Pull request #46: Add rule redacting sponsor companies if preceded by prefix

Merge in RED/redaction-service from RED-301 to master

* commit '6a8f3665198ec2d7452ed0dac43b9a7d220c8bc5':
  Remove entries from must_redact dict, add test and refactor rules
  Add rule redacting sponsor companies if preceded by prefix

redaction-service-v1/pom.xml

@@ -22,7 +22,7 @@
     </modules>
 
     <properties>
-        <pdfbox.version>2.0.16</pdfbox.version>
+        <pdfbox.version>2.0.21</pdfbox.version>
     </properties>
 
 

redaction-service-v1/redaction-service-server-v1/pom.xml

@@ -11,25 +11,6 @@
 
     <artifactId>redaction-service-server-v1</artifactId>
 
-    <properties>
-        <pdfbox.version>2.0.20</pdfbox.version>
-    </properties>
-
-    <dependencyManagement>
-        <dependencies>
-            <dependency>
-                <groupId>org.apache.pdfbox</groupId>
-                <artifactId>pdfbox</artifactId>
-                <version>${pdfbox.version}</version>
-            </dependency>
-            <dependency>
-                <groupId>org.apache.pdfbox</groupId>
-                <artifactId>pdfbox-tools</artifactId>
-                <version>${pdfbox.version}</version>
-            </dependency>
-        </dependencies>
-    </dependencyManagement>
-
     <dependencies>
         <dependency>
             <groupId>com.iqser.red.service</groupId>

redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/redaction/model/Section.java

@@ -44,7 +44,7 @@ public class Section {
     }
 
 
-    public boolean contains(String type) {
+    public boolean matchesType(String type) {
 
         return entities.stream().anyMatch(entity -> entity.getType().equals(type));
     }
@@ -80,6 +80,18 @@ public class Section {
     }
 
 
+    public void redactIfPrecededBy(String prefix, String type, int ruleNumber, String reason) {
+
+        entities.forEach(entity -> {
+            if (entity.getType().equals(type) && searchText.indexOf(prefix + entity.getWord()) != 1) {
+                entity.setRedaction(true);
+                entity.setMatchedRule(ruleNumber);
+                entity.setRedactionReason(reason);
+            }
+        });
+    }
+
+
     public void redactLineAfter(String start, String asType, int ruleNumber, String reason) {
 
         String[] values = StringUtils.substringsBetween(text, start, "\n");
@@ -141,7 +153,8 @@ public class Section {
 
             if (startIndex > -1 && (startIndex == 0 || Character.isWhitespace(searchText.charAt(startIndex - 1)) || isSeparator(searchText
                     .charAt(startIndex - 1))) && (stopIndex == searchText.length() || isSeparator(searchText.charAt(stopIndex)))) {
-                found.add(new Entity(searchText.substring(startIndex, stopIndex), asType, startIndex, stopIndex, headline, sectionNumber));
+                found.add(new Entity(searchText.substring(startIndex, stopIndex), asType, startIndex, stopIndex,
+                        headline, sectionNumber));
             }
         } while (startIndex > -1);
 
@@ -197,7 +210,8 @@ public class Section {
         if (value == null) {
             log.warn("Could not find any data for {}.", cellHeader);
         } else {
-            Entity entity = new Entity(value.toString(), type, value.getRowSpanStart(), value.getRowSpanStart() + value.toString()
+            Entity entity = new Entity(value.toString(), type, value.getRowSpanStart(),
+                    value.getRowSpanStart() + value.toString()
                     .length(), headline, sectionNumber);
             entity.setRedaction(redact);
             entity.setMatchedRule(ruleNumber);
@@ -206,7 +220,7 @@ public class Section {
                     .getSequences()); // Make sure no other cells with same content are highlighted
 
             // HashSet keeps the older value, but we want the new only.
-            if(entities.contains(entity)){
+            if (entities.contains(entity)) {
                 entities.remove(entity);
             }
             entities.add(entity);

redaction-service-v1/redaction-service-server-v1/src/test/java/com/iqser/red/service/redaction/v1/server/RedactionIntegrationTest.java

