How Conversica delivers Conversation Automation with neural networks
Conversica delivers Conversation Automation solutions powered by neural networks to help people engage in authentic two-way conversations, get meaningful answers, and enjoy a delightful customer experience.
October 27 , 2022 by Mark Jancola
At Conversica, we are delivering Conversation Automation solutions powered by the latest in natural language processing (NLP) to offer an experience where people can interact with technology on a deeper level. Our Conversation Automation engages contacts in authentic two-way conversations without strictly relying on pre-programmed exchanges and dialogue flows. Conversica’s Revenue Digital Assistants™ (RDAs) are deployed across multiple digital communication channels and in multiple languages to drive towards the next best action.
This means our RDAs act like a real person and respond with more intuitive answers, determine intent, and tailor responses based on a wealth of data built into the solution. We call this Conversation Automation.
Conversation Automation offers better customer experiences on a larger scale and autonomously identifies and boosts revenue opportunities.
A key component of Conversation Automation is effective search that can understand all incoming messages, questions, and queries and provide accurate, relevant, and quick answers from different types of text sources. The integration of powerful search capabilities in our Conversation Automation platform for marketing, sales, and customer success teams provides a rich and personal experience that drives customer acquisition, engagement, and retention.
Traditional keyword approaches to search based on prior language configuration such as synonyms, stop words, and stemming rules are not ideally suited for a dynamic conversational technology context. Keyword search cannot capture all the different possibilities that a person can use to ask a question for which an answer may exist in a customer’s knowledge base. This can lead to friction in the customer experience or the loss of revenue opportunities if a marketing or sales professional cannot effectively find the information they are looking for.
Conversica’s team of scientists and engineers have a wealth of experience evaluating and implementing NLP systems at scale. To create a meaningful web-based conversational experience that makes Conversica stand out from other solutions, we decided to pursue a “neural first” approach to how we implement search across our Conversation Automation offerings. Our team identified the following critical objectives to successfully deploy LLM-powered search.
- Improve relevance of responses to all types of text queries and find information in any text source.
- Generate summarized answers to natural language questions in a conversational setting.
- Understand and respond to questions in multiple languages, regardless of the knowledge or answer source.
Our engineering team conducted several tests to evaluate neural search with different real world examples of query types and document samples within our web conversation automation offering.
We were impressed with the significant performance jumps in precision (+18 points), recall (+30 points), and F1 (+25 points) that we observed within a short period of time after setting up Vectara’s neural search. The engineering team was very happy with the reduced workload and the streamlined process to deploy the high priority search capabilities for the successful launch of our new conversation automation offering.
The implementation of neural search in our versatile Conversational Automation offering serves marketing teams to attract higher quality leads, sales teams to acquire more customers faster, and customer success teams to grow customer relationships and increase efficiencies. Neural search underlies the information advantage captured by these teams to deliver wins for our customers across the full lifecycle.
You can read the full case study here.