<!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>Knowledge Graphs and Creative Applications ?</article-title>
      </title-group>
      <contrib-group>
        <contrib contrib-type="author">
          <string-name>Jer Hayes</string-name>
          <email>jeremiah.hayes@accenture.com</email>
          <xref ref-type="aff" rid="aff0">0</xref>
        </contrib>
        <contrib contrib-type="author">
          <string-name>Nicholas McCarthy</string-name>
          <xref ref-type="aff" rid="aff0">0</xref>
        </contrib>
        <contrib contrib-type="author">
          <string-name>Xu Zheng</string-name>
          <xref ref-type="aff" rid="aff0">0</xref>
        </contrib>
        <aff id="aff0">
          <label>0</label>
          <institution>Accenture Labs The Dock, 7 Hanover Quay, Dublin https://labs.accenture.com/</institution>
        </aff>
      </contrib-group>
      <abstract>
        <p>We applied computational creativity to the food domain to create a system that uses a domain speci c knowledge graph and graph embeddings to aid a user in nding novel and interesting avour combinations. Metrics like novelty and surprise are calculated using graph embeddings to retrieve the most useful combinations. Computational creativity is the application of AI technologies to emulate and stimulate human creativity. Fast Moving Consumer Goods food (FMCG) companies often nd it di cult to innovate fast enough to satisfy the market in a timely manner, e.g., a market trend showing people would like a healthier snack bar might drive an FMCG to ask what ingredients could be pair with chocolate that also had a health bene t? We created a domain speci c knowledge graph and graph embeddings to aid a user in nding novel and interesting avour combinations with the main goal of 'feeding' the mind of creative professionals in the food business with new unexpected combinations.</p>
      </abstract>
      <kwd-group>
        <kwd>Computational Creativity</kwd>
        <kwd>Knowledge Graphs</kwd>
        <kwd>Graph Em- beddings</kwd>
      </kwd-group>
    </article-meta>
  </front>
  <body>
    <sec id="sec-1">
      <title>Background</title>
    </sec>
    <sec id="sec-2">
      <title>Our approach</title>
      <p>
        A knowledge graph of triples in the format &lt;subject, predicate, object&gt; that
covered ingredients, recipes, nutrients, compounds, avour molecules was compiled
from multiple public data sources. For instance, we extracted triples like &lt;recipe
1, contains, ingredient 1&gt;, &lt;recipe 1, comes from, location 1&gt;, etc. Ampligraph
[
        <xref ref-type="bibr" rid="ref2">2</xref>
        ] was used with the ComplEx scoring function [
        <xref ref-type="bibr" rid="ref3">3</xref>
        ] to input triples into a number
representing a probability that a given triple is correct. The resulting
embeddings are used as the input data for several metrics. This allows us to compute
a probability distribution dI1;:::;In for a set of ingredients fI1; :::; Ing in a space
of known recipes. Then the likelihood p(I1; :::; In) is computed as a mean of
? Copyright c 2020 for this paper by its authors. Use permitted under Creative
Commons License Attribution 4.0 International (CC BY 4.0).
that distribution and the novelty is taken to be 1 p(I1; :::; In). Finally, the one
measure of surprise of fI1; :::; Ing is the Kullback{Leibler divergence between its
distribution dI1;:::;In and likelihood p(I1; :::; In).
      </p>
      <p>
        Other metrics in the system are expectedness - what ingredient is likely
to come next given the recipe data in the knowledge graph, and pleasantness
which is based on the chemical structure information of the avour compounds.
However, this system ultimately relies on the human user to make the nal
selection and as such is an ideation tool that assists the human expert[
        <xref ref-type="bibr" rid="ref1">1</xref>
        ].
      </p>
    </sec>
    <sec id="sec-3">
      <title>Conclusion</title>
      <p>We built a system that shows how knowledge graphs and graph embeddings
can be used in the new food product ideation process. The metrics for novelty
and surprise are calculated from the graph embeddings and help give product
developers fast, curated inspiration and companies a quicker time to market for
products targeting momentary markets. Future work includes a focus on the
formulations of FMCG products and the creation of new formulations.</p>
    </sec>
  </body>
  <back>
    <ref-list>
      <ref id="ref1">
        <mixed-citation>
          1.
          <article-title>Enhancing culinary creativity with ai</article-title>
          , https://www.youtube.com/watch?v= NjT2OjCdGp4
        </mixed-citation>
      </ref>
      <ref id="ref2">
        <mixed-citation>
          2.
          <string-name>
            <surname>Costabello</surname>
            ,
            <given-names>L.</given-names>
          </string-name>
          ,
          <string-name>
            <surname>Pai</surname>
            ,
            <given-names>S.</given-names>
          </string-name>
          ,
          <string-name>
            <surname>Van</surname>
            ,
            <given-names>C.L.</given-names>
          </string-name>
          ,
          <string-name>
            <surname>McGrath</surname>
            ,
            <given-names>R.</given-names>
          </string-name>
          ,
          <string-name>
            <surname>McCarthy</surname>
            ,
            <given-names>N.</given-names>
          </string-name>
          ,
          <string-name>
            <surname>Tabacof</surname>
            ,
            <given-names>P.:</given-names>
          </string-name>
          <article-title>AmpliGraph: a Library for Representation Learning on Knowledge Graphs (Mar</article-title>
          <year>2019</year>
          ). https://doi.org/10.5281/zenodo.2595043, https://doi.org/10.5281/zenodo.2595043
        </mixed-citation>
      </ref>
      <ref id="ref3">
        <mixed-citation>
          3.
          <string-name>
            <surname>Trouillon</surname>
            ,
            <given-names>T.</given-names>
          </string-name>
          ,
          <string-name>
            <surname>Welbl</surname>
            ,
            <given-names>J.</given-names>
          </string-name>
          ,
          <string-name>
            <surname>Riedel</surname>
            ,
            <given-names>S.</given-names>
          </string-name>
          ,
          <string-name>
            <surname>Gaussier</surname>
            ,
            <given-names>E.</given-names>
          </string-name>
          ,
          <string-name>
            <surname>Bouchard</surname>
          </string-name>
          , G.:
          <article-title>Complex embeddings for simple link prediction (</article-title>
          <year>2016</year>
          )
        </mixed-citation>
      </ref>
    </ref-list>
  </back>
</article>