@@ -63,6 +63,7 @@ public class RedactionIntegrationTest {
     private static final String VERTEBRATES_CODE = "vertebrate";
     private static final String ADDRESS_CODE = "address";
     private static final String NAME_CODE = "name";
+    private static final String SPONSOR = "sponsor";
     private static final String NO_REDACTION_INDICATOR = "no_redaction_indicator";
     private static final String REDACTION_INDICATOR = "redaction_indicator";
     private static final String HINT_ONLY = "hint_only";
@@ -117,6 +118,7 @@ public class RedactionIntegrationTest {
         when(dictionaryClient.getDictionaryForType(VERTEBRATES_CODE)).thenReturn(getDictionaryResponse(VERTEBRATES_CODE));
         when(dictionaryClient.getDictionaryForType(ADDRESS_CODE)).thenReturn(getDictionaryResponse(ADDRESS_CODE));
         when(dictionaryClient.getDictionaryForType(NAME_CODE)).thenReturn(getDictionaryResponse(NAME_CODE));
+        when(dictionaryClient.getDictionaryForType(SPONSOR)).thenReturn(getDictionaryResponse(SPONSOR));
         when(dictionaryClient.getDictionaryForType(NO_REDACTION_INDICATOR)).thenReturn(getDictionaryResponse(NO_REDACTION_INDICATOR));
         when(dictionaryClient.getDictionaryForType(REDACTION_INDICATOR)).thenReturn(getDictionaryResponse(REDACTION_INDICATOR));
         when(dictionaryClient.getDictionaryForType(HINT_ONLY)).thenReturn(getDictionaryResponse(HINT_ONLY));
@@ -132,6 +134,11 @@ public class RedactionIntegrationTest {
                         .stream()
                         .map(this::cleanDictionaryEntry)
                         .collect(Collectors.toSet()));
+        dictionary.computeIfAbsent(SPONSOR, v -> new ArrayList<>())
+                .addAll(ResourceLoader.load("dictionaries/sponsor_companies.txt")
+                        .stream()
+                        .map(this::cleanDictionaryEntry)
+                        .collect(Collectors.toSet()));
         dictionary.computeIfAbsent(VERTEBRATES_CODE, v -> new ArrayList<>())
                 .addAll(ResourceLoader.load("dictionaries/vertebrates.txt")
                         .stream()
@@ -176,6 +183,7 @@ public class RedactionIntegrationTest {
         typeColorMap.put(VERTEBRATES_CODE, new float[]{0, 1, 0});
         typeColorMap.put(ADDRESS_CODE, new float[]{0, 1, 1});
         typeColorMap.put(NAME_CODE, new float[]{1, 1, 0});
+        typeColorMap.put(SPONSOR, new float[]{.5f, .5f, .5f});
         typeColorMap.put(NO_REDACTION_INDICATOR, new float[]{0.8f, 0, 0.8f});
         typeColorMap.put(REDACTION_INDICATOR, new float[]{1, 0.502f, 0.1f});
         typeColorMap.put(HINT_ONLY, new float[]{0.8f, 1, 0.8f});
@@ -184,6 +192,7 @@ public class RedactionIntegrationTest {
         hintTypeMap.put(VERTEBRATES_CODE, true);
         hintTypeMap.put(ADDRESS_CODE, false);
         hintTypeMap.put(NAME_CODE, false);
+        hintTypeMap.put(SPONSOR, false);
         hintTypeMap.put(NO_REDACTION_INDICATOR, true);
         hintTypeMap.put(REDACTION_INDICATOR, true);
         hintTypeMap.put(HINT_ONLY, true);
@@ -192,6 +201,7 @@ public class RedactionIntegrationTest {
         caseInSensitiveMap.put(VERTEBRATES_CODE, true);
         caseInSensitiveMap.put(ADDRESS_CODE, false);
         caseInSensitiveMap.put(NAME_CODE, false);
+        caseInSensitiveMap.put(SPONSOR, false);
         caseInSensitiveMap.put(NO_REDACTION_INDICATOR, true);
         caseInSensitiveMap.put(REDACTION_INDICATOR, true);
         caseInSensitiveMap.put(HINT_ONLY, true);
@@ -270,7 +280,7 @@ public class RedactionIntegrationTest {
 
