In Description Logics (DL) knowledge bases (KBs), i.e. information represented through ontological formats and exploited through a concept reasoner, the Open World Assumption (OWA) makes that conclussions not only depend on information contained in the knowledge base, but also on unknown/missing and plausible information. This is, using information contained on it we can conclude that certain statement (query) holds, but we cannot simply deny it if there is no enough evidence. This is because the information that might draw the conclusion false might not be contained in the KB.

The concrete example discussed in the forum was: there are defined two partitions for a class on which the first one identifies individuals with at least one value for the property P (P min 1), meanwhile the second one identifies those individuals that do not have any value for P (P exactly 0). Both are cardinality constraints on property P.

Meanwhile reasoners can easily classify individuals in the first partition, they are not capable of classifying an individual in the second partition when an individual lacks of relations through P. The reason is that the reasoner assumes that that information might be missing in the KB. Then, how can be classified an individual in the second partition? The answer is simple, assert it directly in the KB:

I = P exactly 0 && …

This isn’t intuitive for people that has been educated on relational databases, but is quit clear for Ontology experts and concept reasoners can easily handle this kind of assertions.

An example

In order to clarify the matter you can see the following example, where people is divided in two partitions: those with children (Parent) and those without any children (NoParent).  There are two persons, Joseph and Jesus, such that the second is son of the first and he has no children. In the following Protege snapshots you can see how are classified both of them. Note the declaration of the second individual: Jesus = Person && hasChild exactly 0. Inferences are shown with a yellow background and dashed lines.

Person with children (Parent)

Person without children (NoParent)

Other alternatives have been proposed for solving this problem. For instance, it has been introduced an epistemic operator [2] that allows to specify: given that all about property P is known. In this way you specify in the NoParent concept that information regarding children is completely known in the knowledge base. Nevertheless, as far as I know, concept reasoners do not support this operator yet.

You can find more information about OWA and hor to help your concept reasoner to decide on more complex partitions in [3]. The example is a little unclear as long as authors define Sex as a class and Male and Female as its subclasses. If instead of Sex they would use Animal, the example would make more sense. This is, an animal is male or female, but not both at once.

(In spanish)

References

Para el caso de Bases de Datos DL, esto es, modelos ontológicos explotados con un razonador de conceptos de por medio, la suposición del mundo abierto (Open World Assumption, OWA) establece que las conclusiones a las que se pueden llegar dependen no sólo de la información contenida en la BD-DL sino también de la que se desconoce. Es decir, con base en los hechos contenidos en ella podemos concluir que cierta afirmación (consulta) es cierta, pero no podemos simplemente negarla si no hay evidencia suficiente para ello. Esto se debe a que posiblemente, la información que haría falsa la afirmación podría no estar contenida en la BD.

El ejemplo concreto que se discutía en el foro era el siguiente: se definían dos particiones para una clase en las cuales la primera identificaba a individuos con al menos un valor asignado en la propiedad P (P min 1), mientras que la segunda identificaba a aquellos individuos que no tenían ningún valor asignado en dicha propiedad (P exactly 0). Ambas restricciones son de cardinalidad en la propiedad P.

Mientras los razonadores clasificaban correctamente a los individuos de la primera partición, no era posible clasificar a un individuo en la segunda a partir del hecho de que no estuviera ligado con otro recurso a través de P. La explicación es que el razonador asumía que aunque la propiedad no se había expresado explícitamente ésta aún podía ser cierta (aunque no se hubiera expresado en la BD). La solución al problema era expresar que cierto individuo I no estaba asociado a ningún otro valor o individuo a través de P, esto es: I = P exactly 0 && …

Esto es algo que no se nos ocurre en primera instancia a los que hemos trabajado bajo la suposición del mundo cerrado (OPA). Sin embargo, en una BD-DL si es factible e incluso necesario hacerlo si es que queremos que el razonador haga una correcta clasificación.

Elaboré un sencillo ejemplo en el que las personas se dividen en dos particiones: las que tienen hijos (Parent) y las que no (NoParent). Describí dos individuos, Joseph y Jesus, donde el segundo es hijo del primero y el segundo no tiene hijos. Vean en las imágenes cómo son clasificados cuando se establece explícitamente Jesus = Person && hasChild exactly 0. Las inferencias se muestran en amarillo y con líneas punteadas.

