<!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>Towards a sequent calculus for formal contexts</article-title>
      </title-group>
      <contrib-group>
        <contrib contrib-type="author">
          <string-name>Ondrej Kr dlo</string-name>
          <xref ref-type="aff" rid="aff1">1</xref>
        </contrib>
        <contrib contrib-type="author">
          <string-name>Manuel Ojeda-Aciego</string-name>
          <xref ref-type="aff" rid="aff0">0</xref>
        </contrib>
        <aff id="aff0">
          <label>0</label>
          <institution>Universidad de Malaga. Departamento de Matematica Aplicada.</institution>
          <country country="ES">Spain</country>
        </aff>
        <aff id="aff1">
          <label>1</label>
          <institution>University of Pavol Jozef Safarik</institution>
          ,
          <addr-line>Kosice</addr-line>
          ,
          <country country="SK">Slovakia</country>
        </aff>
      </contrib-group>
      <abstract>
        <p>This work focuses on the de nition of a consequence relation between contexts with which we can decide whether certain contextual information is a logical consequence from a set of contexts considered as underlying hypotheses.</p>
      </abstract>
    </article-meta>
  </front>
  <body>
    <sec id="sec-1">
      <title>Introduction</title>
      <p>In real life, one often faces situations in which the underlying knowledge is
given as a set of tables which can be interpreted as formal contexts, and we
should decide on whether certain contextual information is a consequence
from them.</p>
      <p>This problem clearly resembles the notion of a formula being a logical
consequence of a set of hypotheses, and suggests the possibility or
convenience of de ning a formal (mathematical) notion of logical consequence
between contexts.</p>
      <p>
        The only attempts to introduce logical content within the machinery
of Formal Concept Analysis (FCA), apart from its ancient roots anchored
in the Port-Royal logic, are the so-called logical information systems and
the logical concept analysis [
        <xref ref-type="bibr" rid="ref2 ref9">2, 9</xref>
        ].
      </p>
      <p>
        Of course, di erent links between FCA and logic have been studied
but, to the best of our knowledge, the problem considered in this paper has
not been explicitly studied in the literature. Nevertheless, it is worth to
remark that the concluding section of [
        <xref ref-type="bibr" rid="ref6">6</xref>
        ] states that dual bonds could be
given a proof-theoretical interpretation in terms of consequence relations.
      </p>
      <p>In the present work, we consider the fact that the category ChuCors of
contexts and Chu correspondences is *-autonomous, and hence a model of
linear logic, in order to build some preliminary links with the conjunctive
fragment of this logic.
? Partially supported by grants VEGA 1/0073/15 and APVV-15-0091
?? Partially supported by Spanish Ministry of Science project TIN2015-70266-C2-1-P,
co-funded by the European Regional Development Fund (ERDF)</p>
    </sec>
    <sec id="sec-2">
      <title>Preliminaries</title>
      <p>In order to make the manuscript self-contained, the fundamental notions
and their main properties are recalled in this section.
2.1</p>
      <sec id="sec-2-1">
        <title>Context, concept and concept lattice</title>
        <p>De nition 1. A formal context is any triple C = hB; A; Ri where B and
A are nite sets and R B A is a binary relation. It is customary to
say that B is a set of objects, A is a set of attributes and R represents
a relation between objects and attributes.</p>
        <p>Given a formal context hB; A; Ri, the derivation (or concept-forming)
operators are a pair of mappings " : 2B ! 2A and # : 2A ! 2B such that
if X B, then "X is the set of all attributes which are related to every
object in X and, similarly, if Y A, then # Y is the set of all objects
which are related to every attribute in Y .</p>
        <p>De nition 2. A formal concept of a formal context C = hB; A; Ri is a
pair of sets hX; Y i 2 2B 2A which is a xpoint of the pair of
conceptforming operators, namely, " X = Y and # Y = X. The object part X
is called the extent and the attribute part Y is called the intent. The set
of all formal concepts of a context C will be denoted by CL(C), set of all
extents or intents of C will be denoted by Ext(C) or Int(C) respectively.
2.2</p>
      </sec>
      <sec id="sec-2-2">
        <title>Intercontextual structures</title>
        <p>Two main constructions have been traditionally considered in order to
relate two formal contexts: the bonds and the Chu correspondences.