         System.out.println("redactionTest");
         long start = System.currentTimeMillis();
-        ClassPathResource pdfFileResource = new ClassPathResource("files/Fludioxonil/50 Fludioxonil_RAR_01_Volume_1_2018-02-21.pdf");
+        ClassPathResource pdfFileResource = new ClassPathResource("files/Metolachlor/S-Metolachlor_RAR_02_Volume_2_2018-09-06.pdf");
 
         RedactionRequest request = RedactionRequest.builder()
                 .document(IOUtils.toByteArray(pdfFileResource.getInputStream()))
@@ -435,13 +445,35 @@ public class RedactionIntegrationTest {
         RedactionResult result = redactionController.redact(request);
 
         result.getRedactionLog().getRedactionLogEntry().forEach(entry -> {
-            if(!entry.isHint()){
+            if (!entry.isHint()) {
                 assertThat(entry.getReason()).isEqualTo("Not redacted because row is not a vertebrate study");
             }
         });
     }
 
 
+    @Test
+    public void sponsorCompanyTest() throws IOException {
+
+        long start = System.currentTimeMillis();
+        ClassPathResource pdfFileResource = new ClassPathResource("files/Minimal Examples/sponsor_companies.pdf");
+
+        RedactionRequest request = RedactionRequest.builder()
+                .document(IOUtils.toByteArray(pdfFileResource.getInputStream()))
+                .build();
+        request.setFlatRedaction(false);
+
+        RedactionResult result = redactionController.redact(request);
+
+        try (FileOutputStream fileOutputStream = new FileOutputStream("/tmp/Redacted.pdf")) {
+            fileOutputStream.write(result.getDocument());
+        }
+        long end = System.currentTimeMillis();
+
+        System.out.println("duration: " + (end - start));
+        System.out.println("numberOfPages: " + result.getNumberOfPages());
+    }
+
 
     private static String loadFromClassPath(String path) {
 

redaction-service-v1/redaction-service-server-v1/src/test/resources/dictionaries/must_redact.txt

@@ -1,3 +1,2 @@
-Batches Produced at
 CTL
 determination of residues

redaction-service-v1/redaction-service-server-v1/src/test/resources/dictionaries/sponsor_companies.txt

@@ -0,0 +1,7 @@
+Monthey Syngenta Crop Protection AG, Basel, Switzerland
+Syngenta Crop Protection, Monthey, Switzerland
+Fine Organics Limited, Middlesbrough, United Kingdom
+Syngenta Monthey Switzerland
+Hunan Haili Chemical Industry Co., Ltd., Hunan, China
+Syngenta, Switzerland
+Syngenta Nantong, China

redaction-service-v1/redaction-service-server-v1/src/test/resources/drools/rules.drl

@@ -7,7 +7,7 @@ global Section section
 
 rule "1: Redacted because Section contains Vertebrate"
 	when
-		eval(section.contains("vertebrate")==true);
+		Section(matchesType("vertebrate"))
 	then
 		section.redact("name", 1, "Redacted because Section contains Vertebrate");
 		section.redact("address", 1, "Redacted because Section contains Vertebrate");
@@ -16,7 +16,7 @@ rule "1: Redacted because Section contains Vertebrate"
 
 rule "2: Not Redacted because Section contains no Vertebrate"
 	when
-		eval(section.contains("vertebrate")==false);
+		Section(matchesType("vertebrate"))
 	then
 		section.redactNot("name", 2, "Not Redacted because Section contains no Vertebrate");
 		section.redactNot("address", 2, "Not Redacted because Section contains no Vertebrate");
@@ -25,7 +25,7 @@ rule "2: Not Redacted because Section contains no Vertebrate"
 
