cdmlib-services/src/main/java/eu/etaxonomy/cdm/api/service/NaturalLanguageGenerator.java

   1 package eu.etaxonomy.cdm.api.service;
   2
   3 import java.util.ArrayList;
   4 import java.util.HashSet;
   5 import java.util.Iterator;
   6 import java.util.List;
   7 import java.util.Map;
   8 import java.util.Set;
   9
  10 import org.apache.commons.lang.StringUtils;
  11 import org.apache.log4j.Logger;
  12 import org.springframework.stereotype.Component;
  13
  14 import eu.etaxonomy.cdm.model.description.CategoricalData;
  15 import eu.etaxonomy.cdm.model.description.DescriptionElementBase;
  16 import eu.etaxonomy.cdm.model.description.Feature;
  17 import eu.etaxonomy.cdm.model.description.FeatureNode;
  18 import eu.etaxonomy.cdm.model.description.FeatureTree;
  19 import eu.etaxonomy.cdm.model.description.QuantitativeData;
  20 import eu.etaxonomy.cdm.model.description.TaxonDescription;
  21 import eu.etaxonomy.cdm.model.description.TextData;
  22 import eu.etaxonomy.cdm.model.description.TextFormat;
  23 import eu.etaxonomy.cdm.model.common.Annotation;
  24 import eu.etaxonomy.cdm.model.common.AnnotationType;
  25 import eu.etaxonomy.cdm.model.common.Language;
  26
  27
  28 /**
  29  * Generator of natural language descriptions from TaxonDescriptions.
  30  *
  31  * @author m.venin
  32  * @created 13.04.2010
  33  * @version 1.0
  34  */
  35 @Component
  36 public class NaturalLanguageGenerator implements INaturalLanguageGenerator {
  37         @SuppressWarnings("unused")
  38         private static final Logger logger = Logger.getLogger(NaturalLanguageGenerator.class);
  39
  40         private String firstSeparator = ",";
  41         private String secondSeparator = ".";
  42         private List<Integer> levels = new ArrayList<Integer>();
  43
  44         private DescriptionBuilder<QuantitativeData> quantitativeDescriptionBuilder = new DefaultQuantitativeDescriptionBuilder();
  45         private DescriptionBuilder<CategoricalData> categoricalDescriptionBuilder = new DefaultCategoricalDescriptionBuilder();
  46
  47         private TextData previousTextData;
  48
  49         private Map<String, INaturalLanguageTextDataProcessor> elementProcessors;
  50
  51         private Set<INaturalLanguageTextDataProcessor> applicableElementProcessors = new HashSet<INaturalLanguageTextDataProcessor>();
  52
  53         /**
  54          * Change the first separator used by generateSingleTextData. By default ",".
  55          *
  56          * @param separator
  57          */
  58         public void setFirstSeparator(String separator){
  59                 firstSeparator=separator;
  60         }
  61
  62         public String getFirstSeparator(){
  63                 return firstSeparator;
  64         }
  65
  66         /**
  67          * Change the second separator used by generateSingleTextData. By default ".".
  68          *
  69          * @param separator
  70          */
  71         public void setSecondSeparator(String separator){
  72                 secondSeparator=separator;
  73         }
  74
  75         public String getSecondSeparator(){
  76                 return secondSeparator;
  77         }
  78
  79         /**
  80          * @param quantitativeDescriptionBuilder
  81          */
  82         public void setQuantitativeDescriptionBuilder(DescriptionBuilder<QuantitativeData> quantitativeDescriptionBuilder){
  83                 this.quantitativeDescriptionBuilder = quantitativeDescriptionBuilder;
  84         }
  85
  86         /**
  87          * @param categoricalDescriptionBuilder
  88          */
  89         public void setCategoricalDescriptionBuilder(DescriptionBuilder<CategoricalData> categoricalDescriptionBuilder){
  90                 this.categoricalDescriptionBuilder = categoricalDescriptionBuilder;
  91         }
  92
  93         /**
  94          * @return the element processors of this generator
  95          */
  96         public Map<String, INaturalLanguageTextDataProcessor> getElementProcessors() {
  97                 return elementProcessors;
  98         }
  99
 100         /**
 101          * The keys of the elementProcessors map are regular expressions which are
 102          * being used to identify the those Descriptions to which the mapped
 103          * NaturalLanguageTextDataProcessor is applicable.