De nition 3. Let C1 = hB1; A1; R1i and C2 = hB2; A2; R2i be two formal
contexts. A bond between C1 and C2 is any relation 2 2B1 A2 such
that, when interpreted as a table, its columns are extents of C1 and its
rows are intents of C2. All bonds between such contexts will be denoted by
Bonds(C1; C2).</p>
        <p>Another equivalent de nition of bond between C1 and C2 de nes it
as any relation 2 2B1 A2 such that Ext(hB1; A2; i) Ext(C1) and
Int(hB1; A2; i) Int(C2)</p>
        <p>Dual bonds between C1 and C2 are bonds between C1 and transposition
of C2. Transposition of any context C = hB; A; Ri is de ned as a new
context C = hA; B; Rti with Rt(a; b) holds i R(b; a) holds.</p>
        <p>The notion of Chu correspondence between contexts can be seen as
an alternative inter-contextual structure which, instead, links intents of
C1 and extents of C2.</p>
        <p>De nition 4. Consider C1 = hB1; A1; R1i and C2 = hB2; A2; R2i two
formal contexts. A Chu correspondence between C1 and C2 is any pair
' = h'L; 'Ri of mappings 'L : B1 ! Ext(C2) and 'R : A2 ! Int(C1)
such that for all (b1; a2) 2 B1 A2 it holds that "2 'L(b1) (a2) = #1
'R(a2) (b1).</p>
        <p>All Chu correspondences between such contexts will be denoted by
Chu(C1; C2).</p>
        <p>The notions of bond and Chu correspondence are interchangeable;
speci cally, we can consider the bond ' associated to a Chu
correspondence ' from C1 to C2 de ned for b1 2 B1; a2 2 A2 as follows
'(b1; a2) = "2 'L(b1) (a2) = #1 'R(a2) (b1)
Similarly, we can consider the Chu correspondence ' associated to a
bond de ned by the following pair of mappings:
' L(b1) = #2 (b1) ' R(a2) = "1
t(a2) for all a2 2 A2 and o1 2 B1</p>
        <p>The set of all bonds (resp. Chu correspondences) between two formal
contexts endowed with the ordering given by set inclusion is a complete
lattice. Moreover, both complete lattices are dually isomorphic.
2.3</p>
      </sec>
      <sec id="sec-2-3">
        <title>Categorical products in ChuCors</title>
        <p>
          Recall that it is possible to consider a category in which the objects are
formal contexts and morphisms between two contexts are the Chu
correspondences between them. This category, denoted ChuCors, has been
proved to be *-autonomous and equivalent to the category of complete
lattices and isotone Galois connections, more results on this category and
its L-fuzzy extensions can be found in [
          <xref ref-type="bibr" rid="ref3 ref4 ref5 ref7">4, 3, 5, 7</xref>
          ].
        </p>
        <p>Cartesian product in ChuCors The following de nition provides a
speci c construction of the notion of (binary) cartesian product in the
category ChuCors.</p>
        <p>De nition 5. Consider C1 = hB1; A1; R1i and C2 = hB2; A2; R2i two
formal contexts. The product of such contexts is a new formal context
C1 C2 = hB1 ] B2; A1 ] A2; R1 2i where the relation R1 2 is given by
((i; b); (j; a)) 2 R1 2 if and only if (i = j) ) (b; a) 2 Ri
for any (b; a) 2 Bi</p>
        <p>Aj and (i; j) 2 f1; 2g
f1; 2g.</p>
        <p>If we recall the well-known categorical theorem which states that if
a category has a terminal object and binary product, then it has all
nite products, we need to prove just the existence of a terminal object
(namely, the nullary product) in order to prove the category ChuCors to
be Cartesian.</p>
        <p>Any formal context of the form hB; A; B Ai where the incidence
relation is the full Cartesian product of the sets of objects and attributes
is (isomorphic to) the terminal object of ChuCors. Such a formal context
has just one formal concept hB; Ai; hence, from any other formal context
there is just one Chu correspondence to hB; A; B Ai.</p>
        <p>The explicit construction of a general product (not necessarily either
binary or nullary) is given below:
De nition 6. Let fCigi2I be an indexed family of formal contexts Ci =
hBi; Ai; Rii, the product Qi2I Ci is the formal context given by
*
Y Ci =
i2I
] Bi; ] Ai; R I
i2I i2I
+
where (k; b); (m; a) 2 R I , (k = m) ) (b; a) 2 Rk .</p>
        <p>It is worth to note that the arbitrary product of contexts commutes
