Fernando Gomez

Ph.D., Computer Science, Ohio State University.
M.A., R. Linguistics, Ohio State University.
"Licenciado," Philosophy, University of Valencia, Spain.

Professor of Computer Science

Director, Artificial Intelligence Laboratory

Fernando Gomez's research covers a range of issues in natural language understanding including parsing, semantic interpretation, knowledge acquisition, knowledge representation and problem-solving.

These areas of research are combined in SNOWY, a project that has been underway for over ten years now and which is being used as a test bed for the ideas in these areas. SNOWY is presently reading articles on animals and people randomly selected from the World Book encyclopedia WorldBook and acquiring knowledge from them, and answering questions about the knowledge it has acquired.

Semantic Interpretation

The main effort in semantic interpretation is directed towards achieving it on a large scale as required in the understanding of encyclopedic texts. To that aim, we have integrated WordNet lexical knowledge base into SNOWY. Predicates, or verbal concepts, have been defined for WordNet verb classes, which have been reorganized very considerably following the criteria imposed by the interpretation algorithm. The predicates have been linked to the Wordnet ontology for nouns, which has also undergone some reorganization and redefinition to conform with the entries in the thematic roles of the predicates. As of this writing (August, 29, 00) we have defined predicates for 80% of WordNet verb forms. A semantic interpretation algorithm has been designed and implemented. The algorithm is driven by the definition of the predicates, and presents a solution for the following interpretation problems: determination of the meaning of the verb, identification of thematic roles and adjuncts, and attachments of prepositional phrases. An interesting aspect of the algorithm is that the solution of all these problems is interdependent and is based on the declarative representation of knowledge in the predicates. The algorithm has become an essential tool for testing the correctness of the newly defined predicates.

Ontology

The massive construction of predicates, whose thematic roles have been linked to the WordNet ontology for nouns, has permitted us to look closely into which ontology is needed for semantic interpretation. Moreover, because the correctness of the entries in the predicates is determined by the semantic interpretation algorithm, we have a formal mechanism to decide on a) which primitive categories should be part of the ontology, b) the structure of these categories, and c) the criteria used for deciding to which ontological category a given concept must belong. These decisions become crucial, otherwise the interpretation algorithm may attach a preposition incorrectly, provide a wrong thematic role for a syntactic relation, or fail to identify the meaning of a verb.

Knowledge Acquisition

Recently, we have applied the newly designed semantic interpreter to the acquisition of knowledge from texts. There could be little doubt that a knowledge acquisition component should be based on the output of a semantic interpreter. The more general the semantic interpreter the easier it should be to build different knowledge extraction tasks for different domains. Because the knowledge acquisition component is grounded on the semantic interpreter, uses the predicates of the semantic interpreter to draw inference and shares the same ontology, the construction of different knowledge acquisition tasks reduces to building some inferences in the predicates used by the semantic interpreter. Incompatibilities between ontologies used by diverse components of the system do not exist. Furthermore, the knowledge acquisition designer does not have to be concerned with defining ontological categories, because these have been built for him/her in the semantic interpreter. We have built several applications to acquire biographic knowledge from The World Book. These techniques exceed in generality and extendibility those described in our previous work, which does not use the new semantic interpreter.

Problem-Solving

We are studying the relation between problem-solving and comprehension. We have gathered some evidence showing that word problems have a deep structure, which needs to be revealed if these problems are to be solved in a principled manner. We have also indicated that the comprehension of a word problem is not a frame-based process, but rather is a constructive process by means of which the deep structure is built from the sentences describing the problem, in a way similar to the construction of a house from the building materials. This constructive process is done by means of integration rules, which build the deep structure on the basis of the semantics of the logical forms of the sentences and on the deep structure under construction. These ideas are being applied to the construction of databases from natural language input.

Some Recent Selected Publications

There are three kinds of papers. There are those that you can click on them and download. In order to get the papers marked with a cube, you need to click on the cube and enter your e-mail address below where indicated. The postscript file of the selected papers will be e-mailed to you immediately. You need to cut the headings of the message prior to ghostviewing it. If you are interested in any of the other papers, I can send you a copy if you e-mail me a request.

Gomez, F. (2008).

The acquisition of common sense knowledge by being told: an application of NLP to itself PDF Format. To appear in NLDB 2008 NLDB 08 Copyright Springer-Verlag

File nounsenses-disambiguation text Format. Technical Report, CS-TECH-REPORT-08-02 University of Central Florida, Nov 2008. (test data) This file contains a set of sentences that were part of the test data for the algorithms described in the paper ``The acquisition of common sense knowledge by being told: an application of NLP to Itself.''

Gomez, F. (2007).

Automatic Semantic Annotation of Texts PDF Format. University of Tubingen In the GLDV-2007 Workshop on Lexical-Semantics and Ontological Resources. pp. 59-66. C. Kunze. L. Lemnitzer and R. Osswald (eds). This paper contains a description of the automatic annotation of texts.

Gomez, F. (2007).

Semantic Interpretation and the WordNet Upper-Level Ontology PDF Format. With permission from Journal of Intelligent Systems, Freund Publishing House Ltd. Copyright Freund Publishing House Ltd. This paper contains the latest version of the WordNet ontology as it relates to semantic interpretation.

Gomez, F and Segami, C. (2007).

