Question Answering (Part 1): Why Build Question Answering Systems?
Question Answering is a promising application of Natural Language Processing (NLP). For organizations it offers the promise to unlock the information hiding inside the document stores and make it available to the stakeholders in a timely and swift manner. The aim of this series is to introduce the readers to question answering systems and demonstrate the use of existing libraries to quickly build a question answer system using pre-trained BERT embeddings. In the first part of this series let us look at the relevance of Question Answering systems for a modern organization. We will cover a taxonomy of Question Answering techniques, along with a comparison of the different approaches in the taxonomy in the next article in this series.
Paradigm Shift in Search
Search has gone through a paradigm shift. Although AltaVista one of the early search engines introduced support for natural language queries as early as 1995, it was not a part of popular search engines such as Google. Popular search engines have moved from the traditional approach of only a keyword-based search that returns a list of documents ranked as per their relevance to the search keywords, to also providing direct answers to user’s questions in everyday language. In 2012, Google started a move towards incorporating support for providing direct answers to user’s questions in everyday language using knowledge graphs.
Use Case Scenario for Question Answering system
Let us look at a hypothetical use case which is a fairly common scenario for ground support personals across organizations. We all don’t like call drops. Our favorite mechanic Bob the Builder has been called to a telecom tower to fix call drop issues. Bob is an astute mechanic. However, he is not very experienced with this specific brand of tower. He needs answers like to solve the problem.
Bob is a tech savvy mechanic. He takes out his tablet and enters the query. But lo and behold what he gets. He gets links to documents and now he has to wade his way through terse technical documentation while on the field.
We think Bob could really use some help and this is where an intelligent search solution based on Question Answering can be immensely useful. Bob can ask his question in everyday language as if asking an expert and get the answer along with the relevant passage that contains the answer. The underlying intelligent search solution will go through the technical documentation to get the answer. The solution can be delivered over other interfaces such as voice and chat other than mobile app and web
How Question Answering System Benefit Organizations
The use case above gives an overview of the general services workflow across different organizations and domains. We have customers and agents on the field who often needs quick answers about the products, services, contract etc., and backend customer services staff which need to answer these questions. A Question Answering based intelligent search solution can be beneficial for all the stakeholders and benefit the organization by reducing the time to service requests, increasing the efficiency and thereby giving cost benefits.
In the next part of this series let us look at a taxonomy of Question Answering systems, and the comparison between the different categories.
