<html>
  <head>

    <meta http-equiv="content-type" content="text/html; charset=utf-8">
  </head>
  <body bgcolor="#FFFFFF" text="#000000">
    <small>
      <style><head>
  <style><head>
  <style><head>
  <style><head>
  <style><head>
  <style><head>
  <style><head>
  <style><head>
  <style><head>
  <style><head>
  <style><head>
  <style><head>
  <style><head>
  <style><head>
  <style><head>
  <style><head>
  <style><head>
  <style></style> </small>
    <table class="backgroundTable" cellpadding="0" cellspacing="0"
      bgcolor="#ffffff" width="100%">
      <tbody>
        <tr>
          <td align="left" valign="top">
            <table cellpadding="0" cellspacing="0">
              <tbody>
                <tr>
                  <td style="border-top: 0px solid rgb(0, 0, 0);
                    border-bottom: 1px solid rgb(255, 255, 255);
                    background-color: rgb(255, 255, 255); text-align:
                    center;" align="center"><span style="font-size:
                      10px; color: rgb(96, 96, 96); line-height: 200%;
                      font-family: verdana; text-decoration: none;">Email
not
                      displaying correctly? <a
                        href="http://www.oplin.org/4cast/"
                        style="font-size: 10px; color: rgb(0, 0, 255);
                        line-height: 200%; font-family: verdana;
                        text-decoration: none;">View
                        it in your browser.</a></span></td>
                </tr>
                <tr>
                  <td style="border-top: 0px solid rgb(51, 51, 51);
                    border-bottom: 0px solid rgb(255, 255, 255);
                    background-color: rgb(255, 255, 255);">
                    <center><a href=""><img id="editableImg1"
                          src="cid:part2.00090106.00000008@oplin.org"
                          title="OPLIN" alt="OPLIN 4Cast" border="0"
                          align="middle"></a></center>
                  </td>
                </tr>
              </tbody>
            </table>
            <table style="width: 763px; height: 877px;" cellpadding="20"
              cellspacing="0" bgcolor="#ffffff">
              <tbody>
                <tr>
                  <td style="font-size: 12px; color: rgb(0, 0, 0);
                    line-height: 150%; font-family: trebuchet ms;"
background="http://www.oplin.org/4cast/wp-content/themes/4cast/images/kubrickbgwide.jpg"
                    bgcolor="#ffffff" valign="top">
                    <p><!-- Make sure you modify the 4Cast title in this section -->
                      <span style="font-size: 20px; font-weight: bold;
                        color: rgb(0, 0, 0); font-family: arial;
                        line-height: 110%;">OPLIN 4cast #414: Image
                        search research</span><br>
                      <!-- Make sure you modify the date of the 4Cast in this section -->
                      <span style="font-size: 11px; font-weight: normal;
                        color: rgb(102, 102, 102); font-style: italic;
                        font-family: arial;">December 3rd, 2014</span></p>
                    <!-- Begin copy of Web Source here  -->
                    <p style="text-align: justify;font-size: 16px;
                      font-family: arial; line-height: 110%;"><img
                        src="cid:part4.05050509.01060005@oplin.org"
                        alt="search icon" align="left" height="105"
                        width="100">How many times have you had a
                      library patron say, "I once read a really good
                      book, it had a red cover with a bicycle on the
                      front [or some other cover description] - can you
                      find that for me again?" That kind of request to
                      basically find a described image (the book cover)
                      doesn't just happen in libraries anymore. As the
                      content of the Internet continues to shift from
                      text to graphics, accurately searching for images
                      based on a general description becomes more and
                      more important to some of the biggest Internet
                      companies. In the past couple of weeks,
                      researchers at both Google and Yahoo (owner of
                      Flickr) have posted some interesting news about
                      their recent work to improve image searching.
