Microsoft Technology Licensing, LLC

Maluuba Inc.

Hannes Schulz

Jing He

Kaheer Suleman

Layla El Asri

Shikhar Sharma

Philip Bachman

Seyed Mehdi Fatemi Booshehri

Justin Daniel Harris

Jeremie Zumer

Emery Fine

Rahul Mehrotra

Described herein are systems and methods for providing hierarchical state tracking in a spoken dialogue system. A sequence of turns is received by a spoken dialogue system. Each turn includes a user utterance and a machine act. At each turn, a value pointer and a turn pointer are provided for that turn. The value pointer represents a probability distribution over the one or more words in the user utterance that indicates whether each word in the user utterance is a slot value for a slot. The turn pointer identifies which turn in a set of turns includes a currently-relevant slot value for the slot, where the set of turns includes a current turn for which the turn point is being provided, and all turns that precede the current turn.

Hierarchical attention for spoken dialogue state tracking

Examples of the present disclosure describe systems and methods relating to conversation state management using frame tracking. In an example, a frame may represent one or more constraints (e.g., parameters, variables, or other information) received from or generated as a result of interactions with a user. Consequently, each frame may represent one or more states of an ongoing conversation. When the user provides new or different information, a new frame may be created to represent the now-current state of the conversation. The previous frame may be retained for later access by what is referred to herein as a 'dialog agent,' which is the portion of the system that can search and use previous state-related information. When an utterance is received, a frame to which the utterance relates may be identified. Thus, the dialog agent may track multiple states simultaneously, thereby enabling conversation features that were not previously possible.

Simultaneous dialogue state management using frame tracking

Described herein are systems and methods for two-stage training of a spoken dialog system. The first stage trains a policy network using external data to produce a semi-trained policy network. The external data includes one or more known fixed dialogs. The second stage trains the semi-trained policy network through interaction to produce a trained policy network. The interaction may be interaction with a user simulator.

Two-stage training of a spoken dialogue system

Described herein are systems and methods for providing a natural language generator in a spoken dialog system that considers both lexicalized and delexicalized dialog act slot-value pairs when translating one or more dialog act slot-value pairs into a natural language output. Each slot and value associated with the slot in a dialog act are represented as (dialog act+slot, value), where the first term (dialog act+slot) is delexicalized and the second term (value) is lexicalized. Each dialog act slot-value representation is processed to produce at least one delexicalized sentence as an output. A lexicalized sentence is produced by replacing each delexicalized slot with the value associated with the delexicalized slot.

Natural language generation in a spoken dialogue system

Growing community of inventors

Karlsruhe, Germany

Hannes Schulz