The Inventive, Often Messy World of Textual Information | by TDS Editors | Nov, 2023


For a number of years, the intersection of textual content and knowledge stayed (roughly) inside the realm of pure language processing (NLP) — the big selection of machine studying duties that leverage textual knowledge for prediction, classification, and advice instruments.

The rise of enormous language fashions has launched a number of thrilling new potentialities into the sector, with novel use instances and progressive workflows popping up at a fast clip. Our highlights this week symbolize a large cross-section of ideas and approaches that dig deeper into this rising space. From immediate engineering to text-to-image and text-to-speech purposes, we’re thrilled to share work by authors who discover the artistic potentialities of textual knowledge as each inputs and outputs of those highly effective fashions. Let’s dive in.

  • Lost in DALL-E 3 Translation
    What occurs once you use text-to-image instruments like DALL-E 3 in languages aside from English? Yennie Jun continues to discover the discrepancies in mannequin efficiency for customers working in under-resourced languages and the methods wherein gender and different biases seep by into the generated pictures.
  • How to Convert Any Text Into a Graph of Concepts
    In his newest put up, Rahul Nayak dives deep into the world of Data-Graph Augmented Technology, strolling us by the method of reworking a textual content corpus right into a Graph of Ideas (GC) after which visualizing it to detect patterns and draw significant insights.
Picture by Jas Min on Unsplash
  • RAG: How to Talk to Your Data
    We’ve lined retrieval-augmented era many instances in latest months, however Mariya Mansurova’s addition to the dialog continues to be very a lot price your time: it presents a compelling, sensible workflow for analyzing buyer suggestions utilizing ChatGPT.
  • FastSpeech: Paper Overview & Implementation
    Textual content-to-speech instruments have made main strides in recent times. To achieve a stable understanding of how they work and the way transformers are employed to enhance their efficiency, don’t miss Essam Wisam’s accessible introduction to the FastSpeech paper from 2019, which facilitated a lot of the progress we’ve seen on this area.
  • Unlocking the Power of Text Data with LLMs
    In the event you’re a newbie who’d like to begin experimenting with cutting-edge text-data methods, Sofia Rosa’s step-by-step information will get you rolling up your sleeves very quickly. It walks us by a whole workflow, from downloading knowledge to working with GPT-3 and analyzing outcomes.
  • A Universal Roadmap for Prompt Engineering: The Contextual Scaffolds Framework (CSF)
    Immediate engineering has emerged as a vital part within the interaction between human instinct and enormous language fashions’ capabilities. Giuseppe Scalamogna goes past primary prompting suggestions and methods to current the contextual scaffolds framework (CSF), a “normal function psychological mannequin for efficient immediate engineering.”

Leave a Reply

Your email address will not be published. Required fields are marked *