                    </p>
                    <div> </div>
                    <ul style="text-align: left;">
                      <li style="text-align: justify; font-size: 16px;
                        font-family: arial; line-height: 110%;"><a
href="http://googleresearch.blogspot.com/2014/11/a-picture-is-worth-thousand-coherent.html">A
                          picture is worth a thousand (coherent) words:
                          Building a natural description of images</a>
                        (Google Research Blog | Oriol Vinyals, Alexander
                        Toshev, Samy Bengio, and Dumitru Erhan) "But
                        accurately describing a complex scene requires a
                        deeper representation of what's going on in the
                        scene, capturing how the various objects relate
                        to one another and translating it all into
                        natural-sounding language. Many efforts to
                        construct computer-generated natural
                        descriptions of images propose combining current
                        state-of-the-art techniques in both <a
                          href="http://en.wikipedia.org/wiki/Computer_vision">computer
                          vision</a> and <a
                          href="http://en.wikipedia.org/wiki/Natural_language_processing">natural
                          language processing</a> to form a <a
                          href="https://www.cs.cmu.edu/%7Eafarhadi/papers/sentence.pdf">complete
                          image description approach</a>. But what if we
                        instead merged recent computer vision and
                        language models into a single jointly trained
                        system, taking an image and directly producing a
                        human readable sequence of words to describe
                        it?"</li>
                      <li style="text-align: justify; font-size: 16px;
                        font-family: arial; line-height: 110%;"><a
href="http://www.eweek.com/cloud/image-search-analysis-emerge-as-powerful-tools-privacy-threat-2.html">Image
                          search, analysis emerge as powerful tools,
                          privacy threat</a> (eWeek | Mike Elgan) "In a
                        nutshell, these systems identify objects in a
                        photograph-say, a boy, a dog, a ball, a tree, a
                        park, a bird, some clouds and so on-then use
                        sophisticated artificial intelligence to
                        understand that the boy is throwing the ball for
                        the dog to chase in a park and that the bird
                        isn't involved in the main action of the photo.
                        Combine this technology with face recognition
                        and anyone with access (which will be everyone)
                        will be able to search the Web for people doing
                        things or involved with or associated with some
                        activity."</li>
                      <li style="text-align: justify; font-size: 16px;
                        font-family: arial; line-height: 110%;"><a
href="http://yahoolabs.tumblr.com/post/103469857701/science-powering-product-yahoo-weather">Science
                          powering product: Yahoo Weather</a> (Yahoo
                        Labs | David A. Shamma, Jia Li, Lyndon Kennedy,
                        and Bart Thomée) "But even more difficult than
                        finding a stunning photo that accurately
                        reflects the weather in a given location is the
                        challenge of finding what the Flickr community
                        believes is an interesting weather photo. A
                        little while before we set out to surface our
                        one million photos, we made an observation about
                        how people designate photos on Flickr as
                        'favorites.' Something as simple as favorites
                        and likes on social network sites are rich
                        social signals that can be used to surface
                        themes of images."</li>
                      <li style="text-align: justify; font-size: 16px;
                        font-family: arial; line-height: 110%;"><a
href="http://arstechnica.com/business/2014/11/finding-an-image-with-an-image-and-other-feats-of-computer-vision/">Finding
                          an image with an image and other feats of
                          computer vision</a> (Ars Technica | Megan
                        Geuss) "Yahoo's efforts to make photo search
                        better has a simple mantra: 'more relevant
                        photos for users, not just the most popular
                        photos,' as Li put it. To that extent, Flickr
                        tries to improve general search while also
                        improving search relevance within a person's
                        likely-massive online photo album. Shamma noted
                        that batch upload and the gigabytes and
                        terabytes of storage offered to customers at
                        relatively cheap prices have changed how we
                        photograph things. Accordingly, storage and
                        recall of photographs has to adapt to fit the
                        morphing definition of photography. 'The
                        practice of photography is changing very
                        quickly, using photos for communication has been
                        growing,' Shamma said."</li>
                    </ul>
                    <div style="text-align: left;"> </div>
                    <p style="text-align: left; font-size: 20px;
                      font-family: arial; line-height: 110%;"><small><strong><em>Articles
                            from <a href="http://ohioweblibrary.org">Ohio
