September 2021 Summaries

Symbl.ai is a general-purpose conversation intelligence API platform that enables developers to transform voice, video, message, or broadcast conversations into actionable insights using artificial intelligence. The platform provides a domain-agnostic, cross-channel approach to analyzing contextual information in multi-channel conversations. With Symbl.ai, developers can analyze recorded calls from CPaaS platforms like Telnyx and gain insights such as sentiment analysis, topics, and questions. By integrating with webhooks, developers can receive real-time updates on their HTTP requests through cURL commands. The platform offers a free account sign-up process, allowing developers to access their API keys for authentication purposes. Symbl.ai's Conversation Intelligence API platform accepts various file formats, including .mp3 files, making it easy to analyze and understand conversations. By leveraging artificial intelligence models and continuously deployed technology, Symbl.ai empowers developers to drive real-time engagement and transform channels digitally.

The HackerEarth "Integrate to Disrupt" hackathon, hosted by Symbl.ai, brought together over 2000 participants and generated highly impressive submissions. The five-month hackathon was divided into two phases: ideation and execution, with a focus on connecting developers to conversational intelligence in various applications. Participants were encouraged to innovate, transform, connect, or visualize using Symbl.ai's suite of APIs, which enable the building of intelligent speech-to-text functionality, extracting contextual insights, generating domain-specific insights, and accessing advanced conversation analytics. The top three winners were announced after a thorough review of submissions, with Mehul Nirala taking first prize for his office utility service built on Symbl.ai's comprehensive suite of APIs, while Team Vulcan and Team TriDev took second place for their AI-powered chatbot and video analyzing platform. Third place was awarded to two teams, Sophia Sydir led Augmented Transcripts and Drill Down, respectively. The hackathon aimed to push the limits of conversation intelligence and encouraged developers to continue testing the capabilities of Symbl.ai's Conversation Intelligence API Developer Platform.

Symbl.ai and Telnyx partnered to encourage developers to build innovative solutions at TADHack Global 2021, a yearly conference that brought together hundreds of developers from various companies. The event had over 1000 registrations across three in-person locations and several online platforms, with sponsors including Symbl.ai and Telnyx providing resources and support. A joint prize of $4,000 was offered to teams that combined both APIs, which went to a team led by Lily Madar and Steven Goodwin for their projects Wizard Chess, Colloqui11ly, and Podcast Annotator. These projects showcased innovative uses of Symbl.ai's conversation intelligence and Telnyx's CPaaS APIs, with Wizard Chess enabling voice-controlled chess gameplay for those with visual impairments, Colloqui11ly providing an accessible conferencing solution, and Podcast Annotator allowing hands-free note-taking during podcast listening.

GPT-3 is a probabilistic language model developed by Open.ai, trained on uncategorized internet text to produce human-like text. It has a massive capacity of around 170 billion parameters, making it significantly larger than its predecessors. GPT-3 can be used as a plug-and-play tool for common tasks like sentiment analysis and code generation, but requires fine-tuning for more specialized use cases. The model is being released with beta-only access to mitigate ethical concerns about bias and potential misuse. Despite its impressive capabilities, GPT-3 still struggles with common sense and simple math, and has room for improvement in addressing bias. Its release marks an important milestone in natural language processing, but not a revolutionary paradigm shift.

Symbl.ai is releasing a new product called under its experimental wing of “Labs”, for summarization: the Summary API, which provides developers with the ability to summarize conversations into summaries. The API synthesizes vast amounts of conversation data into a summary and helps filter domain-specific language in source texts. It seeks the most relevant information based upon the context of the conversation and is designed specifically for developers to utilize in their stacks for data science analysis or application development. To use the Summary API, developers can sign up on the platform, upload a conversation through Async APIs with a query parameter set to `?enableSummary=true`, and make a call to the Summary API with the conversation ID. The API currently supports Async APIs, but support for real-time summarization will be added soon, while streaming or telephony API users can use the refresh parameter to generate summaries after the conversation has ended.

The recently launched /labs API endpoint at Symbl.ai is designed for developers who want to try out its most experimental Artificial Intelligence technologies. This API provides real-time engagement builders with new experiences around conversations, including Comprehensive Action Items and Summarization APIs. The Comprehensive Action Items provide a rephrased form of the original action item message enriched with contextual information from conversation details. Symbl.ai's Summarization API captures key discussion points in a conversation to create succinct and context-based summaries. Additionally, the company offers a Personally Identifiable Information (PII) Identification and Redaction API to identify and censor sensitive information. Developers can join the experiment by trying out these APIs and providing feedback, shaping the future course of Symbl.ai's product development.

The analysis of conversation data is being digitally transformed by companies impacted by the shift from office to remote work through Symbl.ai's Conversation Intelligence API platform. The platform enables developers to extract insights from multiple conversations in real-time, analyzing key features such as speaker overlap, silence, pace, and talk time. A developer has built a web application using PHP, Graph.js, and Bootstrap 4 to create a graph displaying the talk time of speakers in a conversation. The application takes user input for the API credentials and conversation ID, generates a token, and uses Symbl.ai's Conversation Analytics API to retrieve the conversation data. The application then displays a bar graph using Chart.js, with each speaker represented by a different color based on their talk time. The developer plans to add features such as a dropdown menu to select the type of return data from Symbl.ai's APIs and a scheme for color patterns to highlight commonalities between speakers.

Symbl.ai is a Conversation Intelligence API platform that enables developers to augment the speaking realities of conversations in real-time. It provides an Async API for separating speakers and processing audio through a URL. The platform's Python SDK allows developers to access programmable intelligence directly in their mobile or web applications, client or server. With Symbl.ai's SDK, developers can perform speaker separation, sentiment analysis, and create separate Numpy arrays for analyzing the speakers separately. The platform also provides access to many of its APIs through its Python SDK, including real-time Telephony API and Async APIs for voice, video, or messaging. Symbl.ai invites developers to reach out to them via email, join their Slack channels, participate in hackathons, fork their Postman public workspace, or git clone their repos at GitHub.

