18457708. MULTI-DOMAIN JOINT SEMANTIC FRAME PARSING simplified abstract (Microsoft Technology Licensing, LLC)

MULTI-DOMAIN JOINT SEMANTIC FRAME PARSING

Organization Name

Microsoft Technology Licensing, LLC

Inventor(s)

Dilek Z. Hakkani-tur of Kirkland WA (US)

Asli Celikyilmaz of Kirkland WA (US)

Yun-Nung Chen of Taipei (TW)

Li Deng of Redmond WA (US)

Jianfeng Gao of Woodinville WA (US)

Gokhan Tur of Kirkland WA (US)

Ye Yi Wang of Redmond WA (US)

MULTI-DOMAIN JOINT SEMANTIC FRAME PARSING - A simplified explanation of the abstract

This abstract first appeared for US patent application 18457708 titled 'MULTI-DOMAIN JOINT SEMANTIC FRAME PARSING

Simplified Explanation

The patent application describes a processing unit that can train a joint multi-domain recurrent neural network (JRNN) for spoken language understanding (SLU). This model can perform tasks such as slot filling, intent determination, and domain classification simultaneously.

• The processing unit can train a model called JRNN, which is a combination of bi-directional recurrent neural network (bRNN) and recurrent neural network with long-short term memory (RNN-LSTM).
• The trained model can estimate a complete semantic frame per query, which includes information about intents and slots across multiple domains.
• The JRNN model enables multi-task deep learning by leveraging data from multiple domains.

Potential Applications

• Spoken language understanding systems in various domains such as customer service, virtual assistants, and voice-controlled devices.
• Natural language processing applications that require understanding user queries and extracting relevant information.

Problems Solved

• Traditional SLU models often focus on a single domain, making it difficult to handle queries that involve multiple domains.
• Training separate models for each domain can be time-consuming and resource-intensive.
• Existing models may struggle to accurately estimate semantic frames that include intents and slots across multiple domains.

Benefits

• The joint multi-domain model allows for more accurate and efficient processing of user queries involving multiple domains.
• Training a single JRNN model reduces the need for separate models for each domain, saving time and resources.
• The model leverages the data from multiple domains, enabling better performance and generalization.

Original Abstract Submitted

A processing unit can train a model as a joint multi-domain recurrent neural network (JRNN), such as a bi-directional recurrent neural network (bRNN) and/or a recurrent neural network with long-short term memory (RNN-LSTM) for spoken language understanding (SLU). The processing unit can use the trained model to, e.g., jointly model slot filling, intent determination, and domain classification. The joint multi-domain model described herein can estimate a complete semantic frame per query, and the joint multi-domain model enables multi-task deep learning leveraging the data from multiple domains. The joint multi-domain recurrent neural (JRNN) can leverage semantic intents (such as, finding or identifying, e.g., a domain specific goal) and slots (such as, dates, times, locations, subjects, etc.) across multiple domains.