Alan Turing in 1950 asked, “Can Machines Think?”. Chatbot Technology has impressively grown since then. Today chatbots are acknowledged as the next big thing, being recognised as the successor to the GUIs.
Basic chatbot framework
The modern chatbot technology incorporates features as:
(a) Dialogic chatbot technology: The chatbot must be able to comprehend the user. The text input is fetched to the chatbots, which is analyzed with natural language processing techniques, and finally, appropriate response is generated.
(b) Rational chatbot technology: The chatbot technology must be able to answer user questions in appropriate context, i.e. the chatbot technology is required to have access to knowledge base, and the chatbot technology should have common sense.
(c) Personified chatbot technology: The chatbot should have a personality so as to engage the user, i.e. the user should feel as if he is interacting with an alive entity.
Architecture and Design of chatbot technology
(a) Speech to text: The speech to text conversion process begins with automatic speech recognition (ASR). Speech is recorded into the microphone and the ASR filters outs the phonemes, i.e. units of speech. Then the features are extracted and finally decoding is performed.
(b) Natural Language Processing: The goal of natural language processing (NLP) is to take the unstructured output of the ASR and produce a structured representation of the text that contains spoken language understanding (SLU) or, in the case of text input, natural language understanding (NLU).
(c) Response Generation: The Response Generator (RG) receives a structured representation of the spoken text. As output, the RG generates a response to deliver to the user, which it will deliver to the Dialogue Manager (DM).
(d) Knowledge Base Creation: Chatbot technology is only as intelligent as the knowledge they have access to. Collecting training data used to train machine learning classifiers used in generative bot models or building corpuses of data used by information retrieval bots is critical to achieving human-like interactions.
(e) Dialog Management: Once the chatbot has selected response, the Dialogue Manager (DM) must choose a number of communication strategies, use language tricks to make it seem human, and deliver the message.
(f) Text to Speech: Text-to-Speech (TTS) is the final step of the process, and converts the response generated to speech, which is returned back to the user. The first step of TTS is text analysis, in which text is converted into phonemes with pitch and duration. The second step is waveform synthesis, in which segments of recorded speech corresponding to each phoneme are concatenated to form speech.
IBM Watson Case study
IBM’s Watson gained popular recognition for its Jeopardy! win. The architecture behind
Watson is a Question-Answer chatbot technology.
Working of Watson is as:
(1) Watson runs natural language processing techniques to classify the problem.
(2) Based on its understanding of the problem, Watson searches its external knowledge bases for an answer, and generates all possible content that could contain the answer.
(3) the system collects evidence on each hypothesis, and applies a ranking algorithm to determine the most likely response.
Applications of chatbot technology
(1) Virtual Personal Assistants (VPAs): Apple Siri, Google Assistant, Microsoft Cortana, Amazon Alexa.
(2) Consumer Domain-Specific Bots: Transportation bots, Dating, Medical bots, Fitness bots, Weather bots.
End of part 5