This talk presents an overview of ontology creation and development in SUMO, including how we envisage using it to help with reasoning about Portuguese texts.

The Suggested Upper Merged Ontology will be introduced (no previous knowledge assumed) and Sigma (the ontology environment) associated with it will be demonstrated. Issues of the capabilities and tradeoffs in first order logic inference are briefly explored. The SUMO mappings to the WordNet lexicon are also discussed.

Hopefully some hands-on axiom writing will be done together, bring your laptops!

Logical systems conceived for providing semantics and logical forms for sentences of English abound. From Montague's original Higher-order Intensional Logic in the seventies, to Situation Theory and Discourse Representation Theory as well as the several frame languages (e.g. KL-ONE) and their descendents, the Description Logics in the 80's and 90's, to vanilla versions of First-order logic (FOL) with bells and whistles more recently, the field is rife with possibilities and issues.

I want to describe one more such language, the product of several years of development of the NLP-based knowledge representation system Bridge, at PARC. While the design of the language was historically tied-up to the development of the software system, I believe that the language and its inferential system are of independent interest. In previous publications, we have been calling this logic language TIL, for Textual Inference Logic. TIL can be considered one of several systems associated with Natural Logic.

Within the traditions of Natural logic, TIL distinguishes itself by the unorthodox treatment of quantification in terms of instantiation of concepts within contexts. I want to describe this mechanism of quantification, necessitated by what we consider a better modelling of negation and other intensional phenomena, ubiquitous in natural language. Further we want to map and relate TIL's expressive power to that of more traditional systems, as for example, the ones described by Moss.

We start from the analysis of Turing's work in On computable numbers, with an application to the Entscheidungsproblem raising evidences about how he proceeded to build the definition of computability. Taking advantage of his stepwise construction that starts from the materialities to achieve a satisfactory level of abstraction, we show how his way of doing mathematic fits an approach of knowledge construction where there is not definite separation between materia and form, and thus, the world and the language are not closed spheres. In the same line of reasoning, the abstract and the concrete, the deduction and the induction, the technical and the social, the objective and the subjective are unthinkable as pure entities. Considering controversies and discussions from the mid-nineteenth century to nowadays, we verify a social component that permanently takes part in what is usually considered "technical content" or “objectivity” undermining, thus, the axis of authority of mathematics, logic and computing. Under this view new possibilities for knowledge construction are acknowledged, allowing maths that are done and lived outside the major centers.

We will start presenting the problem of Truth Maintainance in Database systems. Roughly speaking, this problem addresses the question of how to maintain the integrity constraints during the lifetime of a database. In some recent papers a discussion concerning the use of inconsistency-tolerant methods for obtaining partial integrity constraints satisfaction shows up. Partial integrity is obtained by means of flexible repairs, that is, integrity repair updates on the database that must follow a failed integrity constraint checking in order to try to recover the truth state of the database.

Absolute consistency is out of question due its intractability. On the other hand, naive inconsistency-tolerant repairs can be data-destructive. In order to have a rational flexible repair strategy, one needs criteria. These criteria can be expressed in terms of metrics.

We can consider the extension of what was discussed above to non-SQL DBs. A natural extension is to consider associative based Knowledge Representation bases, such as those based on RDF.