Automatic Semantic Annotation of Texts (Fernando Gomez). Technical Report, UCF-CS-TR-03-04, University of Central Florida, Nov 20, 2003. This zip file contains the semantic interpretation of 500 sentences divided into 10 files containing each 50 sentences. The paper ``Automatic Semantic Annotation of Texts'' (listed above) describes briefly the semantic interpretation of these sentences. These sentences were selected for testing the verb predicates that we have been defining for WordNet 1.6 WordNet verb classes, and the algorithm that uses them. Semantic interpretation is provided for determining the meaning of the verb, or, predicate, its semantic roles, adjuncts, the attachment of prepositional phrases, and also for a limited number of deverbal nominalizations. The senses of nouns are also resolved, but not complex nominals. The introduction section should be more than sufficient to understand the output of the semantic interpreter. However, some sections explaining the predicates and semantic roles are also included. Most (about 96%) of these sentences are taken from the World Book Encyclopedia, WorldBook, but they may have undergone editing before being inputed to the parser and interpreter. This file can be downloaded only for research purposes.

Automatic Semantic Annotation of Texts II (Fernando Gomez). Technical Report, UCF-CS-TR-04-01, University of Central Florida, March 22, 2004. This zip file contains the semantic interpretation of 500 additional sentences. There are 10 files in this zip file each containing 50 sentences. This file was completed on March 20th 2004. Please see the file annotation.ps mentioned in the previous publication for an explanation. Most (about 95%) of these sentences are also taken from the World Book Encyclopedia, WorldBook, and may have undergone editing before being inputed to the parser and interpreter. This file can be downloaded only for research purposes.


Gomez, F. (2004).: Building Verb Predicates: A computational View PDF Format.
Proceedings of the 42nd Meeting of the Association for Computational Linguistics, ACL-2004, Barcelona, Spain, 2004. /Copyright Association for Computational Linguistics. All rights reserved.


Gomez, F. (2001).: An Algorithm for Aspects of Semantic Interpretation Using an Enhanced WordNet PDF Format.
Proceedings of the 2nd Meeting of the North American Chapter of the Association for Computational Linguistics, NAACL-2001, CMU, Pittsburgh, 2001. /Copyright Association for Computational Linguistics. All rights reserved.


Gomez, F. (2000).: Why Base the Knowledge Representation Language on Natural Language? PDF Format
Journal of Intelligent Systems, v.10, no.2, 2000. Copyright Journal of Intelligent Systems. All rights reserved.

R. Hull and F. Gomez, (1999).

Automatic Acquisition of Biographic Knowledge from EncyclopedicTexts
Expert Systems with Applications, v.16, pp 261-270 .


Gomez, F. (1998): A Representation of Complex Events and Processes for the Acquisition of Knowledge from Texts. Knowledge-Based Systems,v.10, no.4, Jan 1998, pp. 237-251. Copyright Elsevier Science. All rights reserved.

Gomez, F. Segami, C. and Delaune, C. (1998)

A System for the Semi-Automatic Generation of E-R Models from Natural Language Specifications.

Knowledge and Data Engineering . Elsevier Science.

Gomez, F. (1998).: Linking WordNet Verb Classes to Semantic Interpretation. PDF Format. Proceedings of the COLING-ACL Workshop on the Usage of WordNet on NLP Systems. Universite de Montreal, Quebec, Canada, Agust, 16 98 pp. 58-64.

Gomez, F., C. Segami, and R. Hull. (1997).: Determining Prepositional Attachment, Prepositional Meaning, Verb Meaning and Thematic Roles. PDF Format. Computational Intelligence, 13(1), February 1997, pp. 1-31. Copyright Computational Intelligence. All rights reserved.

Hull, R. and F. Gomez (1996).

Semantic Interpretation of Nominalizations. PDF Format. Proceedings of the Thirteenth National Conference on Artificial Intelligence, Portland, Oregon, August, 1996, pp. 1062-8.
Copyright 1996, American Association for Artificial Intelligence. All rights reserved.

Gomez, F., R. Hull, and C. Segami. (1994).: Acquiring Knowledge from Encyclopedic Texts. PDF Format. Proceedings of the ACL 4th Conference on Applied Natural Language Processing, Stuttgart, Germany, 84-90. Also in the IEEE Int'l Journal of Artificial Intelligence Tools , 4(3), 1995, pp. 349-67.

Gomez, F. (1996).: Acquiring Intersentential Explanatory Connections in Expository Texts.PDF Format. Int'l Journal of Human-Computer Studies, vol.44, issue 1, Jan. 1996, pp. 19-44. Copyright 1996, Academic Press. All rights reserved.

Gomez, F. (1995).: Learning Word Syntactic Subcategorizations Interactively. Knowledge Based Systems, 8(1), 190-200.

Gomez, F.: On the Deep Structures of Word Problems and Their Construction.PDF Format. In Chapter 3 of AI and Automation, Bourbaki, ed., 28-47. Copyright 1998. World Scientific. All rights reserved.

Gomez, F. and C. Segami. (1991).: Classification-Based Reasoning. IEEE Transactions on Systems, Man, and Cybernetics, 21(3), 644-659.

Gomez, F. and C. Segami. (1989).: The recognition and classification of concepts in understanding scientific texts. Journal of Experimental and Theoretical Artificial Intelligence, 1, 51-77.

Enter your e-mail address: