huginn_naive_bayes_agent 0.1.2 → 0.1.3

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA1:
-  metadata.gz: 6d1a1c1469bb599fe8a5a42a94e89fd6206e519d
-  data.tar.gz: d2ce17b2c8471199add13b73c92047cb84881c4a
+  metadata.gz: c4bf23b64717fc62020b16dcc96b3a9767e71f42
+  data.tar.gz: d3687649e9758f6f57e71478ec4edeb47c9d4c1e
 SHA512:
-  metadata.gz: cb4ae84beb82c52900b99d21c9ca3d542f9e99f1bee34794ef1bb92e6191559d89f4afa6c8a7f5cbf5918e1967df8dadd0c89c90bd9d8f7ffd32382017a4018e
-  data.tar.gz: a4c1e57328ddc1fef840f73aeeeb5072fc251b2d5fe03e649c175e43b46515775f0fa74336d3bbd1964280ee83276e6a8a0d431a786b7c59998c1ec188cdcc07
+  metadata.gz: 97a60c81731ae167e119b8bb68852341f15d7d5e1d6dd03ce728662a9adcdbe42ca5b2886dcfa8ba9e3c4cfe9186fd6f621f916dc3a303cd7a24c29182bcaf27
+  data.tar.gz: 86bc9dfe3906f16cac1851bc38257768ded8fbb7611a527c35946e19ac03939dd61e511e4056c44cc4171331be44f6c244fe13081c34fd2d18c37757c60c746b

data/lib/huginn_naive_bayes_agent/naive_bayes_agent.rb

@@ -18,7 +18,9 @@ module Agents
       
       However, if `nb_cats` is already populated, then the content from `nb_content` will be used as training data for the categories listed in `nb_cats`. For instance, say `nb_cats` consists of `trees`. Then `nb_content` will be used as training data for the category `trees`. The data is saved to the agent memory. 
       
-      When an event is received for classification, the Naive Bayes Agent will assign a value between 0 and 1 representing the likelihood that it falls under a category. The `min_value` option lets you choose the minimum threshold that must be reached before the event is labeled with that category. If `min_value` is set to 1, then the event is labeled with whichever category has the highest value.
+      Data in `nb_content` can be cleaned before classification. If `strip_punctuation` is set to true, the text in `nb_content` is stripped of punctuation before it is sent to the classifier. The changes are not saved to `nb_content`.
+            
+      When an event is received for classification, the Naive Bayes Agent will assign a value between 0 and 1 representing the likelihood that it falls under a category. The `min_value` option lets you choose the minimum threshold that must be reached before the event is labeled with that category. If `min_value` is set to 1, then the event is labeled with whichever category has the highest value. 
       
       The option `propagate_training_events` lets you choose whether the training events are emitted along with the classified events. If it is set to false, then no new event will be created from events that already had categories when they were received.
       
@@ -43,7 +45,8 @@ module Agents
       {
         'min_value' => "0.5",
         'propagate_training_events' => 'true',
-        'expected_update_period_in_days' => "7"
+        'expected_update_period_in_days' => "7",
+        'strip_punctuation' => 'false'
       }
     end
 
@@ -59,6 +62,16 @@ module Agents
     def receive(incoming_events)
       incoming_events.each do |event|
         nbayes = load(memory['data'])
+        # validate incoming payload
+        if !event.payload['nb_cats']
+          error("Missing `nb_cats` field in the event payload. #{event.payload.to_s}")
+		  raise 'Missing `nb_cats` field in the event payload.'
+        end
+        if !event.payload['nb_content']
+          error("Missing `nb_content` field in the event payload. #{event.payload.to_s}")
+          raise 'Missing `nb_content` field in the event payload.'
+        end 
+        # train or modify existing classifier
         if event.payload['nb_cats'].length > 0 and not event.payload['nb_cats'].include?("=class")
           cats = event.payload['nb_cats'].split(/\s+/)
           if cats[0] == "=loadYML"
@@ -73,16 +86,25 @@ module Agents
             nbayes.purge_less_than(event.payload['nb_content'].to_i)
             memory['data'] = YAML.dump(nbayes)
           else
+            nb_content = event.payload['nb_content']
+            if interpolated['strip_punctuation'] == "true"
+              nb_content = nb_content.gsub(/[^[:word:]\s]/, '') #https://stackoverflow.com/a/10074271
+            end
             cats.each do |c|
-              c.starts_with?('-') ? nbayes.untrain(event.payload['nb_content'].split(/\s+/), c[1..-1]) : nbayes.train(event.payload['nb_content'].split(/\s+/), c)
+              c.starts_with?('-') ? nbayes.untrain(nb_content.split(/\s+/), c[1..-1]) : nbayes.train(nb_content.split(/\s+/), c)
             end
             memory['data'] = YAML.dump(nbayes)
-            if interpolated['propagate_training_events'] = "true"
+            if interpolated['propagate_training_events'] == "true"
               create_event payload: event.payload
             end
           end
+        # classify new data
         else
-          result = nbayes.classify(event.payload['nb_content'].split(/\s+/))
+          nb_content = event.payload['nb_content']
+          if interpolated['strip_punctuation'] == "true"
+            nb_content = nb_content.gsub(/[^[:word:]\s]/, '') #https://stackoverflow.com/a/10074271
+          end
+          result = nbayes.classify(nb_content.split(/\s+/))
           if interpolated['min_value'].to_f == 1
             event.payload['nb_cats'] << (event.payload['nb_cats'].length == 0 ? result.max_class : " "+result.max_class)
           else

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: huginn_naive_bayes_agent
 version: !ruby/object:Gem::Version
-  version: 0.1.2
+  version: 0.1.3
 platform: ruby
 authors:
 - Noah Greenstein
 autorequire: 
 bindir: bin
 cert_chain: []
-date: 2017-05-04 00:00:00.000000000 Z
+date: 2017-05-13 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: bundler