KEYNOTE

Title:

Business Process Mining meets Natural Language Processing

Speaker:

Prof. Mohsen Kahani
Director of CERT and Web Techology (WTLab) Labs,
professor of Computer engineering at Ferdowsi University of Mashhad, Iran

ABSTRACT:

Business process mining has gained much popularity recently, due to its unique roles in discovery and monitoring business processes within organizations. Process mining utilizes the real logs of enterprise software to accomplish its tasks. However, these logs usually are either incomplete or insufficient to provide enough feedback. Deploying other information such as small text and description that accompany these logs can help further improve the efficiency of process mining activities. In this talk, after introducing process mining activities, the applications of natural language processing for process mining are explored.

Home Page: http://kahani.profcms.um.ac.ir/

Title:

Computational Social Semantics for Mining Actionable Insights

Speaker:

Dr. Ebrahim Bagheri
Associate Professor in the Department of Electrical and Computer Engineering at the Ryerson University and the Canada Research Chair in Software and Semantic Computing, Senior Member of the Institute of Electrical and Electronics Engineers (IEEE)

ABSTRACT:

Specificity is the level of detail at which a given term is represented. Existing approaches to estimating term specificity are primarily dependent on corpus-level frequency statistics. In this talk, I will discuss how neural embeddings can be used to define corpus-independent specificity metrics. Particularly, I will show how term specificity can be measured based on the distribution of terms in the neighborhood of the given term in an embedding space by leveraging geometric properties between embedded terms to define various classes of specificity metrics. The talk will cover how such specificity metrics can be useful in different information retrieval tasks such as classifying terms into hierarchical categories, estimating query performance, and ad hoc document ranking.

Home Page: https://www.ee.ryerson.ca/~bagheri/

Title:

A theory of anomaly detection in images

Speaker:

Dr. Jean-Michel Morel
Ecole Normale Supérieure Paris-Saclay, France

ABSTRACT:

Anomaly detection can not be formulated in a Bayesian framework: this would require to simultaneously learn a model of the anomaly, and a model of the background. (In the case where there are plenty of examples of the background and for the object to be detected, neural networks may provide a practical answer, but without explanatory power). In the case of anomalies, we often dispose of only one image as unique informer on the background, and of no example at all for the anomaly. Thus one is led to learn a background model from very few samples, and to detect an anomaly as a large deviation from this background model. I’ll show how the anomaly detection problem can be led back to the simpler problem of detecting outliers in noise. I’ll develop the proposed solution as a logical deduction of the huge literature on anomaly detection. Work carried out in collaboration with Axel Davy, Mauricio Delbracio and Thibaud Ehret.

Home Page: https://fr.wikipedia.org/wiki/Jean-Michel_Morel