Dasher’s Corner

A little about me:

• Researcher on Natrual Language Processing and Machine Translation
• Conceptual Contextual Translation (COCOTA) – self learning translation paradigm that translates by building a conceptual framework by inferring contextual relationships between language constructs.
• The current framework environment is built using loosly coupled modula agents that build a conceptual model of language:
• The framework is reflexive – in the sense that agents reflect on concept generation using contextual markers within the text, paragraph, sentence or phrase – and generate a consenus of meaning.
• Agents have access to the same data repositories (previous historical and temporal concept trees as well as proxied self discovery [via the internet and corpus]) but are given different linguistic and programatic parameters (ala devils advocate).
• Agents are able to adjust their own working environment (parameters, working set and code base).
• Proxied self discovery means that requests for information go through a relatively dumb information proxy agent – which searches for information on behalf of the agent.  The search conditions and results are cached to improve performance of the system, and to examine the approaches taken by agents to data discovery.
• Technology:
• Modified SOAR (the original SOAR environment: http://sitemaker.umich.edu/soar) in C++
• .NET (a mixture of C#, VB, Python, LINQ)
• Platform agnostic Remoting framework for intra agent communication
• Stats:
• The system isn’t tied to any language – it self learns – but testing and development is done using European languages (English, Italian, German, French, Spanish)
• The current framework is limited to only 100 words – again – the system is able to learn new words and attribute meaning through Cocota (this is purely for performance reasons – I’m working with limited hardware and resources)
• The system translates (langauges from the above list) from language A -> language B -> Language C -> Langauge A with, on avg, 95% accuracy in terms of meaning at the final stage.  In my mind – this is the only true way to measure the accuracy of the system – if sense and meaning is lost in translation then you have failed.

• Freelance developer (C#, VB.NET 1.0 through to 2.0, PHP, Java, SOAR, C++[rusty], Python, etc)