 104          *
 105          * @param elementProcessors
 106          */
 107         public void setElementProcessors(
 108                         Map<String, INaturalLanguageTextDataProcessor> elementProcessors) {
 109                 this.elementProcessors = elementProcessors;
 110         }
 111
 112         /**
 113          * Looks for technical annotations, if one matches a regular expression of the element processors
 114          * the associated processor is added to the applicable element processors which will then be applied
 115          * when generating the description.
 116          *
 117          * @param annotations the set of annotations of the description
 118          */
 119         private void initNaturalLanguageDescriptionElementProcessors(Set<Annotation> annotations) {
 120
 121                 if(annotations != null){
 122                         for(Annotation annotation : annotations){
 123                                 if(annotation.getAnnotationType().equals(AnnotationType.TECHNICAL())){
 124                                         if (elementProcessors!=null){
 125                                                 for(String regex : elementProcessors.keySet()){
 126                                                         if(annotation.getText().matches(regex)){
 127                                                                 applicableElementProcessors.add(elementProcessors.get(regex));
 128                                                         }
 129                                                 }
 130                                         }
 131                                 }
 132                         }
 133                 }
 134         }
 135
 136
 137         /**
 138          * Applies the list of applicable processors to a TextData.
 139          *
 140          * @param textData the TextData to be modified
 141          * @param previousTextData the TextData corresponding to the feature of the previous level in the tree
 142          */
 143         private void applyNaturalLanguageDescriptionElementProcessors(TextData textData, TextData previousTextData){
 144                 for(INaturalLanguageTextDataProcessor processor : applicableElementProcessors){
 145                         processor.process(textData, previousTextData);
 146                 }
 147         }
 148
 149
 150         /**
 151          * The most simple function to generate a description. The language used is the default one.
 152          *
 153          * @param featureTree the FeatureTree holding the order in which features and their states must be printed
 154          * @param description the TaxonDescription with all the data
 155          *
 156          * @return a list of TextData, each one being a basic element of the natural language description
 157          */
 158         public List<TextData> generateNaturalLanguageDescription(FeatureTree featureTree,TaxonDescription description) {
 159                 return generateNaturalLanguageDescription(featureTree,description,Language.DEFAULT());
 160         }
 161
 162
 163
 164         /**
 165          * Generate a description in a specified language.
 166          *
 167          * @param featureTree the FeatureTree holding the order in which features and their states must be printed
 168          * @param description the TaxonDescription with all the data
 169          * @param language the language in which the description has to be printed
 170          *
 171          * @return a list of TextData, each one being a basic element of the natural language description
 172          */
 173         public List<TextData> generateNaturalLanguageDescription(FeatureTree featureTree, TaxonDescription description, Language language) {
 174                 List<Language> languages = new ArrayList<Language>();
 175                 languages.add(language);
 176                 initNaturalLanguageDescriptionElementProcessors(description.getAnnotations());
 177                 return generatePreferredNaturalLanguageDescription(featureTree,description,languages);
 178         }
 179
 180         /**
 181          * Generate a description with a specified list of preferred languages.
 182          *
 183          * @param featureTree the FeatureTree holding the order in which features and their states must be printed
 184          * @param description the TaxonDescription with all the data
 185          * @param languages the ordered list of languages preferred for printing the description
 186          *
 187          * @return a list of TextData, each one being a basic element of the natural language description
 188          */
 189         public List<TextData> generatePreferredNaturalLanguageDescription(FeatureTree featureTree,TaxonDescription description, List<Language> languages) {
 190                 initNaturalLanguageDescriptionElementProcessors(description.getAnnotations());
 191                 return buildBranchesDescr(featureTree.getRootChildren(), featureTree.getRoot(), description, languages,0);
 192         }
 193
 194         /**
 195          * Generate a description as a single paragraph in a TextData.