Persona con hijos (Parent)

Persona sin hijos (NoParent)

Se han explorado otras opciones como la definición de operadores epistémicos [2], que permite expresar en las consultas: dado que se conoce todo respecto a la propiedad P. Con este operador se podría expresar esta consideración en el concepto de la segunda partición y no habría necesidad de complementar la descripción de los individuos con sentencias como la de la imagen de arriba. Sin embargo, los razonadores DL actuales no lo soportan (as far as I know).

Pueden encontrar más información respecto a OWA y cómo ayudar al razonador en casos de particiones completas en [3]. La explicación está muy completa pero el ejemplo es algo confuso dado que definen Sex como una clase y como subclasses a Male y Female. Si en lugar de Sex utilizaran Animal verían que los ejemplos cobran más sentido. Esto es, un animal es macho o hembra, pero no ambos a la vez.

(En inglés)

Can we predict if a research group will have a sucessful collaboration? Based on the behavior of other groups and the publication history of  a candidate group we might be capable of detetermine it.

The problem

Having observed the formation and evolution of research groups, as well as their scientific production, we have enough information for identifying those factors that foresight a successful colaboration. One of the factors of success is the number of publications that participants produce. Among the diverse factors that intervene on the productivity of the group we are interested on identifying those that contribute to the success of the group, measured in terms of external coauthors and the number of publications.

Members of a group might eventually develop divergent reasearch lines and decide to split. New groups will have a behavior different  to the original  group one: scientific production might increase or decrease.  Based on the expectation of overall scientific production, the university must decide if it will support the formation of the new group or it will discourage the separation.

The approach

Treating a research group as a social network of coauthors, we will try to determine a correlation between centrality measures and success factors. This information will be used for predicting if a new candidate group will have a high scientific production. It can also be useful for assessing cooperation between groups.

* Suitable for masters students.

Professor’s publications are available in the Web, but how can we integrate this information into a single repository when sources are dynamic? We are dealing with information synchronization rather than data integration. The Large Knowledge Collider can help in this task.

The problem

Now we can find multiple public databases on the Web with information of publications. This information is valuable for universities as long as it provides an insightful indicator of professors and research gruops performance. Web sources are dynamic and have multiple formats, not only on the presentation format (RSS, XML, HTML) but also on the information structure (metadata).

Quantifying and qualifying this information cannot be done by simply aggregating and summing up. It is necessary to discarding duplicated publications and classifying them according to a taxonomy of publications which is associated to an evaluation schema.

The approach

We can use the Large Knowledge Collider (LarKC) for identifying suitable web sources, extract publications of our professors, combine them in a repository and assert new publications in an insitutional repository (SIIP).  The SIIP publications repository has a higly specialized structure that can be used as reference framework for distinguising and classifying found publications.

Information synchronization can be implemented through LarKC workflows. We would need components for: identifying publication databases (IDENTIFY), extracting publications (TRANSFORM) from our professors (SELECT), compare them with those stored in our repository (REASON), and distinguish new information (DECIDE).

This topic is suitable for summer interships or a masters student.

Participation of citizens in social projects is becoming a necessity in a world where governments no longer satisfy population needs. Can we provide an organizative framework for NGOs based on CAID? There is a community radio interested in the approach.

The problem

Participation in social projects is chaotic in many cases. The participation is voluntary, irregular and scarce, which may attempt with the continuity of the project. A clear definition and aligning of goals around projects might help to schedule and plan activities, hence optimizing the contribution of participants.

The approach

Modeling processes and priorize goals. Measure differences on the level of participation and goal achievements with and without goal alignment. Test robustness of process specifications when conditions changes; for instance, an expert leaves the project or a new need arises. Optimize the participation of members through a capabilities inventory. Measure the incorporation of new members to the group.

The problem

Currently, information resides mainly in structured or semi-structured formats like databases, excel files and recently XMLs. Information integration is usually made through batch processes on which information from multiple sources is translated into neutral schemas on regular basis. Programming this translation requires intensive application of knowledge by experts on original schemas and the formulation of data cleaning routines that has been systematized through ETLs. Additionally, this process is excessively expensive for small files with few data with unusual formats.

The approach

The same idea of annotating Bayesian networks with semantic descriptors can be applied to structured formats. Dependencies between semantic variables will determine precedence relationships among variables (columns of data). The causal belief propagation algorithm is then used for recovering individuals and asserting relationships and new constants. The information would be integrated into a RDF graph.

