HMS Core Discovery Issue 16 Review | Together with Hudun, play the new "sound" state of AI

What are the latest technical capabilities of HMS Core in the AI field? The theme of this episode of Discovery live broadcast is "Fun with Hudun, the new "voice" state of AI", and invited HMS Core machine learning service product manager, machine translation senior expert and HMS Core's new friend "Hudun" to give everyone Demonstrate the speech and language innovation technology of machine learning, and share the macro development trend of machine learning and artificial intelligence. Let's take a look back at the exciting content of this issue!

【Wonderful review】

1. The Douyin Internet celebrity IP "Hudun Little Escort" is coming strongly

2. Simultaneous interpretation makes a new appearance, and the underlying technology is revealed

3. The AI translation capability has been upgraded, and the language expansion can be broadcast

【expert's point】

Nicolas, Senior Expert of Text Machine Translation at Huawei 2012 Lab

Machine translation cannot replace humans for the time being. At present, the translation needs with low demand can rely on the machine, but the translation needs with high requirements also require manual intervention. The manual can achieve all-round quality control from point to point, such as whether the speech is authentic or not, whether the language is fluent or not, and can also contribute Data and knowledge to improve the quality of machine translation.

Hardy, Senior Product Manager, HMS Core ML Kit

AI is a broad field of intelligent machines, and machine learning is one of the core applications of AI. It refers to any computer application that can "learn" on its own without explicit rules from humans. In the future, machine learning technology will pay more attention to emotional experience, and develop in the direction of multi-modality (voice, text, vision), multi-technology (VR/AR, etc.), multi-platform and multi-system collaboration.

【Wonderful Q&A】

Q1: What is TTS tone customization? What new gameplay will TTS sound customization and Hudun collide with?

A: Relying on Huawei's data accumulation and mature algorithms, TTS timbre customization is supported by speech synthesis technology, and only a small amount of clean recording data is required to conduct model training, and obtain high-reduction, high-definition, and high-stability . The exclusive sound library helps to enhance product features and quickly create personalized brand features. ML Kit's new capability TTS and IP "Hudun Little Escort" are cooperating. TTS restores Hudun's timbre through machine training, and will gradually open Hudun's timbre for developers to use in the future, helping developers to use it in various personalized applications. Scenes.

Q2: How can the translation and simultaneous interpretation capabilities provided by the machine learning service enable the App to create a new "sound" state?

A: The text translation capability solves the pain point of users' poor communication due to language barriers. For example, in the call scene, through real-time speech recognition, the recognized text can be quickly converted into the target language text ; in the reading scene, the prompter function is supported to help users quickly see the translated text; after the video app integrates the text translation service, users can smoothly Experience the AI real-time captioning feature . Through the organic integration of speech recognition, machine translation, and speech synthesis, the simultaneous interpretation capability has the characteristics of low latency and high accuracy . It is suitable for high-real-time scenarios such as conferences, live broadcasts, and speeches. Real-time output of audio content into the target language text, generating bilingual subtitles, and real-time broadcast of the target language text, reducing the cost of understanding, with both quality and efficiency.

Q3: In addition to the above speech and language capabilities, are there other innovative capabilities newly launched by the machine learning service?

A: In the field of financial e-commerce, machine learning services also provide live detection capabilities . The motion detection capability uses technologies such as face key point positioning and face tracking, and can verify whether the user is operating by a real living body in the form of instructions and actions . In the financial fields with high real-name system and security requirements, such as banking, securities, and lending, live detection can be used as an auxiliary verification in the user's remote registration and password retrieval links, helping users identify fraudulent behaviors, effectively resist attacks, and ensure business security.

Q4: What is the macro technical development trend of machine learning?

A: First, machine learning will pay more attention to emotional experience . Machines will have the ability to identify, understand and express emotions, to recognize user needs and changes in environmental information, to understand people's emotional intentions, and to make appropriate responses; secondly, the development of multimodality . Deep learning technology is developing from single modalities such as speech, text, and vision to learning multimodal intelligent learning. In the future, it is even possible to fuse signals that are difficult to quantify, such as smell, taste, and psychology, to realize joint analysis of multiple modalities, and to assist human work in more scenarios and more businesses; again, it is the integration of multiple technologies , such as VR. /AR and the Metaverse, etc. It is believed that in the future, AI will also present a multi-platform and multi-system collaborative situation to achieve a wider range of empowerment. The collaborative combination of general platforms, industry platforms and end-side applications will realize the function customization and expansion of specific applications in a way of integrating software and hardware. .

Welcome to the home page of HMS Core Machine Learning Service for more technical details.

Learn more details>>

Visit the official website of Huawei Developer Alliance
Get development guidance documents
Huawei Mobile Services Open Source Warehouse Address: GitHub , Gitee

Follow us to know the latest technical information of HMS Core for the first time~