Jump to: navigation, search

Implicit User Feedback

Overview

To improve search quality, gather implicit user feedback via Proactive Engagement integration.

For example, sending additional implicit user feedback automatically in cases such as:

  • Customer executed the search, saw search results and left search result page in less then X seconds - > Negative feedback published for all documents in the result set.
  • Customer executed the search, saw result, opened the document and left the page in less then X seconds -> Negative feedback published for particular document.
  • Customer executed the search, saw result, opened the document and remained on the page for X minutes -> Positive feedback to be published for the document.
  • Customer executed the search, saw result, opened the document and opened attachment to the document-> Positive feedback to be published for the document.

    Customizing the Script and reconfiguring the routing strategy

    1. Add additional JavaScript code to the Web Monitoring Agent on the front-end page to monitor described events.
      For example:
      1. Create a small JavaScript function to wrap sending feedbacks:
        function feedbackSender
        (decision, behavior, timeoutSeconds){
            this.decision = decision;//positive,negative
            this.timeoutSeconds = timeoutSeconds;
            this.behavior = behavior;//less,more,equal
            this.openTime = null;
        this.sendPositiveFeedback = function(){
            console.log("Positive feedback");
        };
        this.sendNegativeFeedback = function(){
            console.log("Negative feedback");
        };
        this.sendFeedback = function () {
            switch (decision) {
              case "positive":
              this.sendPositiveFeedback();
                break;
              case "negative":
              this.sendNegativeFeedback();
                break;
              default:
                console.log("Sorry, we are out of " 
        + this.decision + ".");
                }
        };
        }
      2. Add methods to monitor timeout:
        this.onOpenPage = function () {
                this.openTime = new Date();
        };
        this.onLeftPage = function () {
            switch (behavior) {
              case "less":
                if ((this.openTime)&&(((new Date()) 
        - this.openTime) < this.timeoutSeconds * 1000)) {
                        this.sendFeedback();
                    }
                break;
              case "more":
                if ((this.openTime)&&(((new Date()) 
        - this.openTime) > this.timeoutSeconds * 1000)) {
                        this.sendFeedback();
                    }
                break;
              case "equals":
                if ((this.openTime)&&(((new Date()) 
        - this.openTime) == this.timeoutSeconds * 1000)) {
                        this.sendFeedback();
                    }
                break;
              default:
                console.log("Sorry, we are out of " 
        + this.behavior + ".");
                }
        }
      3. Implement provided business cases using this function. For example:
        • Case 1: Customer executed the search, saw search results and left search result page in less then X seconds - > Negative feedback published for all documents in the result set.
          firstCaseFeedback = new feedbackSender
          ('negative','less',5)
          firstCaseFeedback.sendNegativeFeedback=function()
          {
           _gt.push(['event', {eventName: 'Feedback', 
          gks_Reason: 'negative', gks_docIds: strArr, 
          gks_sessionId: gks_sessionId,
           gks_query: gks_query}]);
          }
          $(window).on('hashchange', function(e){
           if (window.location.hash.indexOf('/search/') 
          >= 0) {
           firstCaseFeedback.onOpenPage();
           arr=[];
           var $resultDiv = $('._gk-_wd-_rs');
           $resultDiv.ready(function () {
           $resultDiv.each(
           function(index, a) {
           var basicId = $(a).attr('id');
           arr.push(basicId.substring(18, basicId.length))
           }
           );
           })
           strArr = JSON.stringify(arr);
           gks_query = $('#searchContent').val();
           gks_sessionId = window.localStorage.getItem
          ('sessionId');
           }
          });
          $(window).on('hashchange', function(e){
           var oldURL = e.originalEvent.oldURL;
           if (oldURL.indexOf('/search/') >= 0) {
           firstCaseFeedback.onLeftPage();
           }
           }
          });
        • Case 2: Customer executed the search, saw result, opened the document and left the page in less then X seconds -> Negative feedback published for particular document.
          secondCaseFeedback = new feedbackSender
          ('negative','less',10)
          secondCaseFeedback.sendNegativeFeedback=function()
          {
              _gt.push(['event', {eventName: 'Feedback', 
          gks_Reason: 'negative', gks_docIds: 
          gks_docId, gks_sessionId: gks_sessionId,
           gks_query: gks_query}]);
          }
           
           $(window).on('hashchange', function(e){
              if (window.location.hash.indexOf('/document/') 
          >= 0) {
              gks_query = $('#searchContent').val();
              gks_sessionId = window.localStorage.getItem
          ('sessionId');
              gks_docId = window.location.hash.substr
          (window.location.hash.length - 36);
              secondCaseFeedback.onOpenPage();
              }
          });
           
          $(window).on('hashchange', function(e){
              var oldURL = e.originalEvent.oldURL;
              if (oldURL.indexOf('/document/') >= 0) {
                  secondCaseFeedback.onLeftPage();
              }
          });
        • Case 3: Customer executed the search, saw result, opened the document and remained on the page for X minutes -> Positive feedback to be published for the document.
          thirdCaseFeedback = new feedbackSender
          ('positive','more',20)
          thirdCaseFeedback.sendPositiveFeedback=function()
          {
              _gt.push(['event', {eventName: 'Feedback', 
          gks_Reason: 'positive', gks_docIds: 
          gks_docId, gks_sessionId: gks_sessionId,
           gks_query: gks_query}]);
          }
           
          thirdCaseFeedback.onOpenPage=function(){
              thirdCaseFeedback.openTime = new Date();
              setTimeout(function() 
          {thirdCaseFeedback.onLeftPage();}, 
          (thirdCaseFeedback.timeoutSeconds+1)*1000);   
          }
           
          $(window).on('hashchange', function(e){
              if (window.location.hash.indexOf('/document/') 
          >= 0) {
              gks_query = $('#searchContent').val();
              gks_sessionId = window.localStorage.getItem
          ('sessionId');
              gks_docId = window.location.hash.substr
          (window.location.hash.length - 36);
              thirdCaseFeedback.onOpenPage();
              }
          });
    2. Create a business rule to invoke interaction on this event:
      rule "Rule-101 Feedback"
       
      salience 100001
       
          agenda-group "level0"
       
          dialect "mvel"
       
          when
       
               $event1: Event(eval($event1.getName().equals
      ('Feedback')))
       
          then
       
              sendEvent($event1, ed, drools);
       
      end
    3. Modify the strategy to route the interaction invoked by these events in a separate branch.
    4. Use modules to send REST requests in these branches to publish feedback to the Knowledge server:
      Positive feedback:
       
      POST
      URL: http://<host>:<port>/gks-server/v1/feedback/
      knowledgefaq/documents/<docId>/
      vote?relevant=true&sessionId=<sessionId>
      BODY: {"query":"<query>"}
       
      Negative feedback:
       
      POST
      URL: http://<host>:<port>/gks-server/v1/feedback/
      knowledgefaq/documents/<docId>/
      vote?relevant=false&sessionId=<sessionId>
      BODY: {"query":"<query>"}
This page was last edited on November 25, 2016, at 17:30.
Comments or questions about this documentation? Contact us for support!