Advantages

• Annotations can be reused in similar files.
• The mechanism is time efficient for small files.
• The target schema (ontology) would be defined on the fly.

Abstract

We describe a knowledge and information network approach for managing research assets in a knowledge-oriented organization using a multiagent system. The purpose of the approach is to provide decision makers a knowledge management framework to assist them in generating benefits from the knowledge assets developed by the research groups in a knowledge institution. The research assets under consideration are of three types: research products, intellectual capital and research programs. Research products include publications of various kinds, such as journal articles, research-based books, patents, technology licensing, trademarks, incubation of technology-based startup companies and others. Intellectual capital consists of the talent and expertise of research staff, such as professors, students and researchers. Research programs include academic curricula, research units, research infrastructure, and business incubators. The approach is supported by an intelligent platform that contains an information system, a multiagent-based system, a knowledge management system and a knowledge-information interpreter that coordinate repositories, domain ontologies and databases for handling the various types of research assets. The system provides a means for distributing existing research assets both within the organization and beyond, a variety of research reports, online consultations, a search engine, web services and data mining facilities for knowledge extraction. The knowledge and information generated by the system guide managers in defining strategies on competitiveness, for example, rankings, benchmarking, intellectual property and incubation of technology-based spin offs. The system has been operational at Tecnológico de Monterrey since August 2004, and has proved useful for both acquiring knowledge of research capabilities and for stimulating entrepreneurial science initiatives.

Cite as

Francisco J. Cantú and Héctor G. Ceballos. A multiagent knowledge and information network approach for managing research assets. Expert Systems with Applications. Volume 37, Issue 7. July 2010. pp 5272-5284.

Abstract

The Electronic Institutions (EIs) framework is designed forregulating interactions among heterogeneous agents in opensystems [1]. In EIs, agent interactions are speech acts whoseexchange is organized as conversation protocols called scenes.Agents can participate simultaneously in multiple scenesplaying a single role in each one of them. However, at somepoint, the execution of a given scene may require the pres-ence of an agent playing a particular role. When such anagent is missing, a deadlock may ensue unless the institu-tion or the agents themselves can invoke the participationof an agent to play the missing role. Such functionality isnot provided in the current EI framework. We propose anextension of the framework that addresses that problem in ageneric way: the provision of an institutional agent in chargeof instantiating new agents and dispatching them to scenesthrough a participation request protocol. In this paper wemake the proposal precise and illustrate it with a use case.

Cite as

Hector G. Ceballos, Pablo Noriega and Francisco J. Cantu. Requesting agent participation in Electronic Institutions (Extended Abstract). Ninth International Conference On Autonomous Agents And Multiagent Systems (AAMAS 2010).  pp 1375-1376. Canada. ISBN:978-0-9826571-0-2.

Abstract

In Electronic Institutions [1], agents may be prevented fromachieving their goals if other participants are not present ina given scene. In order to overcome this situation we proposethe addition of an institutional agent in charge of dispatch-ing agents to scenes through a participation request proto-col. We further propose to endow this agent with the ca-pability of instantiating new agents, thus providing groundsfor a self-optimization of the system. Advantages of ourproposal are illustrated with the implementation of an in-formation auditing process.

Cite as

Hector G. Ceballos, Pablo Noriega and Francisco J. Cantu. Dispatching Agents in Electronic Institutions. First Workshop On Infrastructures And Tools For Multiagent Systems (ITMAS 2010), pp 26-33. Canada.

Abstract

In this paper we present a Causal Artificial Intelligence Design (CAID) theory that borrows notions from Classical philosophy for modeling intelligent agents. Principles introduced by this theory are used for extending a goal-driven BDI architecture and implementing what we call Causal Agent. This architecture incorporates causal formalisms like Pearl’s Do calculus and C+ which are adapted to Semantic Web knowledge representations. Our approach includes an ontological agent description that enables and justifies the instantiation of agents as part of a plan. An experimental prototype used for validating experimentally our approach is commented.

Cite as

Héctor G. Ceballos and Francisco Cantú. Towards a Causal Framework for Intelligent Agents Development. In proceedings of the 8th Mexican International Conference On Artificial Intelligence (MICAI). México. pp 67-72. November 2009.