 196          *
 197          * @param featureTree the FeatureTree holding the order in which features and their states must be printed
 198          * @param description the TaxonDescription with all the data
 199          *
 200          * @return a TextData in the default language.
 201          */
 202         public TextData generateSingleTextData(FeatureTree featureTree, TaxonDescription description) {
 203                 return generateSingleTextData(featureTree,description,Language.DEFAULT());
 204         }
 205
 206         /**
 207          * Generate a description as a single paragraph in a TextData.
 208          *
 209          * @param featureTree the FeatureTree holding the order in which features and their states must be printed
 210          * @param description the TaxonDescription with all the data
 211          * @param language the language in which the description has to be printed
 212          *
 213          * @return a TextData in the specified language.
 214          */
 215         public TextData generateSingleTextData(FeatureTree featureTree, TaxonDescription description, Language language) {
 216                 List<Language> languages = new ArrayList<Language>();
 217                 languages.add(language);
 218                 return generatePreferredSingleTextData(featureTree,description,languages);
 219         }
 220
 221         /**
 222          * Generate a description with a specified list of preferred languages.
 223          *
 224          * @param featureTree the FeatureTree holding the order in which features and their states must be printed
 225          * @param description the TaxonDescription with all the data
 226          * @param languages the ordered list of languages preferred for printing the description
 227          *
 228          * @return a TextData using the languages (in the given order of preference)
 229          */
 230         public TextData generatePreferredSingleTextData(FeatureTree featureTree, TaxonDescription description, List<Language> languages) {
 231                 levels.clear(); // before the start, the table containing the levels of each node must be cleared
 232                 // Note: this is not the most efficient way to keep track of the levels of the nodes but it allows some flexibility
 233                 List<TextData> texts = generatePreferredNaturalLanguageDescription(featureTree,description, languages);// first get the description as a raw list of TextData
 234
 235                 StringBuilder descriptionStringBuilder = new StringBuilder(); // the StringBuilder used to generate the description
 236                 int i = 0,j,level; // i is used to store the index of the TextData to use
 237                 boolean startSentence = false, firstOne = true;
 238
 239                 for (j=0 ; j<levels.size() ; j++){
 240                         level = levels.get(j);
 241                         if (level==-1){
 242                                 if ((j+1)<levels.size() && levels.get(j+1).equals(0)){ // if this node is the direct father of a leaf
 243                                         descriptionStringBuilder.append(secondSeparator + " ");
 244                                         startSentence=true;
 245                                         firstOne=false;
 246                                         String asString = texts.get(i).getText(Language.DEFAULT()).toString();
 247                                         if (asString.length()>1) descriptionStringBuilder.append(asString.substring(0,1).toUpperCase() + asString.substring(1));
 248                                 }
 249                                 i++;
 250                         }
 251                         else if (level==0) { // if this node is a leaf
 252                                 if (startSentence) descriptionStringBuilder.append(texts.get(i).getText(Language.DEFAULT()));
 253                                 else descriptionStringBuilder.append(firstSeparator + texts.get(i).getText(Language.DEFAULT()));
 254                                 startSentence=false;
 255                                 i++;
 256                         }
 257                         else {
 258                                 if (!firstOne && levels.get(j-1).equals(0)){ // if this node corresponds to the states linked to the previous leaf
 259                                         if (i<texts.size()) descriptionStringBuilder.append(texts.get(i).getText(Language.DEFAULT()));
 260                                         i++;
 261                                 }
 262                         }
 263                 }
 264                 descriptionStringBuilder.append(secondSeparator);
 265                 String returnString = descriptionStringBuilder.toString();
 266                 returnString = StringUtils.replace(returnString, "  ", " ");
 267                 returnString = StringUtils.removeStart(returnString, secondSeparator + " ");
 268                 return TextData.NewInstance(returnString,Language.DEFAULT(),TextFormat.NewInstance("", "Text", ""));
 269         }
 270
 271
 272
 273         /** recursive function that goes through a tree containing the order in which the description has to be generated,
 274          *  if an element of this tree matches one of the TaxonDescription, a DescriptionBuilder is called which returns a TextData with the corresponding description.
