <abstract>
        <p>The inverse method is a saturation based theorem proving technique; it relies on a forward proof-search strategy and can be applied to cut-free calculi enjoying the subformula property. Here we apply this method to derive the unprovability of a formula in the modal logic K. To this aim, we design a forward calculus to check the K-satis ability of a set of modal formulas. From a derivation of , we can extract a Kripke model of .</p>
      </abstract>
    </article-meta>
  </front>
  <body>
    <sec id="sec-1">
      <title>Introduction</title>
      <p>? This work has been partially funded by INdAM-GNCS (Italy).
an RK( )-derivation is a set containing formulas constructed using the ones in
. Axioms of RK( ) are maximal consistent sets of the set of literals determined
by . Rules of RK( ) are inspired by semantics, so that proof-search for a
derivation of mirrors the construction of a model of . A remarkable result is
that the models extracted from RK( )-derivations are in general \small". This
is mainly due to the fact that in forward derivations we can re-use the same
premise many times, without duplicating it, and this reduces the generation of
redundant worlds. Actually, for every satis able set , we can build a derivation
of in RK( ) such that the height of the extracted model is minimal.
2</p>
    </sec>
    <sec id="sec-2">
      <title>The calculus RK( )</title>
      <p>We consider the language L built from a denumerable set V of propositional
variables and the connectives ^, : and . We denote formulas by , , . . . and
set of formulas by , , . . . ; is the set f j 2 g. By LCL we denote
the set of classical formulas of L, namely the formulas not containing ; with
H, X , . . . sets of classical formulas, we call cl-sets. A literal is a formula of the
kind p or :p, with p 2 V. A set of literals X is lit-consistent i there is no p 2 V
such that fp; :pg X . The set Sf+( ) is the smallest set of formulas such that</p>
      <p>Sf+( ) and:
{ ^
{ :(
{</p>
      <p>2 Sf+( ) implies f ; g Sf+( ); :: 2 Sf+( ) implies
^ ) 2 Sf+( ) implies f: ; : g Sf+( );
2 Sf+( ) implies 2 Sf+( ); : 2 Sf+( ) implies :
2 Sf+( );
2 Sf+( ).</p>
      <p>Let W be a non-empty nite set of worlds and R be an intransitive successor
relation on W (i.e., 8x; y; z(xRy ^ yRz ! :xRz); note that R is irre exive as
well). An intransitive tree with root is a triple (W; R; ) such that the re exive
and transitive closure of R is a tree partial order on W with root [1]. A model
for L is a structure M = hW; R; ; V i, where (W; R; ) is an intransitive tree
with root and V , the evaluation function, maps each p 2 V to a subset of W
(the set of worlds where p is true). A (classical) interpretation is a subset of V;
for w 2 W , by I(w) we denote the interpretation de ned by the propositional
variables true at w (i.e., p 2 I(w) i w 2 V (p)). The relation (M; w) j= (the
formula is valid in M at world w) is de ned as usual by induction on (see
e.g. [1]):
{ (M; w) j= p, with p 2 V, i w 2 V (p);
{ (M; w) j= i , for every v 2 W , wRv implies (M; v) j= .</p>
      <p>The interpretation of : and ^ is classical. Note that the validity of a classical
formula at a world w only depends on I(w). Expressions of the kind ; and
; denote the sets [ f g and [ respectively. By (M; w) j= we mean
that, for every 2 , (M; w) j= . A set is (K-)satis able i there exists
a model M and a world w of M such that (M; w) j= . It is well-known that
the modal logic K is the set of formulas of L such that, for every model M
and every world w of M, (M; w) j= (see [1]); accordingly, is valid in K i</p>
      <p>Lit</p>
      <p>^
1
J (Tin=1 i) ; f :
f
j</p>
      <p>;
; ; ::
::</p>
      <p>; : k
; : k; :( 1 ^ 2) :^</p>
      <p>n H
j : 2 Sin=1 i g K; H</p>
      <p>H
2 Sf+( )g; H
no0
9
&gt;
&gt;
&gt;
=
&gt;
&gt;
&gt;
;
no
n</p>
      <p>1
Cl-rules
Applied only if the formula
introduced in the
conclusion belongs to Sf+( )
