Continuing from my previous post, Bots – The Good, The Bad and The Ugly, I would like to specifically look at chatbot or talkbots which are likely to make its presence felt in the coming days. You might have heard of Facebook chatbot platform in Messenger where one can browse internet for news, shop from inside the messenger. Large companies like Lloyds, Royal Bank of Scotland are deploying chatbot as on-line automated assistants sometimes replacing even humans in call centers. Tech giants like Microsoft, Google, Apple (Siri), and many tech start-ups are developing new techniques for consumers to interact with services using chatbot, we will be hearing plenty about them in near future.
A chatbot is a computer program designed to simulate a conversation like a human via textual method or thru’ sensing sound. A simple program scans for keywords from within the input and then uses similar sounding words and matching keywords from the database to reply. On the other hand sophisticated programs are known to run natural language processing (NLP) systems.
Chatbot may be categorized into retrieval and regenerative based models.
Retrieval based models – These systems do not generate any new text but use a database of predefined responses and analytics to deliver an appropriate reply based on the input and context. Since they use database, chances of grammatical errors are non-existent but these systems come up short when pitted against obscure cases where there are no pre-defined responses in the data bases.
Regenerative based models – These systems are not reliant on predefined responses, but are based on machine translating techniques where input is translated to an output. These systems are smarter as their responses have references to the input they are more likely to give impression of human presence at the other end.
Chatbots can also be classified on the length and type of the conversation they facilitate.
Short Text conversation – Single response to a given input, and is easier to automate.
Long conversation – Conversations with multiple turns with necessity to keep track of previous mentions like in customers support interactions, these are difficult to automate.
Open domain – The user can take the conversation anywhere without any goal or intention, conversations in social media sites like Twitter and Reddit are examples of this and can go anywhere and this makes it harder to automate.
Closed domain – The conversations happens for a very specific goal and is very limited in the setting, technical customer support or shopping assistants are very good examples of closed domain conversations.
Chatbots are majorly incorporated in dialog systems like automated on-line assistance to provide customer service during 24 /7 on the website. Other usages include deployments as smart phone apps in dating and romance bot category.
Unfortunately, chatbots find applications for malicious purpose like filling chatrooms with adverts and spam, more vicious are the ones used to entice people and solicit personal information for use in banking fraud etc.
For more on chatbots please visit: http://www.wildml.com/2016/04/deep-learning-for-chatbots-part-1-introduction/