Systems, methods, and computer program products for identifying a relevant candidate product in an electronic marketplace. Embodiments perform a visual similarity comparison between candidate product image visual content and input query image visual content, process formal and informal natural language user inputs, and coordinate aggregated past user interactions with the marketplace stored in a knowledge graph. Visually similar items and their corresponding product categories, aspects, and aspect values can determine suggested candidate products without discernible delay during a multi-turn user dialog. The user can then refine the search for the most relevant items available for purchase by providing responses to machine-generated prompts that are based on the initial search results from visual, voice, and/or text inputs. An intelligent online personal assistant can thus guide a user to the most relevant candidate product more efficiently than existing search tools.

Intelligent online personal assistant with multi-turn dialog based on visual search

According to an example embodiment of the present disclosure, there is provided a method and a device for cross-type conversation, an apparatus, and a computer-readable storage medium. The method includes receiving a message from a user and determining a reply type for the reply message based on the constructed map, wherein nodes in the map represent semantic content and edges in the map representsemantic associations between nodes. In addition, the method includes providing a reply to the message based on the reply type. According to an embodiment of the present disclosure, by using a prioriknowledge in a map to assist in determining a reply type, it is possible to reduce the dependence of a chat robot on training data, so that the chat robot can actively guide a cross-type conversationand maintain the logicality of the semantic content of the conversation.

Method and device for cross-type conversation, apparatus, and computer-readable storage medium

In one embodiment, a character engine models a character that interacts with users. The character engine receives user input data from a user device, and analyzes the user input data to determine a user intent and an assessment domain. Subsequently, the character engine selects inference algorithm(s) that include machine learning capabilities based on the intent and the assessment domain. The character engine computes a response to the user input data based on the selected inference algorithm(s) and a set of personality characteristics that are associated with the character. Finally, the character engine causes the user device to output the response to the user. In this fashion, the character engine includes sensing functionality, thinking and learning functionality, and expressing functionality. By aggregating advanced sensing techniques, inference algorithms, character-specific personality characteristics, and expressing algorithms, the character engine provides a realistic illusion that users are interacting with the character.

Modeling characters that interact with users as part of a character-as-a-service implementation

A dialogue system comprising: an input for receiving input data relating to a speech or text signal originating from a user; an output for outputting speech or text information specified by a dialogue act; and a processor configured to: generate features from the input signal; for each of a plurality of classifier models, each classifier model corresponding to a dialogue slot, and for one or more values corresponding to the dialogue slot, input features generated from the input signal, the classifier model outputting a probability corresponding to each of three or more relations, wherein the relations specify the relation of the value to the dialogue slot; update a belief state based on the outputs of the classifier models; determine a system dialogue act by inputting information relating to the belief state into a policy model; output speech or text information specified by the determined dialogue act at the output.

Dialogue system, a dialogue method, a method of generating data for training a dialogue system, a system for generating data for training a dialogue system and a method of training a dialogue system

An artificial intelligence problem is solved using an artificial intelligence memory graph data structure and a lexical database to identify supporting knowledge. A natural language input is received and classified into components. A starting node of an artificial intelligence memory graph data structure, which comprises one or more data nodes, is selected to begin a search for one or more supporting knowledge data nodes associated with the classified components. Starting at the starting node, the artificial intelligence memory graph data structure is searched using a lexical database to identify the one or more supporting knowledge data nodes. An artificial intelligence problem is identified and solved using the one or more identified supporting knowledge data nodes of the artificial intelligence memory graph data structure.

Adaptive, interactive, and cognitive reasoner of an autonomous robotic system

Systems and methods for generating prompts for further data from a user in a multi-turn interactive dialog. Embodiments improve searches for the most relevant items available for purchase in an electronic marketplace via a processed sequence of user inputs and machine-generated prompts. Question type prompts, validating statement type prompts, and recommendation type prompts may be selectively generated based on whether a user query has been sufficiently specified, user intent is ambiguous, or a search mission has changed. Detection of a new dominant object denotes search mission change. Contextual associations between prompts and user replies are maintained, but a search mission change results in previous context data being disregarded. Prompts can be directed to unspecified knowledge graph dimensions based on data element association strength values, relative data element positions and depths in the knowledge graph, and generated following a predetermined order of data element knowledge graph dimension types.

Selecting next user prompt types in an intelligent online personal assistant multi-turn dialog

Vehicles and other items often have corresponding documentation, such as registration cards, that includes a significant amount of informative textual information that can be used in identifying the item. Traditional OCR may be unsuccessful when dealing with non-cooperative images. Accordingly, features such as dewarping, text alignment, and line identification and removal may aid in OCR of non-cooperative images. Dewarping involves determining curvature of a document depicted in an image and processing the image to dewarp the image of the document to make it more accurately conform to the ideal of a cooperative image. Text alignment involves determining an actual alignment of depicted text, even when the depicted text is not aligned with depicted visual cues. Line identification and removal involves identifying portions of the image that depict lines and removing those lines prior to OCR processing of the image.

Document optical character recognition

Many real world problems require fast and efficient lexical comparison of large numbers of short text strings. Search personalization is one such domain. We introduce the use of feature bit vectors using the hashing trick for improving relevance in personalized search and other personalization applications. We present results of several lexical hashing and comparison methods. These methods are applied to a user's historical behavior and are used to predict future behavior. Using a single bit per dimension instead of floating point results in an order of magnitude decrease in data structure size, while preserving or even improving quality. We use real data to simulate a search personalization task. A simple method for combining bit vectors demonstrates an order of magnitude improvement in compute time on the task with only a small decrease in accuracy.

The Bitwise Hashing Trick for Personalized Search

Selecting types of generated prompts for further data from a user in a multi-turn interactive dialog. In one scenario, a processed sequence of user inputs and machine-generated prompts improves searches for the most relevant items available for purchase in an electronic marketplace. The number of prompts may be limited to a predetermined maximum value. Prompt generation is minimized by incorporating into a knowledge graph world knowledge that helps user intent inference. Prompt generation may be suppressed if a search indicates the reply to a prompt will not lead to any satisfactory search results. Prompts can provide suggestions for available search results that either meet all query constraints, or meet only some query constraints if a search indicates no search results are available that meet all query constraints. Prompts can provide suggested incisive reply phrasing likely to improve search results through an affirmation or negation reply.

Selecting next user prompt types

Many real world problems require fast and efficient lexical comparison of large numbers of short text strings. Search personalization is one such domain. We introduce the use of feature bit vectors using the hashing trick for improving relevance in personalized search and other personalization applications. We present results of several lexical hashing and comparison methods. These methods are applied to a user’s historical behavior and are used to predict future behavior. Using a single bit per dimension instead of floating point results in an order of magnitude decrease in data structure size, while preserving or even improving quality. We use real data to simulate a search personalization task. A simple method for combining bit vectors demonstrates an order of magnitude improvement in compute time on the task with only a small decrease in accuracy.

Braddock Gaskill

Papers

Selecting next user prompt types in an intelligent online personal assistant multi-turn dialog

Document optical character recognition

The Bitwise Hashing Trick for Personalized Search

Selecting next user prompt types

The Bitwise Hashing Trick for Personalized Search