Chatbots
Conversational interfaces built on natural language processing, from FAQ assistants to context-based voice systems.
What are Chatbots?
Chatbots are conversational interfaces for individual conversations with computer systems, conducted through messages.
They allow users to write or say messages — sometimes only questions — in normal, human-understandable form, and to receive answers in the same form. The core of every chatbot is natural language processing, which lets it not only understand the written or spoken message but mimic human conversation.
The variety is huge, from simple FAQ answer assistants to systems as complex as Siri or Alexa. The benefits are service around the clock, improved customer satisfaction, increased revenue and reduced costs.
Our Chatbots Projects
FAQ ChatBot (NLU)
An FAQ chatbot is a conversational interface for individual conversations with computer systems using messages. The chatbot asks predefined questions and analyses the answers from the user with RNN techniques, LSTM in particular.
Its responses can take different forms: a message, a link to an article, an internal object, an external source. It also learns from the user’s answers, and asks whether the answer given was relevant.
The usage is wide — any field that requires a conversation with agents, such as sales, services or customer support. For all the benefits of an FAQ chatbot, complete human intelligence remains out of reach so far.
CBAI CB (NLP)
A context-based artificial intelligence chatbot is a conversational interface for individual conversations with computer systems using voice commands, and only rarely messages. Apple Siri, Amazon Alexa and Google Assistant are prime examples.
It receives one or more vocal commands from the user, analyses them and responds accordingly. The algorithm is complex and works in several steps. First it translates the vocal signal to text using automatic speech recognition, eliminating noise and long pauses and dividing the input into separate parts according to the words recognised. Where a word is unknown — not found in the bag of words, the algorithm’s dictionary — it is added and remembered, so the algorithm teaches itself.
Next the algorithm classifies the text produced in the previous step, using machine learning classification techniques such as a hidden Markov model, or SVM with or without naive Bayes features. Lastly a text-to-speech unit delivers the response to the user, or reports the completion of the current command, in normal human-understandable form.
CBAI chatbots are used not only in IoT systems but as a personal assistant in everyday life.
