Document-classification system, method and software
Every year, professional classifiers at West Group manually classify over 350,000 headnotes, or abstracts of judicial opinions, across approximately 82,000 separate classes of the Key Number System. Although most headnotes are classified from the memory of the classifiers, a significant number are difficult and thus costly to classify manually. Accordingly, the inventors devised systems, methods, and software that facilitate manual classification of headnotes and documents generally and hard-to-classify headnotes particularly. One exemplary system provides a graphical user interface that concurrently displays an unclassified headnote, a ranked list of one or more candidate classes, a candidate class in combination with adjacent classes of the classification system, and at least one classified headnote that is associated with one of the candidate classes.
1. A method comprising:
providing a classification scheme including two or more classes, with each class associated with one or more classified document headnotes;
summarizing a legal opinion to define at least one particular document headnote;
automatically generating a ranked list of two or more classes of the classification scheme, with each listed class associated with one or more classified document headnotes which are similar to the one particular document headnote; and
classifying the one particular document headnote based on the ranked list of classes:
wherein automatically generating a list of one or more of the classes comprises:
defining one or more natural-language queries based on the particular document summary;
performing one or more searches of the classified document summaries based on one or more of the queries, with one or more of the searches yielding two or more found document summaries;
ranking the one or more found document summaries based on relative similarity to the particular document summary to define one or more ranked document summaries; and
generating the list based on one or more of the ranked document summaries.
2. The method of claim 1 , wherein summarizing a particular document comprises manually summarizing the particular document.
3. The method of claim 1 , wherein classifying the one particular document summary based on the list of classes, comprises manually selecting one or more of the classes using a graphical user interface, wherein the graphical user interface includes:
means for displaying at least a portion of the one particular document summary;
means for displaying information identifying one or more of the list classes as candidate classes for the one particular document summary;
means for displaying a logical relationship between at least one of the candidate classes and another class in the classification scheme; and
means for displaying at least one classified document associated with one of the candidate classes;
wherein the portion of one of the documents, the information identifying one or; more of the classes, the logical relationship, and the one classified document associated with one of the candidate classes are displayed concurrently within the graphical user interface.
4. The method of claim 1 , wherein classifying the one particular document summary based on the list of classes, comprises:
displaying at least a portion of the one particular document summary;
displaying information identifying one or more of the ranked list of classes as candidate classes for the one particular document summary, the information displayed concurrently with the portion of the one particular document summary;
displaying a logical relationship between at least one of the candidate classes and another class in the classification scheme, the logical relationship displayed concurrent with the information; and
displaying at least a portion of one classified document associated with one of the candidate classes, the portion of the one classified document displayed concurrent with the logical relationship.