Due to the number of spam that every Internet user receive today, uses of an automated filter, is highly recommended.

In order to develop automated classification solution, I have collected, in the pas month, many email messages including spams and regular one.

Following this stage, two solutions were developed: The first one is a plugin for the Spamihilator program. The second one is a full solution in order to remove spams.

Design and development of a Spamihilator plugin (Domain Filter)

One rule which emerge from the mail corpus study is that most of spams messages contains links to external web sites. Study of the domain names linked by the spam messages to show that many times they look unusual.

Heuristics have been defined according to preceding remarks.

The Domain Filter plugin, first extract all domain names from the email message and use the defined heuristics to identify spam message. The filter is available for free from the Spamihilator web site. A document about Spamihilator configuration is available in the Technology section.

Tools used:
- C++

Design and development of a multi-platform antispam solution

This solution is based upon a client-server architecture and uses of specialized agents. Emails messages are collected by dedicated agents (working on their own thread). Spam detection is done by specialized agent. Regular email message are dropped in a local mailbox.

Mail client connect to the local mailbox, through a local POP3 server, in order to read mail messages.

The selected architecture allows processing a high number of messages without any problem and some tests collecting several hundred messages have been done without any problem.

Spam identification is performed using a multi-criteria approach. Most of the filter used implements a supervised learning capability.

The program is actually in beta test phase. A detailed description will be published soon.

Tool used:
- Java

Candidate information's are automatically extracted from their resume:
 - personal data (last name, first name, age and/or date of birth, situation),
 - address, phone number, email address,
 - education,
 - work experience,
 - computing skills and foreign languages,
 - miscellaneous information (driving license, ...).

A more detailed description of this project is available in thetechnology section.

Tools used:
 - C++,
 - C#,
 - Java,
 - Python,
 - XML Spy,
 - BayesiaLab (Bayesia),
 - Neurosolutions (NeuroDimension),
 - VisualText (Text Analysis International),
 - Machinese Phrase Tagger (Connexor).

Design and development of an application dedicated to on-line resume management.
- Creation and modification of the resume using specialized forms,
- Possibility to upload an electronic resume (Microsoft Word file),
- Creation and modification of accompanying letter,
- Resumes are exported to HTML, PDF and RTF formats (the resumes are first generated in XML and converted using XSLT and XSL:FO).

Tools used:
 - Java,
 - Javascript,
 - SQL Server,
 - XML Spy.

Design and development of an application for determining the likelihood that two people are the same person.
   People matching is made upon personal data (last name, first name, address, phone number, email address, date of birth).
   Method is based on fuzzy logic, and is not sensitive to difference due to the fill-in stage (typos, accentuated words, ...).
Tools used:
 - Java,
 - fuzzyTech (fuzzy logic).

   Specialized agents have been built in order to permit real time integration of information flow into a corporate web site (TF1).
   Agents were in charge of collecting, analysing and storing into a database all flow's information.

Tools used:
 - ColdFusion,
 - Oracle 8.

Spare parts information's are stored into an AS/400 database. Spare parts catalogs publishing was done by hand and constitute tedious work.
   A fully automated solution was developed. Due to the short publishing time, on demand catalogs can be done.
   Page layout is done automatically using an export file from the AS/400. Application generate a MML file (Maker Markup Language) which is automatically rendered using Adobe FrameMaker software.

Tools used:
 - FrameMaker,
 - Prograph.