with both the concept lattice construction and the bonds between
contexts. These two results are explicitly stated below.</p>
        <p>Lemma 1. Let Ci = hBi; Ai; Rii be a formal context for i 2 I. It holds
that CL(Qi2I Ci) is isomorphic to Qi2I CL(Ci).</p>
        <p>Lemma 2. Let I and J be two index sets, and consider the two sets of
formal contexts fCigi2I and fDj gj2J . The following isomorphism holds
Bonds Y Ci; Y</p>
        <p>Dj
i2I
j2J
=</p>
        <p>Y
(i;j)2I J</p>
        <sec id="sec-2-3-1">
          <title>Bonds(Ci; Dj ) :</title>
          <p>Tensor product Another product-like construction can be given in the
category ChuCors.</p>
          <p>Note that if ' 2 Chu(C1; C2), then we can consider ' 2 Chu(C2 ; C1 )
de ned by 'L = 'R and 'R = 'L.</p>
          <p>De nition 7. The tensor product C1 C2 of contexts Ci = hBi; Ai; Rii
for i 2 f1; 2g is de ned as the context hB1 B2; Chu(C1; C2 ); R i where
R</p>
          <p>(b1; b2); ' = #2 'L(b1) (b2):</p>
          <p>
            The properties of the tensor product were shown in [
            <xref ref-type="bibr" rid="ref7">7</xref>
            ], together with
the result that ChuCors with is symmetric and monoidal. Those results
were later extended to the L-fuzzy case in [
            <xref ref-type="bibr" rid="ref3">3</xref>
            ]. In both papers, the
structure of the formal concepts of a tensor product context was established
as an ordered pair formed by a bond and a set of Chu correspondences.
Lemma 3. Let ( ; X) be a formal concept of the tensor product C1
it holds that = V 2X and X = f 2 Chu(C1; C2 ) j g.
C2,
          </p>
          <p>Due to the monoidal properties of on ChuCors we can add a notion
of n-ary tensor product of n formal contexts in=1Ci of any n formal
contexts Ci for i 2 f1; : : : ; ng. Hence, it is possible to consider a notion of
n-ary bond that we can imagine as any extent of n-ary tensor product.
De nition 8. Let Ci = hBi; Ai; Rii be formal contexts for i 2 f1; : : : ; ng.
A dual n-ary bond between fCigin=1 is an n-ary relation Qin=1 Bi
such that for all i 2 f1; 2; : : : ; ng and any (b1; : : : ; bi 1; bi+1; : : : ; bn) 2
Qn
j=1;j6=i Bi it holds that</p>
          <p>(b1; : : : ; bi 1; ( ); bi+1; : : : ; bn) 2 Ext(Ci) :
Lemma 4. Let fC1; : : : ; Cng be a set of n formal contexts and be some
n-ary bond between such contexts. Let Di be a new formal context
dened as hBi; Qn</p>
          <p>j=1;j6=i Bj ; Rii where Ri(bi; (b1; : : : ; bi 1; bi+1; : : : ; bn)) =
(b1; : : : ; bn) for any i 2 f1; 2; : : : ; ng. Then Ext(Di )
Ext(Ci).
3</p>
        </sec>
      </sec>
    </sec>
    <sec id="sec-3">
      <title>Conjunctive linear logic in FCA</title>
      <p>One of the main di erences between linear and classical logic is the
coexistence of two di erent conjunctions in linear logic, in both cases the
underlying semantics is that two actions are possible, or can be executed,
but the di erence relies on how these actions are actually performed: on
the one hand, we have the multiplicative conjunction (times) which
expresses that both actions will be performed; on the other hand, the
additive conjunction &amp; (with) states that, although both actions are
possible, actually just one will be performed.
3.1</p>
      <sec id="sec-3-1">
        <title>Additive conjunction</title>
        <p>The categorical product on ChuCors plays the role of additive
conjunction &amp;. Recall that the product C1 C2 is de ned as hB1]B2; A1]A2; R1 2i
where R1 2(b; a) = (i = j) ) (b; a) 2 Ri for all b 2 Bi and a 2 Aj for
all i; j 2 f1; 2g.</p>
        <p>The semantics of the additive conjunction is that, in order to perform
the action of C1 C2, we have rst to choose which among the two
possible actions we want to perform, and then to do the one selected. And
this is exactly what happens here. From the previous section about the
categorical product of ChuCors, it is known that a concept lattice of a
product of formal contexts is equal to a product of concept lattices of the
input formal contexts. Hence no interaction or parallel action of input
formal contexts occurs in case of the categorical product.