9
&gt;
&gt;
&gt;
=
&gt;
&gt;
&gt;
;</p>
      <p>Join rules
J K =
\ Sf+( )
X is a maximal lit-consistent subset of Sf+( )
the set f: g is not satis able. In this paper we present a forward calculus to
constructively prove the satis ability of a set of formulas.</p>
      <p>The calculus RK( ) is a forward refutation calculus to prove that a nite
nonempty set of formulas of L is satis able, meaning that the formula :(V )
is not valid in K. In forward calculi proof-search starts from axioms and rules
are applied from premises to the conclusion (forward direction). The rules are
displayed in Fig. 1. Axioms of RK( ), introduced by the axiom rule Lit, are
the cl-sets X such that X is a maximal lit-consistent subset of Sf +( ), namely:
there is no lit-consistent set Y such that X Y Sf+( ). Cl-rules are standard
refutation rules for classical connectives; they introduce a formula of the kind
generating f^orm)ualansdin::Sf +. (To):gtehtistemrmotinivaattieosn,thweesiodnelycoanddmitiitonruolen
athpeplaicpaptliiocnas^ , :(
tion of cl-rules. Join rules on and on0 allow one to introduce formulas of the kind
and : . The rule on has n + 1 2 premises, where at least one premise H
is a cl-set. Rule on0 is the degenerated instance of on only having the premise H.
In both cases, the formulas introduced by the rule applications must belong to
Sf+( ) (see the de nition of the J : K operator). An RK( )-derivation of is a
derivation in RK( ) with the set as root. A set is provable in RK( ) i
there exists an RK( )-derivation of such that ; we also say that
is proved by .</p>
      <sec id="sec-2-1">
        <title>Soundness</title>
        <p>Given an RK( )-derivation of , we can build a K-model satisfying ; this
proves the soundness RK( ). The height of , denoted h( ), is the maximum
distance from the root of and an axiom sequent of . We de ne the structure
Mod( ) inductively on the height of .
{ If h( ) = 0, then only consists of an application of Lit and is a maximal
lit-consistent subset of Sf +( ). We set Mod( ) = hf g; ;; ; V i, where ; is the
empty successor relation and V (p) = f g if p 2 and V (p) = ; otherwise.
{ If h( ) &gt; 0, let R be the rule applied at the root of .</p>
        <p>
          (
          <xref ref-type="bibr" rid="ref1">1</xref>
          ) If R is a cl-rule or R =on0, let 0 be the immediate subderivation of .
        </p>
        <p>
          Then Mod( ) = Mod( 0). Note that, if is a derivation of a cl-set H,
only consists of one instance of Lit followed by instances of cl-rules,
accordingly Mod( ) consists of a single world.
(
          <xref ref-type="bibr" rid="ref2">2</xref>
          ) Otherwise, R =on. Let i be the subderivations with root i and n+1 the
one with root H; let Mod( i) = hWi; Ri; i; Vii (1 i n + 1). Since
H is a cl-set, as discussed in Point (
          <xref ref-type="bibr" rid="ref1">1</xref>
          ), we have Wn+1 = f g. Without
loss of generality, we assume that the Wj 's are pairwise disjoint. We set
Mod( ) = hW; R; ; V i where:
        </p>
        <p>n+1
W = [ Wi;
i=1
n+1
[ Vi;
i=1
V =</p>
        <p>R =
n
[ ( Ri [ f ( ; i) g )
i=1
It is easy to check that Mod( ) is a well-de ned model. Moreover, proceeding
by induction on the height of one can prove:
Theorem 1 (Soundness of RK( )). Let
be the root of Mod( ). Then (Mod( ); ) j=
be an RK( )-derivation of
.</p>
        <p>and</p>
      </sec>
      <sec id="sec-2-2">
        <title>Completeness</title>
        <p>We prove that the calculus RK( ) is complete, namely: if is a nite satis able
set, then is provable in RK( ). We give a constructive proof by exhibiting
how to build an RK( )-derivation of starting from a model M =
hW; R; ; V i of . The height of a world w0 2 W , denoted by h(w0), is the
maximal length of an R-chain w0Rw1Rw2 : : : ; since hW; R; i is a nite tree, the
de nition is well-found. The height of M, denoted by h(M), coincides with h( ).
