Fethi Ferjani

Automatic handedness detection from off-line handwriting

In forensics, the handedness detection or the classification of writers into left or right-handed helps investigators focusing more on a certain category of suspects. However, only a few studies have been carried out in this field. Classification of handwriting into a demographic category is generally performed in two steps: feature extraction and classification. In this study, we propose a system which extract characterizing features from handwritings and use those features to perform the classification of handwritings with regards to handedness. Classification rates are reported on the QUWI dataset, reaching almost 70% for Left and right Handwriting.

Formal context coverage based on isolated labels: An efficient solution for text feature extraction

Different available data as images, texts, or database may be mapped into an equivalent or approximate binary relation. A text may be considered as a binary relation relating sentences to words, while a numerical table may be represented by a binary relation after using some scaling approach. A social network may be also represented by a formal context. The objective of this paper is to present an original approach for covering a binary relation by formal concepts based on isolated single or multiple properties, i.e., those belonging to only one concept. As a matter of fact, isolated properties are efficiently used for discriminating and labeling concepts. The latter are used for browsing in a corpora, or in a document by navigating through associated labels. By using fringe relations, the presented approach compared to those of the literature has the advantage of offering a relevant feature of a context by significant labels. Carried out experiments show the benefits of the introduced approach.

General learning approach for event extraction: Case of management change event

Starting from an ontology of a targeted financial domain corresponding to transaction, performance and management change news, relevant segments of text containing at least a domain keyword are extracted. The linguistic pattern of each segment is automatically generated to serve initially as a learning model. Each pattern is composed of named entities, keywords and articulation words. Some generic named entities like organizations, persons, locations, dates and grammatical annotations are generated by an automatic tool. During the learning step, each relevant segment is manually annotated with respect to the targeted entities (roles) structuring an event of the ontology. Information extraction is processed by associating a role with a specific entity. By alignment of generic entities to specific entities, some strings of a text are automatically annotated. An original learning approach is presented. Experiments with the management change event showed how recognition rates are improved by using different generalization tools.

Financial events detection by conceptual news categorization

In the scope of Financial Watch project, several targeted events have been required by contacted users in banking and investment domains. Financial news are classified with respect of the list of desired events. In this paper, a conceptual approach for indexing short English news in the financial domain is presented. By using a supervised original learning approach, a categorization method is proposed. Experimentation of the method on a sample of news showed that in almost all cases, occurrences of the right events in each document have been recognized, with respect to a corpus on the financial domain.

Hyper rectangular trend analysis application to Islamic rulings (fatwas)

Trend analysis is a research field with a large number of applications ranging from monitoring potential rivals to analyzing interests of a certain category of people. Applying trend analysis to the Islamic domain makes it possible to have a general idea about topics discussed by Muslims all over the world. It helps both scholars and social science researchers understanding the needs and the interest domains of each Muslim society. In this paper, we present a trend analysis method based on hyper-concepts. Hyper-concepts make it possible to decompose any corpus into non-overlapping rectangular relations and to highlight the most representative attributes or keywords. We illustrate the effectiveness of our method in identifying relevant keywords related to the Islamic context and we show how to use our method for identifying trending topics with respect to time.

Conceptual ascendant feature extraction of a financial corpus

In the scope of Financial Watch project, in order to create a feature of a corpus reflecting its semantic structure, a general algorithm of text micro structuring and browsing is proposed. The same conceptual algorithm has been extended recursively to derive the macro feature of a space of documents, in an ascendant way. Obtained browsing system based on the extracted feature proved to help for a fast convergence to the user request. The same tool is extendable for report generation from a corpus, based on a general feature. The main goal of this paper is to propose new experiences on document feature composition, in order to compare different varieties of algorithms and find out better paradigms with respect to some well defined criteria.

A multilingual financial watch alerting system

AbstractDepending on user profiles expressed by associated expected events, and conditions for raising alerts, from manual meticulous news annotation of an adequately selected corpus, an ontology domain is created. A cross-language information retrieval approach is used for automatic translation of financial documents corresponding to the particular domain corresponding to the user requirement. By this way, users may receive alerts and news expressed in their own language even if they are initially expressed in a different language. Manual annotation is used for knowledge extraction composed of general rules useful for automatic annotation of financial news arriving instantly to the system from reliable providers of information. As a first step in the loop, news are filtered, split into different sub-documents each one corresponding to a particular event and categorized. News are then mapped automatically to formatted data as instantiations of a sequence of predefined entities defined an event. By using a...