Telnyx Voice API Initiates Conversationally Intelligent Transfers to Live Operators with Symbl.ai’s APIs in a Python Flask App with an IVR/AVA`, This demo application utilizes Telnyx's developer-centric CPaaS platform and Symbl.ai's Conversation Intelligence APIs to create a conversational intelligence-powered call routing system. The system uses Python Flask as the backend framework, ngrok for secure webhooks, and Telnyx's Voice API for call routing. Symbl.ai's APIs are used to analyze customer sentiment and trigger transfers to live operators when sentiments drop below a preset threshold. The application is designed to provide an omnichannel, contextual, artificially intelligent calling experience with programmable intelligence that can be extended beyond mere automated speech recognition to analyze sentiments on messages.

Symbl.ai is launching a new product under its "Labs" experimental wing, offering an API for Comprehensive Action Items. This allows developers to capture the most important action items from a conversation, going beyond mere transcription, and providing a deeper understanding of the conversation itself. The API provides contextual insights such as questions, follow-ups, or action items without requiring specialized expertise in data science, natural language processing, or Python. Comprehensive Action Items differ from traditional Action Items by analyzing contextuality, including speaker names, context, and overall descriptions of the action items. These provide access to an understanding of actionable speech throughout conversations, with timestamps, enabling time series analysis over multiple conversations. The API is fully enabled and available for public use, with Symbl.ai recommending using Postman or a similar tool for making API calls.

Symbl.ai is a Conversation Intelligence API platform that enables developers to augment the speaking realities of conversations in real-time by extending the functionalities of streaming platforms through a few API calls, including its Async API which separates speakers' voices and provides programmable intelligence directly in mobile or web applications. Developers can access Symbl.ai's Python SDK to process audio files, URLs, or streams, and configure parameters for speaker separation, enabling features like diarization, speaker counting, and message analysis. The platform also offers various APIs, including Telephony APIs, Async Text API, and others, allowing developers to expand the experiences around calls in real-time and access contextual insights. Symbl.ai invites developers to join its community through email, Slack channels, hackathons, and GitHub repositories.

Symbl.ai's Async API provides a REST interface for submitting recorded or saved conversations to be processed asynchronously, with the ability to retrieve job status and completion via webhooks. To use the API, developers must register for an account, authenticate using their appId and appSecret, and establish a WebSocket endpoint for live transcription. The webhook URL parameter is used to receive automatic updates on job status, and servers must respond with a 200 or 201 status code upon receipt of successful POST requests. A sample implementation of a simple Node.js server using Express is provided, which can be used as a starting point for handling webhooks and interacting with the Async API. Once a status: 'completed' event is received, conversation transcripts and insights are available via the Conversation API, and additional logic can be added to automate this process.

I can provide you with an objective, neutral, and interesting summary of the text in one paragraph. Here it is: Business intelligence systems have been used to help companies make informed decisions by analyzing data for decades, but they can be limited to traditional data sources. By incorporating conversation analytics, businesses can unlock new insights from human-to-human conversations, such as trending topics and sentiments, and gain a deeper understanding of how their business operates. Integrating BI with conversation intelligence allows businesses to create systems that offer the benefits of both worlds, providing real-time access to granular aspects of conversation data and actionable insights at scale. This can be applied across various industries, including distance learning, sales calls, and telehealth, enabling businesses to gain a competitive edge by using data from conversations to inform their operations.

Improve speech to text accuracy in real-time transcription by capturing clear audio, using lossless audio formats and techniques like speaker diarization or separate channels. Real-time transcription can be challenging due to human-to-human conversations with accents, overlapping voices, and unfamiliar words. To address this, AI-assisted transcription can help improve accuracy when combined with factors like sentence boundary detection, punctuation detection, and custom vocabulary training. Using APIs like Symbl.ai's context-aware conversation intelligence platform can also enhance the accuracy of real-time transcription by providing tools for speaker diarization, custom vocabulary, and advanced contextual AI.

The conversation intelligence landscape is evolving, driven by advancements in speech recognition and context understanding. Larry Heck, a pioneer in speech technology, shares his insights on human-to-human conversations, augmentation of meetings, and the importance of building feedback loops into conversation intelligence products. He emphasizes the need for passive systems to learn from humans before being active, citing technical challenges such as open microphone as obstacles to progress. The industry is moving towards more robust and contextual AI solutions that can augment human conversations, rather than interrupting them. Heck believes that augmentation technology has the potential to enrich communication channels between humans. The key to successful conversation intelligence lies in understanding the importance of feedback loops, which require businesses to effectively communicate value from day one, even when the technology is not fully polished. Heck's vision for the future of conversation intelligence involves creating systems that can recognize context, understand conversations, and make higher-level inferences and reasoning. This requires injecting knowledge into the technology stack, including business knowledge, to enable more effective human-machine interactions. With his expertise and experience, Heck is poised to drive innovation in this rapidly evolving field.

To access Symbl.ai's WebSocket for real-time audio transcription, developers need to register for an account and authenticate using their appId and appSecret. This authentication process allows users to receive their x-api-key, which is necessary for accessing the WebSocket. The x-api-key enables developers to choose from different types of message objects, such as recognition_result or message_response, depending on their specific use case. Recognition_result is suitable for live captioning, while message_response is better suited for transcripts or beyond. By understanding the differences between these message objects and following the guide provided by Symbl.ai, developers can effectively utilize the WebSocket to access real-time audio transcription services.