We show that h(Mod( )) h(M); accordingly, we can use RK( ) to generate
\small" models of . To formalize this, we introduce the following de nitions:
{
{ If
is h-satis able i there is a model M of such that h(M) h.</p>
        <p>
          is satis able, h( ) is the minimum h such that is h-satis able.
The rank Rn( ) of an RK( )-derivation is the maximum number of
applications of rule on along a branch of . Formally, let r be the root rule of and let
1; : : : ; n be the immediate subderivations of . Then:
Rn( ) =
(0 if r is the axiom rule Lit
c + maxf Rn(
          <xref ref-type="bibr" rid="ref1">1</xref>
          ); : : : ; Rn( n) g
otherw.
        </p>
        <p>c =
(1 if r =on
0 otherw.</p>
        <p>Note that h(Mod( )) = Rn( ). In the proof of next lemma we show how to
build a derivation of a satis able set. Note that Theorem 2 immediately follows
from Lemma 1. By j j we denote the number of symbols occurring in .
Lemma 1. Let be a nite set of formulas and let
satis able set. Then, there exists an RK( )-derivation
and Rn( ) h. Moreover, LCL implies LCL.
of</p>
        <p>Sf+( ) be a
hsuch that
Proof. We prove the assertion by induction hypothesis (IH) on j j. Since
is h-satis able, there exists a model M = hW; R; ; V i and w 2 W such that
(M; w) j= and h(w) h. We proceed through a case analysis, only detailing
some representative cases.</p>
        <p>Case 1: only contains literals.</p>
        <p>Let X be the set of literals l 2 Sf+( ) such that (M; w) j= l. Then, X is a
maximal lit-consistent subset of Sf +( ), hence X is an axiom of RK( ). Since</p>
        <p>X LCL, the assertion holds.</p>
        <p>Case 2: = :( 1 ^ 2); 0, with :( 1 ^ 2) 62 0.</p>
        <p>Since (M; w) j= :( 1 ^ 2), there exists k 2 f1; 2g such that (M; w) j= : k.
Let k = 0 [ f: kg; note that (M; w) j= k, hence k is h-satis able. Since
j kj &lt; j j and k Sf+( ), by (IH) there exists an RK( )-derivation k of
k such that k k and Rn( k) h. Let be:
.
.
.</p>
        <p>k
: k; 0;</p>
        <p>k
= :( 1 ^ 2); 0;
Since and Rn( ) = Rn( k)
LCL, by (IH) we get k LCL, hence</p>
        <p>h, the assertion holds. Moreover, if
Case 3: = ; : 1; : : : ; : n; H, with n 1.</p>
        <p>Let j 2 f1; : : : ; ng and j = [ f: j g. Since (M; w) j= : j , there exists
wj 2 W such that wRwj and (M; wj ) j= : j . We also have (M; wj ) j= ,
hence j is h(wj )-satis able. Since j j j &lt; j j and j Sf+( ), by (IH) there
exists an RK( )-derivation j of j such that j j and Rn( j ) h(wj ),
hence Rn( j ) &lt; h. Since H is h-satis able (indeed, (M; w) j= H) and H LCL
and jHj &lt; j j, by (IH) there exists an RK( )-derivation c of G such that
H G LCL. We can de ne as follows:
...
where we leave understood the formulas in the (possibly empty) set . We have
. Note that c cannot contain applications of rule on, hence Rn( c) = 0.
