ASTRI’s speech recognition incorporates the researches in speech-to-text in Cantonese mixed with English, speaker diarization, voiceprint verification and indexing to address the challenges of information retrieval from a large quantity of audio.
Supports both real time and offline transcribing
Supports both real time and offline transcribing
Cantonese / English mixed
Putonghua / English mixed
Accuracy 85+% in specific domain
Emotion detection
Audio indexing
ASTRI’s AI Agent support customizable chatbots through innovations in leveraging deep learning, semantic analytics, and speech recognition & generation with NLP. It is proven effective in enhancing operational efficiency and improving customer experience by accurately addressing over 80% of customer enquiries and transcription needs in mixed language environments.
Innovations
The AI Chatbot with speech-to-text capabilities understands Hong Kong’s local mix of Cantonese, English, as well as professional terms, industrial jargons and slangs. The innovation also includes the followings:
- Large Language Model are deep learning algorithms that can recognize, summarize, translate, predict, analyse and generate content using very large datasets.
- Domain-specific Large Langue Model is fine-tuned large language model with domains dataset.
- Text-to-speech support Cantonese mixed with English, Mandarin, English with voice cloning and voice conversion.
- Knowledge Graph provides a richer context and understanding of information by representing entities and their relationships, enabling both humans and machines to better interpret and interact with the data.
- Virtual Avatar generates the talking face for making conversation more interactive.
- Speech recognition engine that can recognise and segregate voices of multiple speakers in a group conversation
- Voicebot that can respond to inquiries in the corresponding language
- Domain-specific database management platform for non-IT users to build and customise client-specific knowledgebase
- Big data and NLP analysis for analysing enquiries that are not supported by prior data training or any existing keywords databases
- Real-time language analysis with automated punctuation insertion into transcribed texts
- Version control enabling transparent updates minimizing interference
- Server-based system offering enhanced security to accommodate sensitive client information and to reduce cost for future modifications
ASTRI’s NLP for analysing enquiries that are not supported by prior data training or any existing keywords database. We help to build Domain-specific database management platform for non-IT users to build and customise client-specific knowledgebase. NLP can parse Chinese sentences and understand the contextual meanings.
For any inquiries or feedback on using our products in your organisation, please reach out to us here. Thank you!