May 2021 Summaries

Trackers are designed by Symbl.ai to recognize contextual similarities rather than exact keyword matching, making it easier for developers to build deeper and more sophisticated domain-specific intelligence at scale. The Tracker API propels Symbl.ai into the next phase of its objective in further enabling contextual comprehension in any and all verticals. Trackers support both real-time as well as asynchronous voice and video channels, and are now available in beta. They can be applied to any conversation-driven use case, such as improving customer satisfaction or monitoring upsell potential, by configuring key-value pairs with a few example keywords. The Tracker API is supported with Symbl's Async APIs and Streaming API, enabling it to be integrated with any voice, audio, or text conversation streams and formats.

Symbl.ai has partnered with Agora to accelerate the rise of more intelligent and robust real-time engagement app experiences, providing developers with a suite of comprehensive RTE products and tools to build and scale applications that create meaningful human interactions. With minimal coding required, developers can now augment their RTE applications to add live captioning, real-time coaching, compliance, content moderation, and intelligence-driven search through videos and live content. Agora's cloud platform provides access to AI capabilities such as speech recognition, conversation analytics, contextual insights, sentiments, topics, and custom domain-specific insights using custom trackers. The partnership enables developers to leverage conversation intelligence with just a few lines of code, empowering them to add AI and ML in their RTE apps without requiring upfront training or maintenance of ML models.

The conversation intelligence landscape encompasses various technologies, including text analysis, human-to-machine conversations, and human-to-human conversations. A closed domain system is built to understand specific types of conversations, while an open domain system can handle broad and versatile conversations with varying outcomes. Building a conversation intelligence system involves three stages: speech recognition, machine learning framework development, and continuous training and maintenance. The choice between a closed and open domain system depends on the scope and complexity of the conversation and the data sources available. An open-domain system offers more flexibility but requires more complex machine learning models, while a closed domain system is more practical for specific use cases but may be biased towards that domain. To build an effective conversation intelligence system, it's essential to gather high-quality training data, design a suitable machine learning framework, and continuously train and maintain the model to mitigate biases and improve accuracy.

The key points of this text revolve around the importance of context in conversation AI, how it enables more intelligent and efficient conversations between humans and machines, and enhances an AI's ability to capture insights and sentiments from conversations. Context is crucial in both human-to-machine (H2M) and human-to-human (H2H) conversations, as it allows the AI to leverage relevant information at the right time, improve user experience, and make informed decisions. The text also highlights various sources of context, including user input, enterprise knowledge, user/task information, session context, and emotion and tone, which can be used to augment an interaction. Additionally, it discusses how capturing context can significantly improve user interactions, such as in sales calls, and provides examples of how context can be used to enhance user experience. Finally, the text touches on the importance of using third-party APIs that have mastered natural language processing (NLP) of voice and text conversations, which can provide substantial contextual understanding without the complexity of building it oneself.

The Devpost Hackathon was launched by Symbl.ai in February 2021, attracting over 200 participants who competed to win up to $10,100 in prizes. The focus of the hackathon was on building or extending applications with Symbl.ai's APIs to extract insights from voice and video inputs. A diverse range of AI-powered use-cases were showcased, including conversational intelligence, conversation platform analytics, Automated Speech Recognition, and more. Three impressive submissions stood out, winning first place with an e-learning enablement app that leveraged Symbl.ai's APIs for topic extraction and sentiment analysis, second place with a call center satisfaction app that utilized real-time transcription and analytics, and third place with a virtual assistant app that automatically took notes from lectures and videos. The hackathon saw a wide array of innovative solutions, and the team at Symbl.ai is grateful to all participants, with some submissions still available for viewing on their Devpost Hackathon Project Gallery page.

There are five main ways in which conversation intelligence (CI) is currently being used or envisaged: real-time action, coaching, predictive analysis, knowledge, and search. Conversation intelligence provides the ability to analyze natural human-to-human conversations in real-time, harnessing mission-critical communications and optimizing them. By leveraging CI, developers can create products that solve problems and enhance human productivity, automating manual repetitive tasks and recommending useful actions at the right moment. Additionally, conversation intelligence can be used for coaching, providing insights to improve skills and knowledge, predicting outcomes with deal intelligence and proactive compliance monitoring, storing valuable information in a knowledge base, and enabling search capabilities through natural language processing. By combining these aspects, businesses can optimize their operations and make humans smarter and more efficient with little effort on the user's part.