This implies that Rn( ) = m + 1, where m is the maximum Rn( j ). Since m &lt; h,
we get Rn( ) h, and this concludes the proof of this case.
tu</p>
        <p>As a consequence of the previous lemma we get:
Theorem 2 (Completeness of RK( )). Let
formulas. Then, there exists an RK( )-derivation
that h(Mod( )) h( ).
be a
of
nite set of satis able</p>
        <p>such that such
k
:^</p>
        <p>LCL.</p>
        <p>k =</p>
        <p>
          k n k
Lit
Lit
no (
          <xref ref-type="bibr" rid="ref1">1</xref>
          ) (
          <xref ref-type="bibr" rid="ref2">2</xref>
          )
no (
          <xref ref-type="bibr" rid="ref3">3</xref>
          ) (
          <xref ref-type="bibr" rid="ref1">1</xref>
          )
^(
          <xref ref-type="bibr" rid="ref4">4</xref>
          )
        </p>
        <p>
          Derivation 1
Let = p ^ : p. Note that corresponds to the negation of the transitivity
axiom (
          <xref ref-type="bibr" rid="ref4">4</xref>
          ) p ! p. It is well-known that (
          <xref ref-type="bibr" rid="ref4">4</xref>
          ) is not valid in K, hence 1 = f g
is satis able. We populate the database DB of sets provable in RK(
          <xref ref-type="bibr" rid="ref1">1</xref>
          ) according
with the naive recipe of [3]; the obtained DB is displayed in Fig 2 (for the sake
of conciseness, we only show the sets needed to get the goal 1). We start by
inserting the axioms ((
          <xref ref-type="bibr" rid="ref1">1</xref>
          ) and (
          <xref ref-type="bibr" rid="ref2">2</xref>
          )); then we enter a loop where, at each iteration,
we apply the rules to the set collected in previous steps. The iterations stop when,
either a superset of the goal is generated (as in our example) or the database
saturates (no more new sequents can be generated). The tree-like structure is
given by derivation 1. As for the model Mod(
          <xref ref-type="bibr" rid="ref1">1</xref>
          ), the world wi is generated by
the rule applied to get the sequent (i) in the database, an arrow from wi to
wj indicates that wiRwj , while the interpretation I(wi) associated with wi is
displayed after the colon (hence I(w1) = I(w4) = ; and I(w3) = fpg). It is easy
to check that (Mod(
          <xref ref-type="bibr" rid="ref1">1</xref>
          ); w4) j= . We conclude that Mod(
          <xref ref-type="bibr" rid="ref1">1</xref>
          ) is a model of f g
and a countermodel for the transitivity axiom. We remark that h(Mod(
          <xref ref-type="bibr" rid="ref1">1</xref>
          )) = 2,
which is the minimal height of a model of 1.
        </p>
        <p>
          A more signi cant example is given in Fig. 3, where we show an RK(
          <xref ref-type="bibr" rid="ref2">2</xref>
          )derivation of a satis able set 2 (read as : :) and the related model.
3
        </p>
      </sec>
    </sec>
    <sec id="sec-3">
      <title>Future work</title>
      <p>
        We have presented the naive forward proof-search strategy; we leave as future
work the investigation of clever strategies (e.g., the use of subsumption to reduce
redundancies) and the implementation of the calculus exploiting the full- edged
Java Framework JTabWb [5]. In this paper we only focus on K; we plan to
extend the techniques here introduced to other modal logics.