3.2</p>
      </sec>
      <sec id="sec-3-2">
        <title>Multiplicative conjunction and dual bonds</title>
        <p>Any dual bond between two formal contexts C1 and C2 plays a role of
multiplicative conjunction . From the de nition of bonds one can see
the parallelism of use of input contexts, where every value in the bond,
as a relation, is from both extents of input contexts.</p>
        <p>The existing isomorphism between Chu correspondences and bonds,
and monoidal properties of Chu correspondences which follows from the
fact that Chu correspondences form a category, bonds satis es all
properties of conjunction.</p>
        <p>In the sense of category theory, any dual bond can be seen as a Galois
connection between concept lattices of their input contexts, because of
the categorical equivalence between category ChuCors and the category
of complete lattices and isotone Galois connections.</p>
        <p>
          In [
          <xref ref-type="bibr" rid="ref1 ref8">1, 8</xref>
          ] one can nd that the tensor product is used as multiplicative
conjunction, due to its monoidal properties on category of Chu Spaces or
on any monoidal category in general. Here, in FCA, the tensor product
is a special formal context with nice properties that generates all bonds
between the input formal contexts. Hence tensor product shows all
possibilities how we can connect or use in parallel two formal contexts.
4
        </p>
      </sec>
    </sec>
    <sec id="sec-4">
      <title>Sequent calculus</title>
      <p>The idea here is to develop a logic between contexts and, speci cally, a
proof theory for this logic.</p>
      <p>We will consider the problem of de ning a consequence relation j=
which allows for developing formally a sequent calculus between contexts.
De nition 9. Two contexts C and D are isomorphic if their concept
lattices are isomorphic.</p>
      <p>De nition 10. Given formal contexts C1; : : : ; Cn; C, we say that C is a
consequence of contexts C1; : : : ; Cn, denoted as C1; : : : ; Cn j= C, if C is
isomorphic to a bond between all contexts in the left hand side. Speci cally,
C1; : : : ; Cn j= C if and only if C is isomorphic to some n-ary bond between
input formal contexts C1; : : : ; Cn.</p>
      <p>The relation just de ned satis es the properties of closure operator
and, hence, can be considered as a relation of logical consequence between
contexts, since any consequence relation can be viewed as a nitary
closure operator on a set (of sentences or formulas).</p>
      <p>Lemma 5. The relation j= above is a consequence relation.</p>
      <p>Recall the notion of sequent of Gentzen calculus. Any sequent is of
the form ` and has the following meaning: from the conjunction of
all hypothesis of follows some formula of . Hence as a conjunction of
all contexts we use some n-ary bond between input contexts.</p>
      <p>Without entering into the details, the following sequent rules from
conjunctive linear logic can be proved in terms of this de nition.
Axiom rule: As a unary bond we use a context itself
Constants rule Due to isomorphism Bonds (C; &gt;) = Ext(C) where &gt; =
hf g; f g; 6=i we can write the following rule</p>
      <p>C j= C</p>
      <p>C1; : : : ; Cn j= C
&gt;; C1; : : : ; Cn j= C
-left From the de nition of j= and associativity of tensor product, or
of associativity inside of n-ary product, we can write
-left Due to the distributivity of tensor and categorical product on
formal contexts, n-ary bond between C1; : : : ; Cn 1; Cn</p>
      <sec id="sec-4-1">
        <title>D is equal to</title>
        <p>product of n-ary bonds between C1; : : : ; Cn 1; Cn and C1; : : : ; Cn 1; D.
Hence it is easy to use a full relation as the n-ary bond between</p>
      </sec>
      <sec id="sec-4-2">
        <title>C1; : : : ; Cn 1; D to obtain the following rule.</title>
        <p>-right One of the possibilities here is to add to hypothesis a special
context, product of two singletons &gt; &gt;.
We have obtained a preliminary notion of logical consequence relation
between contexts which, together with the interpretation of the
multiplicative (resp. additive) conjunction as the cartesian product (resp. bond) of
contexts, enable to prove the correctness of the corresponding rules of the
sequent calculus of the conjunctive fragment of linear logic.</p>
        <p>Of course, we are just scratching the surface of the problem of
providing a full calculus since we still need to nd the adequate context-related
constructions to interpret the rest of connectives. This is future work.</p>
      </sec>
    </sec>
  </body>
  <back>
    <ref-list>
      <ref id="ref1">
        <mixed-citation>
          1.
          <string-name>
            <given-names>S.</given-names>
            <surname>Abramsky</surname>
          </string-name>
          and
          <string-name>
            <given-names>N.</given-names>
            <surname>Tzevelekos</surname>
          </string-name>
          .
          <article-title>Introduction to categories and categorical logic</article-title>
          .
