java - How to integrate two Parse Tree data structures from two different NLP Tools -

i using both stanford corenlp , fudan nlp process chinese natural language. these 2 tools both generate parse tree, i.e. stanford corenlp parse tree , fudan nlp parse tree (let me call them stree , ftree).

and need make use of stree , ftree, , costumed method on them, sharing same function signature , various in implement details. best practice define class can generate both stree , ftree.

however, these 2 kind of parse tree totally different in aspect of data structure. can think of 2 solutions:

1. define class tree generic type, passed content type of stree , ftree. plus treefactory pass these 2 kind of content types , generate relative trees. if follow way, cannot take apart 2 kind of implementation of same method.

2. define interface or abstract class tree, contains several methods. extend interface 2 different subclass corresponding stree , ftree. if follow way, children in subclasses not subclass of super.children.

   class treenode {      list<treenode> children;      //...  }; class stree extends treenode {      list<stree> children; // problem: not subclass of super.children     //...  }; class ftree extends treenode {      list<ftree> children; // problem: not subclass of super.children     //...  }; 

i wonder better choice. or can provide more adaptable solution.

following brief definition of ftree:

// declaration edu.fudan.nlp.parser.dep.dependencytree; public class dependencytree implements serializable {      // tree node content     public string word;     public string pos;     // sequence number in sentence     public int id;     private int size=1;     // dependancy relation type     public string relation;      public list<dependencytree> leftchilds;     public list<dependencytree> rightchilds;     private dependencytree parent = null;      // ... }; 

definition of stree:

// definition edu.stanford.nlp.trees.labeledscoredtreenode public class labeledscoredtreenode extends tree {      // label of parse tree.     private label label; // = null;     // score of <code>treenode</code>     private double score = double.nan;      // daughters of parse tree.     private tree[] daughtertrees; // = null;      // ... }; 

i think more fundamental question how going use syntactic analysis result 2 different parsers. 2 trees totally different in structure , not sure how going use 2 parsers result or part of speech level!

another possible option if doing annotation on each word in sentence, can combine output standford nlp tool , fudannlp tool.

technically it's possible use option suggested mike have metamodelnode interface concrete implementation classes of streemetamodelnode , ftreemetamodelnode.