Detecting semantic errors in text using ontology-based extraction rules
Semantic errors in a natural language text document are automatically detected by matching sentences in the document with stored ontology-based extraction rules that express both logically correct and logically incorrect relationships between the classes and properties of an ontology for a predefined knowledge domain of relevance to the natural language text document. The matching identifies logically correct and incorrect statements in the document which may be used for various applications such as automatic grading.
1. A computer-implemented method for automatically detecting semantic errors in a natural language text document using ontology-based extraction rules, the method comprising:
inputting to a computer the natural language text document;
storing the ontology-based extraction rules in the computer, wherein the ontology-based extraction rules express formal logical relationships between classes and properties of an ontology for a predefined knowledge domain of relevance to the natural language text document, wherein the ontology-based extraction rules comprise extraction rules that express logically correct relationships between the classes and properties of the ontology, and wherein the ontology-based extraction rules comprise extraction rules that express logically incorrect relationships between the classes and properties of the ontology;
matching by the computer sentences in the natural language text document with the ontology-based extraction rules to identify logically correct and incorrect statements in the natural language text document;
outputting from the computer a list of the logically correct and incorrect statements in the natural language text document.
2. The method of claim 1 further comprising deriving a grade for the natural language text document based in part on the logically correct and incorrect statements in the natural language text document.
3. The method of claim 1 further comprising parsing the natural language text document into the sentences.
4. The method of claim 1 wherein the matching comprises matching each sentence separately with the ontology-based extraction rules, where the matching of each sentence separately is performed sequentially or in parallel.
5. The method of claim 1 wherein the ontology-based extraction rules comprise extraction rules that express logically incomplete relationships between the classes and properties of the ontology.
6. The method of claim 1 further comprising pre-processing the natural language text document by performing spelling correction, completion of sentences, and eliminating non-informative words.
7. The method of claim 1 further comprising matching the sentences in the natural language text document using machine learning generated information extractors to identify logically correct and incorrect statements in the natural language text document.