 rule "3: Do not redact Names and Addresses if no redaction Indicator is contained"
 	when
-		eval(section.contains("vertebrate")==true && section.contains("no_redaction_indicator")==true);
+		Section(matchesType("vertebrate"), matchesType("no_redaction_indicator"))
 	then
 		section.redactNot("name", 3, "Vertebrate was found, but also a no redaction indicator");
 		section.redactNot("address", 3, "Vertebrate was found, but also a no redaction indicator");
@@ -34,7 +34,7 @@ rule "3: Do not redact Names and Addresses if no redaction Indicator is containe
 
 rule "4: Redact Names and Addresses if no_redaction_indicator and redaction_indicator is contained"
 	when
-		eval(section.contains("vertebrate")==true && section.contains("no_redaction_indicator")==true && section.contains("redaction_indicator")==true);
+		Section(matchesType("vertebrate"), matchesType("no_redaction_indicator"), matchesType("redaction_indicator"))
 	then
 		section.redact("name", 4, "Vertebrate was found and no_redaction_indicator and redaction_indicator");
 		section.redact("address", 4, "Vertebrate was found and no_redaction_indicator and redaction_indicator");
@@ -43,15 +43,16 @@ rule "4: Redact Names and Addresses if no_redaction_indicator and redaction_indi
 
 rule "5: Do not redact in guideline sections"
 	when
-		eval(section.headlineContainsWord("guideline") || section.headlineContainsWord("Guidance"));
+		Section(headlineContainsWord("guideline") || headlineContainsWord("Guidance"))
 	then
 		section.redactNot("name", 5, "Section is a guideline section.");
 		section.redactNot("address", 5, "Section is a guideline section.");
 	end
 
+
 rule "6: Redact contact information if applicant is found"
 	when
-		eval(section.headlineContainsWord("applicant") || section.getText().contains("Applicant") || section.headlineContainsWord("Primary contact") || section.headlineContainsWord("Alternative contact"));
+		Section(headlineContainsWord("applicant") || text.contains("Applicant") || headlineContainsWord("Primary contact") || headlineContainsWord("Alternative contact"))
 	then
 		section.redactLineAfter("Name:", "address", 6, "Applicant information was found");
 		section.redactBetween("Address:", "Contact", "address", 6, "Applicant information was found");
@@ -78,7 +79,7 @@ rule "6: Redact contact information if applicant is found"
 
 rule "7: Redact contact information if Producer is found"
 	when
-		eval(section.getText().toLowerCase().contains("producer of the plant protection") || section.getText().toLowerCase().contains("producer of the active substance") || section.getText().contains("Manufacturer of the active substance") || section.getText().contains("Manufacturer:") || section.getText().contains("Producer or producers of the active substance"));
+		Section(text.toLowerCase().contains("producer of the plant protection") || text.toLowerCase().contains("producer of the active substance") || text.contains("Manufacturer of the active substance") || text.contains("Manufacturer:") || text.contains("Producer or producers of the active substance"))
 	then
 		section.redactLineAfter("Name:", "address", 7, "Producer was found");
 		section.redactBetween("Address:", "Contact", "address", 7, "Producer was found");
@@ -109,7 +110,7 @@ rule "8: Not redacted because Vertebrate Study = N"
 
 rule "9: Redact if must redact entry is found"
 	when
-		eval(section.contains("must_redact")==true);
+		Section(matchesType("must_redact"))
 	then
 		section.redact("name", 9, "must_redact entry was found.");
 		section.redact("address", 9, "must_redact entry was found.");
@@ -124,3 +125,10 @@ rule "10: Redact Authors and Addresses in Reference Table if it is a Vertebrate
         section.redact("address", 10, "Redacted because row is a vertebrate study");
         section.highlightCell("Vertebrate study Y/N", 10, "must_redact");
     end
+
+rule "11: Redact sponsor company"
+    when
+        Section(text.toLowerCase().contains("batches produced at"))
+    then
+        section.redactIfPrecededBy("batches produced at", "sponsor", 11, "Redacted because it represents a sponsor company");
+    end