 275          *
 276          * @param children the children of the feature node considered
 277          * @param parent the feature node considered
 278          * @param description the TaxonDescription element for which we want a natural language output
 279          * @param language The language in which the description has to be written
 280          * @param floor integer to keep track of the level in the tree
 281          * @return a list of TextData elements containing the part of description corresponding to the feature node considered
 282          */
 283         private List<TextData> buildBranchesDescr(List<FeatureNode> children, FeatureNode parent, TaxonDescription description, List<Language> languages, int floor) {
 284                 List<TextData> listTextData = new ArrayList<TextData>();
 285                 floor++; // counter to know the current level in the tree
 286
 287                 if (!parent.isLeaf()){ // if this node is not a leaf, continue recursively (only the leaves of a FeatureTree contain states)
 288                         levels.add(new Integer(floor)); // the level of the different nodes in the tree are kept, thus it is easier to build a structured text out of the List<TextData>
 289                         Feature feature = parent.getFeature();
 290                         TextData featureName;
 291                         if (feature!=null && feature.getLabel()!=null){ // if a node is associated to a feature
 292                                 featureName = categoricalDescriptionBuilder.buildTextDataFeature(feature, languages);
 293                                 levels.add(new Integer(-1)); // it is indicated by a '-1' after its level
 294                                 listTextData.add(featureName); // the TextData representing the name of the feature is concatenated to the list
 295                         }
 296                         else featureName = new TextData(); // else an empty TextData is created (because we keep track of the features, it is useful to inform when the upper node has no feature attached)
 297
 298                         for (Iterator<FeatureNode> ifn = children.iterator() ; ifn.hasNext() ;){
 299                                 previousTextData = featureName; // this allows to keep track of the name of the feature one level up in the tree
 300                                 FeatureNode fn = ifn.next();
 301                                 listTextData.addAll(buildBranchesDescr(fn.getChildren(),fn,description, languages, floor));
 302                         }
 303                 }
 304                 else { //once a leaf is reached
 305                         Feature feature = parent.getFeature();
 306                         if (feature!=null && (feature.isSupportsQuantitativeData() || feature.isSupportsCategoricalData())) {
 307                                 Set<DescriptionElementBase> elements = description.getElements();
 308                                 for (Iterator<DescriptionElementBase> deb = elements.iterator() ; deb.hasNext() ;){ // iterates over all the descriptions enclosed in the TaxonDescription
 309                                         DescriptionElementBase descriptionElement = deb.next();
 310                                         if (descriptionElement.getFeature().equals(feature)){ // if one matches the corresponding feature associated to this leaf
 311                                                 if (descriptionElement instanceof CategoricalData || descriptionElement instanceof QuantitativeData){
 312                                                         TextData featureTextData;
 313                                                         TextData statesTextData;
 314                                                         if (descriptionElement instanceof CategoricalData) { // if this description is a CategoricalData, generate the according TextData
 315                                                                 CategoricalData categoricalData = (CategoricalData) descriptionElement;
 316                                                                 statesTextData = categoricalDescriptionBuilder.build(categoricalData, languages);
 317                                                                 featureTextData = categoricalDescriptionBuilder.buildTextDataFeature(feature, languages);
 318                                                         }
 319                                                         else { // if this description is a QuantitativeData, generate the according TextData
 320                                                                 QuantitativeData quantitativeData = (QuantitativeData) descriptionElement;
 321                                                                 statesTextData = quantitativeDescriptionBuilder.build(quantitativeData, languages);
 322                                                                 featureTextData = quantitativeDescriptionBuilder.buildTextDataFeature(feature, languages);
 323                                                         }
 324                                                         applyNaturalLanguageDescriptionElementProcessors(featureTextData, previousTextData);
 325                                                         levels.add(new Integer(0)); // 0 indicates a feature, which is a leaf of the tree
 326                                                         listTextData.add(featureTextData);
 327                                                         levels.add(new Integer(floor)); // this represents the level of the feature and means it is followed by a TextData containing the states of the feature
 328                                                         listTextData.add(statesTextData);
 329                                                 }
 330                                         }
 331                                 }
 332                         }
 333                 }
 334                 return listTextData;
 335         }
 336
 337 }