AI Based Natural Language Processing (NLP)


We Have Advanced Technologies That Can Be Readily Used In The Field

Although natural language processing has been studied academically for many years, solutions that work well and fast enough to be used in real life applications are few or still lacking in many areas. Both with the technologies we have developed and with the experience we have gained so far, we provide our customers with effective and fully-functional solutions on Natural Language Processing Based on Artificial Intelligence. The highlights of our technology, experience and capabilities in this field are:

  • The largest labelled and unlabeled Turkish language data set (Over 2 billion tweets, blogs, news of which over 250 thousand are labelled)
  • End to end text processing system (SaaS)
  • Text summarization, finding relationships between texts, text cleaning and correction, text segmentation, morphology detection, subject detection
  • Language support for Turkish, Arabic (5 dialect), Persian, English and Kurdish (2 dialect)
  • Named Entity Recognition on 24 classes
  • Flexibility for adding more named entity classes
  • Sentiment Analysis on 14 classes (Most sentiment analysis tools only support positive, negative and neutral)

End to end text processing system (SaaS)

We develop end-to-end solutions on Natural Language/Text Processing. We both integrate our solutions to our products and also provide them as a service. Our solutions include:

  • Text/Document Summarization Service
  • Text Relationship Extraction Service
  • Text Cleaning and Correction Service
  • Text Segmentation Service
  • Named Entity Recognition Service
  • Sentiment Analysis Service
  • Morphology Detection Service

Named Entity Recognition for 24 Classes

Named entity recognition aims to identify entities for predefined classes from texts such as person, institution name, organization, location, etc. Named entity recognition, which is one of the most important steps of information extraction, is used in many fields. We analyze data from various channels and by using modern Natural Language Processing techniques, we detect the content and scope of any text.

Sentiment Analysis for 14 Emotion Classes

Sentiment analysis is the task of determining whether a text contains emotions, and if yes, finding out the emotion. With this technology, people's emotions towards a particular subject can be determined. Our sentiment analysis solutions are perfect especially for social media texts. With our sentiment analysis tools, it is possible to automatically detect the emotion within given text.


Hash Based Voice Search Systems

We provide solutions on Voice ID Recognition and Voice Search In this scope, we transfer voice signals in time domain to frequency domain with Fast Fourier Transform (FFT) and generate voice spectrums. We collect samples on voice record which will be searched to create voice hashes. To get best performance on voice search, we apply optimum amount of sampling on voice records. With this, we can keep reference database size at an optimum level. Via our solutions on voice search:

  • Created voice hashes can be searched within voice database
  • Records with highest matching score or records matching above certain threshold are identified
  • Voice search can be performed for any language