How it works

CCL (Corpus Cube Linguistics) uniquely integrates 3 different approaches to machine learning: supervised, unsupervised and manual rule-based learning. By studying the language in use (corpus linguistics), the system teaches itself to understand the nuances of any language and jargon.

Based on the interaction between the human mind
and computer intelligence. 


Keyword-based search and predefined dictionaries are the most common solutions in today's market. The advantage is low user threshold, but it is well known that the accuracy is low. High analysis accuracy requires an understanding of natural language. 

Research has delivered so-called learning-based solutions that can capture more complex relationships and automatically connect single words or word pairs to a theme. However, these solutions typically require manual (thus time–consuming and expensive) tagging of large text collections in the training phase, with thousands of manually-curated theme examples per category. The more examples one tags manually, the better the accuracy, simultaneously unlocking the analysis of the language and the genre use, as an example. Each new genre and each new language requires extensive manual tagging process. There are solutions for unsupervised learning (cluster formation), but these are seldom suitable to identifying accurately fine-grained categories and are also very skills-intensive (subsequently low usage among analysts).

CCL attacks the problem in an entirely different manner, demonstrating its disruptive and innovative nature. By supplying a large amount of raw text, the language and genre can be analysed. CCL builds up a rich understanding of the language through an analysis of data, thereby pulling language knowledge straight out of raw text (based on corpus linguistics). This process works seamlessly.

The end user specifies analysis tasks simply by using illustrative words and phrase samples (typically no more than 5-15 words / phrases per theme). CCL combines these examples of language knowledge and builds a robust tailored analysis model in a fraction of a second, where the prevailing language and genre are captured, including specialist terms, typos, rearrangement, and slang. 

By analyzing data, such as structured and unstructured documents, chats and emails, customer feedback, scientific research articles, electronic patient records and even social media-channels, our CCL technology can automatically derive valuable insights and help you find what you are looking for. 

The results are presented in a dashboard, where findings are easily visualized and key performance indicators can be tracked. A management dashboard provides you with automatically updated overviews, as well as detailed information. Through the use of Artificial Intelligence, Anzyz is able to properly understand any language, in all its richness and nuances.  

CCL is based on ten years of research performed by Professor Ole-Christoffer Granmo at the University of Agder (UiA). The algorithm developed by Professor Granmo is unlike any other existing model in the world of Artificial Intelligence today. The unique linguistic quality of CCL accelerates the contextual understanding, thereby providing a tool which is far more superior than neural networks, naïve bayes, logistic regression and decision trees in benchmarks. 

Anzyz offers customized applications to fit your specific need.

© Copyright 2018. www.anzyz.com