We’re all familiar with ‘virtual assistant’ chats that appears in most of the commercial sites.
You can see examples below:
The common denominator is that all of them are based on pre-defined set of automated rules that are trying to guide you through predefined path in order to answer you or promote some sales activities.
Those bots can’t really understand your questions and solve the issue but redirect it to relevant agent or line of business instead of IVR menu.
The root cause for this is the limitation of techniques such NLP and NLU from understanding the real intent and the flow and context of the conversation.
NLP – stands for Natural Language Processing. Started around 1950 and until today there is no much progress with that.
The main challenge is to bring the right answer on a specific domain with limited knowledge base.
NLU – Stands for Natural Language Understanding and the difference between NLP and NLU is mainly that NLU is sub domain which aims to deal with the understanding part of the language and not in the processing itself.
Many companies and software techniques aims to solve this issue by tackling part of the problem such as: Identify part of speech, Speech recognition and Linguistic processing.
Although computing power grow exponentially the NL domain didn’t have such a break and the growth is almost linear.
In order to improve the NL results many companies are ‘helping’ the NL technique by adding some heuristics such as knowledge base and pre-defined definitions and limitations. This can help the vendor on specific domains and as first answer for a customer but not more than that.
In order to make a break through there are two major factors that need to be break:
1. Unlimited unique data with various domains and with concrete answers
2. Some feedback from human activity that can help with A/B testing like approach
Will give some examples and more details in one of my upcoming posts.
Have great week,