Text & Natural Language Processing (NLP)
Figure Eight’s human-in-the-loop machine learning platform is the most powerful solution for the creation of text and natural language training data at enterprise scale.
With a decade of experience and tens of millions of text annotations, Figure Eight is the industry leader in NLP training data. We can help label, create, and validate domain-specific text data in a wide array of languages.
Trusted by the world’s most innovative companies
Our platform powers:
Sentiment Analysis
Figure Eight provides the accuracy and nuance that out-of-the-box sentiment solutions simply cannot. That’s because our human-in-the-loop approach allows you to tailor every step of the process to your exact specifications, including going beyond simple sentiment and getting to the “why” behind the positive and negative.
Machine Learning Assisted Text Annotation
Our platform powers highly accurate named entity recognition (NER) tasks, via our Machine Learning Assisted Text Annotation solution, with dedicated tooling and robust quality controls. Whether you’re looking to identify parts of speech, classify proper nouns like places and people, or any other important entities for your project, with Figure Eight, you create your own custom ontology and our platform will take care of the rest.
Intent Classification
Intent classification is an important step in training chatbots, virtual assistants, and other conversational agents. And since intent is nuanced, a human-in-the-loop approach to creating training data is the best strategy to train a model that understands what your users actually want.
Utterance Collection
Free text utterance collection jobs let you leverage the power of a distributed workforce of fluent annotators to collect text strings based on prompts or scenarios to power your conversational agents. We support dozens of languages and employ workflows and quality controls to make sure your free-text utterances are of the highest quality.
Optical Character Recognition and Transcription
Figure Eight has created ground truth for optical character recognition (OCR) models for some of the most ambitious companies in the world. Whether you’re looking to identify important areas of text in PDFs with bounding boxes, validate model outputs, or transcribe key sections of PDFs or other documents, we have a robust set of solutions to make your algorithms more accurate and more confident.
Information Extraction for NLP
When you need to understand the subject or topic of a text, our platform’s information extraction solutions are a great way to do it. We can do everything from simply classifying sections or whole documents to enabling highlighting or copy/paste functionality to identify the strings of text that are important for your NLP models.
Relevance Scoring
Conversational agents don’t always return the best results. That’s where relevance scoring jobs can help. Have our human-in-the-loop platform identify if your search or conversational agent outputs are relevant to a user’s request so you can iterate on your model and return the best possible results for every query.
Text Categorization
You can categorize any kind of text with Figure Eight. Whether you’re looking to understand a newswire article or detect spam, our platform provides a fully customizable instructions, ontologies, and form functionalities to get you the training data you need to optimize your model for understanding.