Thinking Solutions Pty Ltd

ANY Language ANY Meaning ANY Device

Multilingual Context Tracking

Project Turing provides a language independent/multilingual pattern matching engine. Here we look at how the context tracking works with different languages at once. The German and English examples are only indications of the types of patterns to be retained and disambiguated by the context-tracking engine.

These are very early days ion our 2-month long initiative, so feedback to us now is essential to ensure our R&D effort is maximised to year-end.

Expanding the context introduction

Today we explore the prototype with the first implementation tracking just the use of nouns in sentences - building up context automatically.

We will do one more video today to consider the application with multiple languages linking to the common layer of meaning.

Introducing Conversation

Today, we introduce the boiler plate screens for conversational interaction with Project Turing. Context is an essential aspect of language, and is therefore explored with our language system.

The system as of today is not operational, however we want to receive as much feedback as possible from our extended team as we develop this interface, so feel free to comment on this blog.

The next step will be to add the server capabilities to automate the tracking demonstrated here.

Beth Carey Introduces the Thinking Solutions Language System

Beth Carey, the Thinking Solutions Director of Business Development, introduces the reasons for the Language System in this short introductory video and outlines the historical perspective and market demand.

Beth is due back next week to report on other developments now leaving the lab at Thinking Solutions.

Indonesian Introduction

After lunch today we decided to implement an introduction to the Indonesian language...

This video shows the result of the effort - a few phrases matched, some conjugation, and some translation into Indonesian. It is great to be working with RRG and Patom Theory as both are bidirectional and therefore able to leverage existing patterns.

By recognising an English sentence, there is typically enough information available to generate a target language sentence, provide the target language has a sufficiently mature set of patterns as well.

Translation of Languages

Translating between languages requires a few things: an ability to understand what is in the source text, a knowledge of the meaning of the words in the target language, and a knowledge of the correct word orders (and inflections) in the target language.

This video introduces the Thinking Solutions translation capability used to do simple conjugations between English and French/German and Latin.

Introducing Latin

Latin has been taught through the ages to help students appreciate some of the origins of their language and to broaden their knowledge in general. At Thinking Solutions, we see it as a good way to see how a less rigid language works, in terms of word order. As Patom Theory considers a brain to be a machine that simply stores, matches and uses patterns, our model deals with this lack of word order.

This video introduces some of the vocabulary available on the demonstration site for Latin. The first part shows how matching identifies the key elements of simple Latin phrases. This, of course, requires the user to know enough words to enter them!

Word-Sense Disambiguation - Summary

This video summarises the problem of Word-Sense Disambiguation: the need to choose from among a number of possible meanings for a word. It is a problem resolved by the application of Patom theory, in which a pattern is identified from those learned through experience.

Coming up: more translation options between languages: relying on an accurate comprehension step followed by appropriate language generations.

Upgrade Progress - Translation based on meaning

This video introduces our recent upgrade in capability. The new system maps words to a universal meaning level. Once the meaning and its related logical structure is stored, translation becomes a simple mechanical task to generate an appropriate target language sequence. Provided a match is correctly made and the target language has equivalent meaning, a target result is created.

There are a few more posts we will be doing in the coming days as we go over the system's new capabilities.

Longer translation example

This video is meant to clarify some of the translation features we currently include in our prototype. There are a few ways to see the same result. As always, we are interested in any feedback you may have on what we are displaying.

We plan to produce, over the coming weeks, a few more videos in this area now that we have finished some work with the translation engine.