March 2020 Summaries

The Symbl API Platform is a comprehensive suite of APIs for analyzing natural human conversations, offering real-time analysis of free-flowing discussions and AI-powered conversational intelligence for various channels including voice, video, and text. The platform provides a range of features such as contextual language understanding, conversation APIs, summary UI, flow manager, and unmatched developer support to help developers build unique product experiences.

The Business Intelligence (B.I.) landscape is evolving with the adoption of Artificial Intelligence (AI), expanding its capabilities beyond simple data presentation to predictive analysis, real-time insights, and action-oriented decision-making. As AI enhances B.I., new sources of data are being explored, including IoT, click tracking, RPA systems, and conversations, which are analyzed using big data technologies like Apache Hadoop and Spark. Predictive business analysis is becoming increasingly popular, enabling organizations to predict future outcomes based on historical data, while conversational intelligence platforms can extract insights from real-time customer conversation experiences. The influx of data encourages the use of proactive, real-time systems for reporting and analysis, with a focus on actionable intelligence and the exploitation of all available data sources.

The Symbl Voice SDK is a tool that analyzes voice conversations on SIP or PSTN networks, generating actionable outcomes through contextual conversation intelligence. It allows developers to integrate this functionality into their existing telephony system capabilities, such as call center applications and PBX systems. The SDK can be used to provide real-time or post-conversation intelligence to customers, enabling them to gain insights from voice conversations. The SDK supports dialing through a simple phone number using the PSTN endpoint. It also allows developers to push events on an ongoing connection, sending speaker events that contain details of users who started and stopped speaking. The SDK can be used to send summary emails after the conversation is finished, containing insights such as meeting topics, action items, and follow-ups. Developers can tune their summary page with query parameters, such as minimum score threshold for rendering insights, enabling or disabling certain features, and ordering topics by importance score or position in the transcript. The SDK provides a way to generate insights from voice conversations, which can be pushed to downstream channels like RPA, business intelligence platforms, task management systems, and others using the conversation API.

Capturing audio and deriving real-time insights from customer interactions can be achieved using Twilio Media Streams and Symbl's WebSocket API. To get started, developers need to set up a local server that can handle WebSocket connections, connect their Twilio number to the server, and configure the server to stream media packets from the call to the Symbl server. This is done by creating an HTTP route that returns TwiML instructing Twilio to stream audio to the server. Once the server is set up, developers can use ngrok to expose the port to the internet, navigate to their Twilio Studio Dashboard, and create a new flow to join Symbl to customer and agent conversations. The integration allows for real-time insights to be generated from the conversation, which can be fetched using the Symbl API and pushed to downstream channels such as Trello or Slack.

Conversations are complex and challenging to analyze due to their inherent variability in data, context, and user influence, making traditional NLP approaches less effective. The lack of clear structure and hierarchical meaning in conversations adds to the difficulties, whereas deep learning models, although powerful for certain tasks, struggle with the train-test mismatch and context injection. Understanding conversation intelligence requires addressing these challenges and exploring hybrid learning approaches that can effectively handle uncertainty and incomplete information.

Deep Learning models are used to predict patterns in complex sequences by learning from data, rather than relying on simple rules or algorithms. The models work by constantly iterating over the training data, making predictions and adjusting their parameters until the error between the prediction and ground truth is minimal. In conversations, using conversational data is challenging due to its rarity, but ensembling multiple Deep Learning models can provide exponential payoffs. This approach allows each model to learn specific things from different data sets, and combining them results in a better overall model. While even ensemble models can fail to spot simple patterns, additional techniques such as standard rule-based approaches can be used to refine the results.

We are Symbl, formerly Rammer.ai, a company that aims to break down language into its essence by identifying symbols - small semantic units that carry high meaning intensity - in conversations. The name change reflects the company's growth and purpose as individuals and as a team, embodying an overwhelming sense of empathy across the board. With this new identity, Symbl is poised to continue empowering teams with its mission-driven approach.