                              Web Library</a>:</em></strong></small><br>
                    </p>
                    <div style="text-align: justify; font-size: 16px;
                      font-family: arial; line-height: 110%;">
                      <ul>
                        <li><a
href="http://web.a.ebscohost.com.proxy.oplin.org/chc/detail?sid=4526ea1f-cbdc-4b82-8091-5653d181d008%40sessionmgr4001&vid=0&hid=4106&bdata=JnNpdGU9Y2hjLWxpdmU%3d#db=cmh&AN=86149983">Leveraging
                            multimedia in web search.</a> (<em>Online
                            Searcher</em>, March/April 2013, p62-64 |
                          Greg R. Notess)</li>
                        <li><a
href="http://web.a.ebscohost.com.proxy.oplin.org/lrc/detail/detail?sid=a99c1848-0729-4bc6-8ece-8b1c8c1f2732%40sessionmgr4003&vid=0&hid=4106&bdata=JnNpdGU9bHJjLWxpdmU%3d#db=lfh&AN=25902122">Getting
                            the picture.</a> (<em>European Journal of
                            English Studies</em>, Aug. 2007, p193-206 |
                          Julia Thomas)</li>
                        <li><a
href="http://web.a.ebscohost.com.proxy.oplin.org/ehost/detail/detail?sid=c1fb0d11-4c1c-4536-88b4-31ce07ab5bff%40sessionmgr4004&vid=0&hid=4106&bdata=JnNpdGU9ZWhvc3QtbGl2ZQ%3d%3d#db=aph&AN=74267899">Finding
                            images in an online public access catalogue:
                            Analysis of user queries, subject headings,
                            and description notes.</a> (<em>Canadian
                            Journal of Information & Library
                            Sciences</em>, Sept. 2010, p271-295 |
                          Youngok Choi and Ingrid Hsieh-Yee)</li>
                      </ul>
                    </div>
                    <div style="text-align: left;"> </div>
                    <!-- End paste of web source here --> </td>
                </tr>
                <tr>
                  <td style="" solid="" background-color:="" rgb(255,=""
                    255,="" 255);="" >=""
background="http://www.oplin.org/4cast/wp-content/themes/4cast/images/kubrickbgwide.jpg"
                    valign="top" width="760"><span style="font-size:
                      10px; color: rgb(96, 96, 96); line-height: 100%;
                      font-family: verdana;">
                      <hr><!-- Begin standard subscription verbiage -->
                      <div style="text-align: justify;">The <strong><em>OPLIN
                            4cast</em></strong>
                        is a weekly compilation of
                        recent headlines, topics, and trends that could
                        impact public
                        libraries. You can subscribe to it in a variety
                        of ways, such as: <br>
                      </div>
                      <div style="text-align: left;"> </div>
                      <ul>
                        <li style="text-align: justify;"><strong>RSS
                            feed.</strong>
                          You
                          can receive the OPLIN 4cast
                          via RSS feed by subscribing to the following
                          URL:
                          <a class="moz-txt-link-freetext" href="http://www.oplin.org/4cast/index.php/?feed=rss2">http://www.oplin.org/4cast/index.php/?feed=rss2</a>.
                        </li>
                        <li style="text-align: justify;"><strong>Live
                            Bookmark.</strong>
                          If you're using the Firefox
                          web browser, you can go to the 4cast website
                          (<a class="moz-txt-link-freetext" href="http://www.oplin.org/4cast/">http://www.oplin.org/4cast/</a>) and click on the
                          orange "radio wave" icon
                          on the right side of the address bar. In
                          Internet Explorer 7, click on
                          the same icon to view or subscribe to the
                          4cast RSS feed. </li>
                        <li style="text-align: justify;"><strong>E-mail.</strong>
                          You
                          can have the OPLIN 4cast
                          delivered via e-mail (a'la OPLINlist and
                          OPLINtech) by subscribing to
                          the 4cast mailing list at
                          <a class="moz-txt-link-freetext" href="http://mail.oplin.org/mailman/listinfo/OPLIN4cast">http://mail.oplin.org/mailman/listinfo/OPLIN4cast</a>.
                        </li>
                      </ul>
                    </span> </td>
                </tr>
                <tr>
                  <td style="border-top: 0px solid rgb(255, 255, 255);
                    background-color: rgb(255, 255, 204);"
background="http://www.oplin.org/4cast/wp-content/themes/4cast/images/kubrickfooter.jpg"
                    valign="top" width="760"> <br>
                  </td>
                </tr>
              </tbody>
            </table>
          </td>
        </tr>
      </tbody>
    </table>
    <small>
      <title>OPLIN 4Cast</title>
      <style>
 .headerTop { background-color:#FFFFFF; border-top:0px solid #000000; border-bottom:1px solid #FFFFFF; text-align:center; }
 .adminText { font-size:16px; color:#0000FF; line-height:200%; font-family:verdana; text-decoration:none; }
 .headerBar { background-color:#FFFFFF; border-top:0px solid #333333; border-bottom:0px solid #FFFFFF; }
 .title { font-size:20px; font-weight:bold; color:#000000; font-family:arial; line-height:110%; }
 .subTitle { font-size:11px; font-weight:normal; color:#000000; font-style:italic; font-family:arial; }
 .defaultText { font-size:12px; color:#000000; line-height:150%; font-family:trebuchet ms; }
 .footerRow { background-color:#FFFFCC; border-top:0px solid #FFFFFF; }
 .footerText { font-size:10px; color:#996600; line-height:100%; font-family:verdana; }
 a { color:#0000FF; color:#0000FF; color:#0000FF; }
  </style>
    </small>
  </body>
</html>