In the rapidly growing AI market that especially focused on foundation models and Generative AI, the quality of datasets directly impacts the performance. In real-world applications, data is messy and improving models is not the only way to get better performance. As AI continues to transform industries, the need for high quality datasets has become critical for developing responsive, adaptable, and intelligent systems.
At the Interspeech 2024, Dataocean AI, a global leader in AI data solutions, officially launched its latest offerings: high-quality off-the-shelf datasets. This exciting announcement further illustrates the company’s position as a pioneer in the AI technology domain.
Dataocean AI introduced its newest corpus designed to meet the demands of various application scenarios – “Massively Multilingual Speech Corpus”. This corpus was recording from 215,891 speakers with total of 259,672 hours, covering over 100 languages. Along with this corpus, Dataocean AI also showcased its datasets in European languages. These meticulously labeled high quality datasets, covering English, French, Spanish, Turkish and Swedish, known for their diversity and accuracy, promise to enhance the performance of AI models across industries, such as smart finance, AI assistant, in-cabin, smart home, and other trendy topics related to AI.
The key strength of Dataocean AI’s datasets lies in their ability to deliver high precision across different fields.
- For data collection process, Dataocean AI leverages its extensive global network, comprising native speakers who professionally record in over 200+ languages. The company owns a team of native and professional speakers for these recordings and employs high-fidelity equipment within professional recording studios including indoor, outdoor, and in-cabin environments.
- For data labeling process, the company offer datasets that are labeled with their advanced self-developed platform with human in the loop. The expert team consist of scholars and specialists that covering multiple scenarios, and they have successfully build over 1100 speech datasets that match top quality standards, fulfilling the evolving needs of the AI industry.
In addition to speech datasets, Dataocean AI also owns over 1600 high-quality training datasets with proprietary intellectual property rights, covering a wide range of fields including foundation models, autonomous driving, finance, healthcare, and law. At the same time, its self-developed data processing platform, DOTS, equipped with more than 200 algorithms and hundreds of data processing tools, can achieve powerful functions such as automated labeling and assisted labeling, better helping customers reduce costs and increase efficiency. Additionally, they have earned data security regulations such as the European GDPR, and obtained certifications for ISO 9001, ISO 27001, and ISO 27001, ensuring safety and compliance.
Along with the high-quality datasets, Dataocean AI also empower LLMs through world-class live data collection for pre-trained and SFT/RLHF/red teaming for fine-tuning, as well as model evaluation.
Dataocean AI’s goal is to deliver one-stop data solution that ensuring their partners and clients can build reliable, adaptable AI models. This commitment to excellence is central to the company’s mission of driving innovation in AI.