Combining Search, Chatbots, and Question Answering to Deliver Holistic Enterprise Knowledge
How do we give users and stakeholders access to the wealth of knowledge scattered in numerous enterprise repositories, such as technical documents, customer information, employee information, vendor relationships, locations, procedures, business reports, project documents, etc.?
Without holistic knowledge and understanding of the organization, it is difficult to innovate and improve business results. Imagine the various problems enterprises can solve when disparate data sources are pooled together to support use cases like task fulfillment, account updates, content search, question/answer, troubleshooting, determining applicability, and decision making.
Thus, helping users find information and derive insights easily will ultimately add value across all industries and business functions, from corporate, customer support, insurance, to manufacturing, government services, financial services, and many others. In today’s digital workplace, Natural Language Processing (NLP) solutions like chatbots and Question Answering (QA) systems, coupled with search, have emerged as innovative solutions to address this demand.
Discovering Enterprise Knowledge: Going Beyond Manuals and Text Documents to Knowledge Graphs
Using AI and NLP techniques, we can extract knowledge from various sources and bring them together to create knowledge graphs of insightful relationships between enterprise data points. In this perspective, data lakes are particularly useful as repositories for storing and handling all types of structured and unstructured data. Finally, combining all knowledge graphs from all data sources in a single view leads to a holistic enterprise knowledge system that can help users find answers and solve problems effectively.
It's worth noting that not all questions can be answered simply. Complex questions, like contract, policies, procedures, or applicability questions, remain challenging for today’s chatbots and QA systems to answer. But with growing developments in this space, we expect that policy and procedure documents will one day be replaced by knowledge graphs that can enable answers to complex questions directly. There will be no paragraphs of text, manuals, wiki pages, or reference documentation. There will only be the knowledge base which is rendered, on-demand, in any way required. Read more about this observation by our Innovation Lead.
Ultimately, knowledge graphs power the enterprise chatbots and QA systems through which users can interact with knowledge sources. The diagram below illustrates this workflow.
Chatbots, Question Answering (QA), and Semantic Search – How Do They Work Together?
Chatbots handle deep dialogs and specific domains while QA systems handle broad domains of knowledge. But chatbots and QA systems can be complementary depending on the interface where the user starts to look for information (a search box first or a chatbot interface first). Semantic search and search engines can be used as fallbacks. Based on costs, the depth of knowledge, and other potential criteria required by your organization, an assessment can help you select one or a combination of these solutions.
The architecture diagram below showcases how chatbots, QA, and search interact with the business system and each other to create a fully-integrated, intelligent knowledge system within the enterprise.
- Chatbots provide deep dialogs to help perform specific tasks.
- QA systems interface with business systems and knowledge graphs to answer questions.
- Semantic search understands what you are searching for and returns highly-targeted documents or records.
- Search engines find documents that best match the list of words and tokens from the user.
Using AI and NLP techniques to merge content sources and create knowledge graphs, we can then leverage chatbots, QA, and search to deliver holistic knowledge and understanding of the enterprise to the user. This is where we think the industry is heading to.