          <source>Lecture Notes in Physics</source>
          ,
          <volume>813</volume>
          :3{
          <fpage>94</fpage>
          ,
          <year>2011</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref2">
        <mixed-citation>
          2.
          <string-name>
            <given-names>P.</given-names>
            <surname>Cellier</surname>
          </string-name>
          ,
          <string-name>
            <given-names>S.</given-names>
            <surname>Ferre</surname>
          </string-name>
          ,
          <string-name>
            <given-names>M.</given-names>
            <surname>Ducasse</surname>
          </string-name>
          , and
          <string-name>
            <given-names>T.</given-names>
            <surname>Charnois</surname>
          </string-name>
          .
          <article-title>Partial orders and logical concept analysis to explore patterns extracted by data mining</article-title>
          .
          <source>Lecture Notes in Arti cial Intelligence</source>
          ,
          <volume>6828</volume>
          :
          <fpage>77</fpage>
          {
          <fpage>90</fpage>
          ,
          <year>2011</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref3">
        <mixed-citation>
          3.
          <string-name>
            <surname>O.</surname>
          </string-name>
          <article-title>Kr dlo</article-title>
          , S. Krajci, and
          <string-name>
            <given-names>M.</given-names>
            <surname>Ojeda-Aciego</surname>
          </string-name>
          .
          <article-title>The category of L-Chu correspondences and the structure of L-bonds</article-title>
          .
          <source>Fundamenta Informaticae</source>
          ,
          <volume>115</volume>
          (
          <issue>4</issue>
          ):
          <volume>297</volume>
          {
          <fpage>325</fpage>
          ,
          <year>2012</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref4">
        <mixed-citation>
          4.
          <string-name>
            <given-names>O.</given-names>
            <surname>Kr</surname>
          </string-name>
          dlo and M.
          <string-name>
            <surname>Ojeda-Aciego</surname>
          </string-name>
          .
          <article-title>On L-fuzzy Chu correspondences</article-title>
          .
          <source>Intl J of Computer Mathematics</source>
          ,
          <volume>88</volume>
          (
          <issue>9</issue>
          ):
          <year>1808</year>
          {
          <year>1818</year>
          ,
          <year>2011</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref5">
        <mixed-citation>
          5.
          <string-name>
            <surname>O.</surname>
          </string-name>
          <article-title>Kr dlo and M. Ojeda-Aciego. Revising the link between L-Chu correspondences and completely lattice L-ordered sets</article-title>
          .
          <source>Annals of Mathematics and Arti cial Intelligence</source>
          ,
          <volume>72</volume>
          :
          <fpage>91</fpage>
          {
          <fpage>113</fpage>
          ,
          <year>2014</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref6">
        <mixed-citation>
          6. M. Krotzsch, P. Hitzler, and
          <string-name>
            <given-names>G.-Q.</given-names>
            <surname>Zhang</surname>
          </string-name>
          . Morphisms in context.
          <source>Lecture Notes in Computer Science</source>
          ,
          <volume>3596</volume>
          :
          <fpage>223</fpage>
          {
          <fpage>237</fpage>
          ,
          <year>2005</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref7">
        <mixed-citation>
          7.
          <string-name>
            <given-names>H.</given-names>
            <surname>Mori</surname>
          </string-name>
          .
          <article-title>Chu correspondences</article-title>
          .
          <source>Hokkaido Mathematical Journal</source>
          ,
          <volume>37</volume>
          :
          <fpage>147</fpage>
          {
          <fpage>214</fpage>
          ,
          <year>2008</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref8">
        <mixed-citation>
          8.
          <string-name>
            <given-names>V.</given-names>
            <surname>Pratt</surname>
          </string-name>
          .
          <article-title>Chu spaces as a semantic bridge between linear logic and mathematics</article-title>
          .
          <source>Theoretical Computer Science</source>
          ,
          <volume>294</volume>
          :
          <fpage>439</fpage>
          {
          <fpage>471</fpage>
          ,
          <year>2003</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref9">
        <mixed-citation>
          9.
          <string-name>
            <given-names>O.</given-names>
            <surname>Ridoux</surname>
          </string-name>
          and
          <string-name>
            <given-names>S.</given-names>
            <surname>Ferre</surname>
          </string-name>
          .
          <article-title>Introduction to logical information systems</article-title>
          .
          <source>Information Processing and Management</source>
          ,
          <volume>40</volume>
          :
          <fpage>383</fpage>
          {
          <fpage>419</fpage>
          ,
          <year>2004</year>
          .
        </mixed-citation>
      </ref>
    </ref-list>
  </back>
</article>