(A ^
(
        <xref ref-type="bibr" rid="ref1">1</xref>
        ) :p ; :q
(
        <xref ref-type="bibr" rid="ref2">2</xref>
        ) p ; :q
(
        <xref ref-type="bibr" rid="ref3">3</xref>
        ) :p ; q
(
        <xref ref-type="bibr" rid="ref4">4</xref>
        ) A ; :p ; :q
(
        <xref ref-type="bibr" rid="ref5">5</xref>
        ) B ; p ; :q
(
        <xref ref-type="bibr" rid="ref6">6</xref>
        ) C ; :p ; q
(
        <xref ref-type="bibr" rid="ref7">7</xref>
        ) D ; E ; :p ; :q
(
        <xref ref-type="bibr" rid="ref8">8</xref>
        ) :: D ; :: E ; D ; E ; :p ; :q
(9) D ; E ; A ; :p ; :q
(10) A ^ D ; A ^ E ; D ; E ; A ; :p ; :q
(11) ::(A ^ D) ; ::(A ^ E) ; : : :
(12) (A ^ D) ; (A ^ E) ; B ; p ; :q
(13) (A ^ D) ; (A ^ E) ; C ; :p ; q
(14) ::(B ^ (A ^ D)) ; B ^ (A ^
(15) ::(C ^ (A ^ E)) ; C ^ (A ^
(16) (B ^ (A ^ D)) ; (C ^ (A ^
Lit
Lit
Lit
^ (
        <xref ref-type="bibr" rid="ref1">1</xref>
        )
^ (
        <xref ref-type="bibr" rid="ref2">2</xref>
        )
^ (
        <xref ref-type="bibr" rid="ref3">3</xref>
        )
no0 (
        <xref ref-type="bibr" rid="ref1">1</xref>
        )
:: (
        <xref ref-type="bibr" rid="ref7">7</xref>
        ) (two times)
no (
        <xref ref-type="bibr" rid="ref8">8</xref>
        ) (
        <xref ref-type="bibr" rid="ref4">4</xref>
        )
^ (9) (two times)
::(10) (two times)
no (11) (
        <xref ref-type="bibr" rid="ref5">5</xref>
        )
no (11) (
        <xref ref-type="bibr" rid="ref6">6</xref>
        )
D) ; : : : ^ (12) followed by ::
E) ; : : : ^ (13) followed by ::
E)) ; A ; :p ; :q no (14)(15)(
        <xref ref-type="bibr" rid="ref4">4</xref>
        )
Sf+(
        <xref ref-type="bibr" rid="ref2">2</xref>
        ) =
8 q; p; :p; :q; A; B; C; D; E; D; E; ::
&gt;&gt;&gt;&gt; D; (A ^ D); ::(A ^ D); (A ^
&gt;&lt;&gt;&gt; E; (A ^ E); ::(A ^ E); (A ^
      </p>
      <p>(B ^ (A ^ D)); ::(B ^ (A ^
&gt;&gt;&gt;&gt; B ^ (A ^ D); (C ^ (A ^ E));
&gt;
&gt;&gt;: ::(C ^ (A ^ E)); C ^ (A ^ E);</p>
      <p>D; :: E 9</p>
      <p>D); &gt;&gt;&gt;&gt;</p>
      <p>E); &gt;=&gt;&gt;
D));
&gt;
&gt;
&gt;
&gt;
&gt;
&gt;
&gt;
;
w1:
w9:
w10:
w90:
w16:</p>
    </sec>
  </body>
  <back>
    <ref-list>
      <ref id="ref1">
        <mixed-citation>
          1.
          <string-name>
            <given-names>A.</given-names>
            <surname>Chagrov</surname>
          </string-name>
          and
          <string-name>
            <given-names>M.</given-names>
            <surname>Zakharyaschev</surname>
          </string-name>
          .
          <source>Modal Logic</source>
          . Oxford Univ. Press,
          <year>1997</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref2">
        <mixed-citation>
          2.
          <string-name>
            <given-names>K.</given-names>
            <surname>Chaudhuri</surname>
          </string-name>
          ,
          <string-name>
            <given-names>F.</given-names>
            <surname>Pfenning</surname>
          </string-name>
          , and
          <string-name>
            <given-names>G.</given-names>
            <surname>Price</surname>
          </string-name>
          .
          <article-title>A logical characterization of forward and backward chaining in the inverse method</article-title>
          . In U. Furbach et al., editor,
          <source>IJCAR 2006</source>
          , volume
          <volume>4130</volume>
          <source>of LNCS</source>
          , pages
          <volume>97</volume>
          {
          <fpage>111</fpage>
          . Springer,
          <year>2006</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref3">
        <mixed-citation>
          3.
          <string-name>
            <given-names>A.</given-names>
            <surname>Degtyarev</surname>
          </string-name>
          and
          <string-name>
            <given-names>A.</given-names>
            <surname>Voronkov</surname>
          </string-name>
          .
