MPACT is a software Research and Development company dedicated to developing leading edge software.

It’s software is based on work by its founder and CEO, Dr. Pedro V. Marcal. Dr. Marcal has contributed extensively to the field of Computer Aided Design. He is well known in FEA, specializing in nonlinear materials, large displacements as well as failure by Fracture Mechanics and / or low-cycle fatigue. His theory of macro mechanics has led to practical analysis of micro lattices and fiber composites. In conjunction with his colleagues , he has clarified the conditions of element selection and convergence. He has shown the use of the hexa 27 node element provides the best answer that is also applicable to thin shells. The above topics have been imbedded in the MPACT suite of programs.

Dr. Marcal’s other interest is in Natural Language Understanding ( NLU ). NLU depends on Natural Language Processing ( NLP ) which takes a sentence and parses it in a context free manner to establish the syntax of a sentence. Subsequent to this, the sentence is parsed for meaning. This is called a semantic parse. Primarily the semantic parse of English is to resolve all the ambiguities in the sentence. This is because English is a phonetic based language ( alphabet ). Dr. Marcal first established a statistical parsing method based on Design of Experiment (DEX) that is more than two orders of magnitude more efficient than current traditional finite state search methods for sentences with about twenty words.

Next, he observed that the Asian Languages namely Chinese (simplified Chinese) and Japanese are based on ideographs which may be regarded as contracted drawings. This means that they are orthogonal to the European Languages. Chinese words ( character combinations ) are unique in meaning but there are many words that have the same meaning. Hence a translation of an English sentence into Chinese resolves all the ambiguities in the English Sentence. That is to say a translation is the most effective semantic parse. Fortunately the same DEX procedure could be applied in the semantic parse. In order to achieve a translation Dr. Marcal built a bilingual Lexical Dictionary, based on the Open Source Balanced Chinese Corpora from Lancaster University. In addition because English and Chinese grammar is only about 35% similar, it was necessary to map the results of a context-free parse into the syntax of the target translation language. This results in almost error free translation that only depends on the quality of the Lexical Dictionary. After a semantic parse it is necessary to process the result further to extract meaning, including contextual from the processed text. Dr. Marcal adopted Schank’s Conceptual Dependency as the basis of NLU. Some extensions were added to Schank’s original theory to reflect the availability of computing resources to handle large scale Corporas. ( circa 1.5 million words in each Language ). The first product resulting from NLP is a General Purpose Semantic Reader (GPSR) that forms the basis of all subsequent NLU products. Dr. Marcal developed an English to Chinese translator ( ECMAP ) and vice versa ( CEMAP ).

The first NLU product ANLAP is based on ECMAP and Conceptual Dependency. It takes a series of unstructured reports and converts each report to a structured database format, phrases are stored in a python dictionary. The structured reports are then correlated. Significant progress has been made in this area through the discovery that each sentence can be split up into a number of simpler concepts (SVO ). These simpler concepts have hyper generic format in a Conceptual Dependency and /or Wordnet sense, called hyper concepts. The hyper concepts were found to obey Zipf’s Law. The next version of ANLAP will be based on these hyper concepts.

Dr. Marcal’s interest in AI began when he co-led the team that developed the Stress Analysis Consultant (SACON ) Expert System. Rules were developed to guide the Finite Element Analyst in the use of the general purpose nonlinear program, MARC. At the time such analysts usually had a Ph D. In Applied Mechanics. Two programs were subsequently developed by Japanese Companies which claimed that Engineering Aides were able to use the MARC program.

Optimal Design, automatic mesh generation, Domain Languages, Automatic programming are fields that may generally be classified under the term Applied AI. Dr. Marcal is actively pursuing R& D in these areas. For example, he is investigating the use of Q learning and delayed rewards in Optimal multi constrained design.

The interested reader may find further information in Dr. Marcal’s publications or from the appropriate sections in this website. Page 2 gives a list of publications most of which are available through Research Gate.