November 2022 Summaries

API World` was an in-person event held in San Jose on Oct. 25-27 and a virtual conference on Nov. 1-3, focusing primarily on developers. Symbl.ai had a notable presence at the conference with multiple speaking sessions, giveaways, and connections made. The company's goal was to inspire developers to create APIs that are intuitive and user-friendly, making it easier for them to access the platform's capabilities. This was achieved through creating an open-source Go SDK for the Symbl.ai Platform, which enables users to build virtual assistants on various platforms, including edge devices and scalable systems like Kubernetes. The release of version v0.1.2 of the Go SDK marks a significant milestone in its adoption, and Symbl.ai is encouraging feedback from users.

The Context is Everything Hackathon, sponsored by Symbl.ai and Vonage, was held at VoiceSummit 2022, where participants were challenged to create a project combining Vonage's communications platforms with Symbl.ai's capabilities. The winner of the hackathon was S. Deepak Kumar for his project LiveCaptioner, which provides real-time captioning, complete transcription, topics, and questions asked during video calls/meetings using Symbl.ai's Conversation Intelligence platform and Vonage's Video APIs. Kumar faced challenges while building the app but was able to complete it by teaching himself to use the platforms' APIs quickly with the help of their documentation. The LiveCaptioner features live captioning, real-time transcription, and storage of conversation topics and questions in a tabbed format. To build this project, participants need to have Symbl.ai and Vonage accounts, which can be created for free. Kumar's experience showcases his ability to turn an idea into a functioning app with the help of the platforms' APIs.

I'm excited to join Symbl.ai as a technical writer and share my passion for data analysis, text/sentiment analysis, and machine learning with you. I believe that Symbl.ai's platform has the potential to become a household name, especially when it comes to conversation intelligence and feedback analysis. Conversation intelligence is the key to effective conversation analysis, which can help businesses determine customer intent and interest levels, leading to improved customer satisfaction and increased loyalty. The three main tools used in conversation analytics are keyword extraction, topic summarization, and sentiment analysis, but intelligent analysis engines like Symbl.ai's can improve these processes by flagging keywords in context, separating topics, and understanding speaker emotions. By using these tools accurately, businesses can make informed decisions that lead to better customer experiences and increased loyalty. In summary, conversation intelligence is the future of feedback analysis, and Symbl.ai's platform is at the forefront of this technology. With its ability to analyze business interactions, provide accurate insights, and inform business efforts through AI, I'm excited to share more about how it can help businesses improve their customer experience and achieve success in the competitive online marketplace.

Symbl.ai's Bookmarks API is a convenient way to pinpoint key phrases or moments within human-to-human conversations, allowing users to revisit and share them with others in their organization. This feature can be used for various purposes such as sales, customer experience, and support, by flagging important moments and generating summaries that can be shared with others. The Bookmarks API allows users to bookmark parts of a conversation both in real time and after it has been recorded, link them to user profiles, and discover other users' bookmarks. It also complements Symbl.ai's other conversation intelligence features, enabling users to track topics of interest and revisit important moments on their own or with others.

The Symbl.ai Go SDK has been released with several usability enhancements, including improved named structs and fixed issues with Trackers in the Streaming configuration. The SDK allows developers to process conversation insights in real-time using WebSockets, enabling bi-directional exchange of information between a client and server. This involves logging into the Symbl.ai platform, setting up a WebSocket protocol, defining an input source for audio or conversation, and implementing an InsightCallback interface to feed conversational insights into the SDK. The example uses PortAudio to handle audio processing and demonstrates how to start streaming using WebSockets with the Symbl.ai platform.

NLP is a branch of artificial intelligence that enables more natural human-to-computer communication by bridging human and machine language. NLU is a subset of NLP that converts input text or speech into structured data to extract meaning, recognize entities, and extract information. NLG is another subcategory of NLP that constructs sentences based on a given semantic frame to explain structured data in meaningful sentences humans can understand. These concepts are applied in various use cases such as chatbots, sentiment analysis, automated ticketing support and routing, and automatic text summarization, enabling machines to understand and process natural language, generate responses, and provide actionable insights.

BERT and GPT-3 are two influential modern tools in natural language processing (NLP) and natural language generation (NLG), leading the charge towards artificial general intelligence (AGI). BERT, introduced by Google in 2018, is a neural network-based NLP pre-training technique that enables anyone to train their own state-of-the-art question answering system. It performs well in classification tasks like sentiment analysis and answering questions, and has been used by companies such as Google and Facebook. GPT-3, on the other hand, is a general language model trained on uncategorized text data from the internet, with its largest model boasting around 170 billion parameters, ten times larger than the nearest notable NLP model. It learns these parameters from historical training data and applies its "knowledge" to downstream tasks such as language inference, paraphrasing, and sentiment analysis. GPT-3 has displayed promising performance in zero, one, or a few multitask settings, and is used in various applications including building websites, assisting the development of written content, and generating machine learning code. The two models have different architectural distinctions, with BERT being the encoder part and GPT-3 being the decoder part, and have been shown to be competitive or even better than state-of-the-art models in certain tasks. However, GPT-3's sheer size and black box nature can be restrictive for smaller operations, while BERT requires additional training which uses many resources. The future of NLP and NLG will be bright, busy, and full of possibilities with new models emerging such as BLOOM.