          <article-title>The inverse method</article-title>
          .
          <source>In J.A. Robinson</source>
          et al., editor,
          <source>Handbook of Automated Reasoning</source>
          , pages
          <volume>179</volume>
          {
          <fpage>272</fpage>
          . Elsevier and MIT Press,
          <year>2001</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref4">
        <mixed-citation>
          4.
          <string-name>
            <given-names>K.</given-names>
            <surname>Donnelly</surname>
          </string-name>
          ,
          <string-name>
            <given-names>T.</given-names>
            <surname>Gibson</surname>
          </string-name>
          ,
          <string-name>
            <given-names>N.</given-names>
            <surname>Krishnaswami</surname>
          </string-name>
          ,
          <string-name>
            <given-names>S.</given-names>
            <surname>Magill</surname>
          </string-name>
          , and
          <string-name>
            <given-names>S.</given-names>
            <surname>Park</surname>
          </string-name>
          .
          <article-title>The inverse method for the logic of bunched implications</article-title>
          . In F. Baader et al., editor,
          <source>LPAR 2004</source>
          , volume
          <volume>3452</volume>
          <source>of LNCS</source>
          , pages
          <volume>466</volume>
          {
          <fpage>480</fpage>
          . Springer,
          <year>2004</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref5">
        <mixed-citation>
          5.
          <string-name>
            <given-names>M.</given-names>
            <surname>Ferrari</surname>
          </string-name>
          ,
          <string-name>
            <given-names>C.</given-names>
            <surname>Fiorentini</surname>
          </string-name>
          , and
          <string-name>
            <surname>G. Fiorino.</surname>
          </string-name>
          <article-title>JTabWb: a Java framework for implementing terminating sequent and tableau calculi</article-title>
          .
          <source>Fundamenta Informaticae</source>
          ,
          <volume>150</volume>
          :
          <fpage>119</fpage>
          {
          <fpage>142</fpage>
          ,
          <year>2017</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref6">
        <mixed-citation>
          6.
          <string-name>
            <given-names>Camillo</given-names>
            <surname>Fiorentini</surname>
          </string-name>
          and
          <string-name>
            <given-names>Mauro</given-names>
            <surname>Ferrari</surname>
          </string-name>
          .
          <article-title>A forward unprovability calculus for intuitionistic propositional logic</article-title>
          . In R. A.
          <string-name>
            <surname>Schmidt</surname>
          </string-name>
          et al., editor,
          <source>TABLEAUX 2017</source>
          , volume
          <volume>10501</volume>
          <source>of LNCS</source>
          , pages
          <volume>114</volume>
          {
          <fpage>130</fpage>
          . Springer,
          <year>2017</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref7">
        <mixed-citation>
          7.
          <string-name>
            <given-names>L.</given-names>
            <surname>Kovacs</surname>
          </string-name>
          ,
          <string-name>
            <given-names>A.</given-names>
            <surname>Mantsivoda</surname>
          </string-name>
          ,
          <article-title>and</article-title>
          <string-name>
            <given-names>A.</given-names>
            <surname>Voronkov</surname>
          </string-name>
          .
          <article-title>The inverse method for many-valued logics</article-title>
          . In F. Castro-Espinoza et al., editor,
          <source>MICAI 2013</source>
          , volume
          <volume>8265</volume>
          <source>of LNCS</source>
          , pages
          <volume>12</volume>
          {
          <fpage>23</fpage>
          . Springer,
          <year>2013</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref8">
        <mixed-citation>
          8.
          <string-name>
            <surname>S. Ju. Maslov.</surname>
          </string-name>
          <article-title>An invertible sequential version of the constructive predicate calculus</article-title>
          .
          <source>Zap. Naucn. Sem. Leningrad. Otdel. Mat. Inst. Steklov. (LOMI)</source>
          ,
          <volume>4</volume>
          :
          <fpage>96</fpage>
          {
          <fpage>111</fpage>
          ,
          <year>1967</year>
          .
        </mixed-citation>
      </ref>
    </ref-list>
  </back>
</article>