Abstract: A third-party company may assist companies in providing natural language interfaces for their customers. To implement a natural language interface for a company, a configuration may be received that includes information, such as a list intents, seed messages for the intents, and hierarchical information of the intents. An intent classifier may be trained using the configuration, and the natural language interface may be deployed for use with customers. Usage data of the natural language classifier may be collected and used to improve the natural language interface. Messages corresponding to an intent may be clustered into clusters of similar messages, and a prototype message may be obtained for each cluster to provide a human understandable description of the cluster. The information about the clusters may be used to improve the natural language interface, such as by creating a new intent with a cluster or moving a cluster to a different intent.

Abstract: An intent classifier may be used to increase the efficiency of a communications system. A company may provide assistance to a first user using automated processing or by a second user manually responding to the first user. To reduce costs, the company may prefer to use automated processing for assistance where it is available. While a second user is assisting a user, a message from the first user may be processed with an intent classifier to determine that automated support is available to assist the first user, and a suggestion may be presented to the second user to transfer the first user to the automated processing, such as by presenting a button to the second user to transfer the first user to the automated processing. The second user may then transfer the first user to the automated processing and assist other users.

Abstract: A company may desire to maintain a quality level for messages sent by customer service representatives to customers. The company may receive a message input by a customer service representative, modify the message with one or more neural networks, and transmit the modified message to a customer. To modify a message, an input vector may be created for each word of the message where the input vector is created using a word embedding of the word and a feature vector that represents the characters of the word. The input vectors for the words of the message may be sequentially processed with an encoding neural network to compute a message encoding vector that represents the message. The message encoding vector may then be processed by a decoding neural network to sequentially generate the words of a modified message. The modified message may then be transmitted to the customer.

Abstract: A language model may be used in a variety of natural language processing tasks, such as speech recognition, machine translation, sentence completion, part-of-speech tagging, parsing, handwriting recognition, or information retrieval. A natural language processing task may use a vocabulary of words, and a word hash vector may be created for each word in the vocabulary. A sequence of input words may be received, and a hash vector may be obtained for each word in the sequence. A language model may process the hash vectors for the sequence of input words to generate an output hash vector that describes words that are likely to follow the sequence of input words. One or words may then be selected using the output word hash vector and used for a natural language processing task.

Abstract: A user may respond to a request of another user by entering text, such as a customer service representative responding to a customer. Suggestions of resources may be provided to the responding user to assist the responding user in providing a response. For example, a resource may provide information to the responding user or allow the responding user to perform an action. Previous messages between the two users and other information may be used to select a resource. A conversation feature vector may be determined from previous messages, and feature vectors may be determined from the resources. The conversation feature vector and the feature vectors determined from the resource may be used to select a resource to suggest to the responding user.

Abstract: A third-party service may be used to assist entities in responding to requests of users. A third-party service may receive, directly or indirectly, a request of a first user for assistance from a first entity. The third-party service may request information about the first user by sending a request to a computer of the first entity. The third-party service may use the request of the first user and the information about the first user to automatically generate a response to the request of the first user. The third-party service may then transmit, directly or indirectly, the response to the first user.

Abstract: Text of customer support sessions of a company may be processed to detect events that are taking place. For example, an event may be a service disruption. The company may desire to detect the events and take action to address them. The events may be detected by processing customer support sessions during a time window and computing counts, where each count corresponds to a topic and a customer parameter. The topics may be determined from the customer support sessions and the customer parameters may relate to information about the customer (e.g., services received by the customer). Baseline counts may also be computed that correspond to typical or expected behavior when no event is occurring. Event detection scores may be computed by processing the counts and baseline counts and used to determine if an event has occurred. The process may be repeated for subsequent time windows.

Abstract: A third-party company may assist companies in providing natural language interfaces for their customers. To implement a natural language interface for a company, a configuration may be received that includes information, such as a list intents, seed messages for the intents, and hierarchical information of the intents. An intent classifier may be trained using the configuration, and the natural language interface may be deployed for use with customers. Usage data of the natural language classifier may be collected and used to improve the natural language interface. Messages corresponding to an intent may be clustered into clusters of similar messages, and a prototype message may be obtained for each cluster to provide a human understandable description of the cluster. The information about the clusters may be used to improve the natural language interface, such as by creating a new intent with a cluster or moving a cluster to a different intent.

Abstract: A neural network may be used to remove personal information from text (such as names, addresses, credit card numbers, or social security numbers), and replace the personal information with a label indicating the type or class of the removed information. The neural network may comprise multiple layers that compute a context vector for words of the text, compute label scores for words of the text using the context vectors, and select a label for each word using the label scores. Words corresponding to certain labels may be replaced with a label, such as replacing the digits of a credit card number with a label <cc_number>. The redacted text may then be presented to a person or stored for later processing.

Abstract: A language model may be used in a variety of natural language processing tasks, such as speech recognition, machine translation, sentence completion, part-of-speech tagging, parsing, handwriting recognition, or information retrieval. A natural language processing task may use a vocabulary of words, and a word hash vector may be created for each word in the vocabulary. A sequence of input words may be received, and a hash vector may be obtained for each word in the sequence. A language model may process the hash vectors for the sequence of input words to generate an output hash vector that describes words that are likely to follow the sequence of input words. One or words may then be selected using the output word hash vector and used for a natural language processing task.

Abstract: Text of customer support sessions of a company may be processed to detect events that are taking place. For example, an event may be a service disruption. The company may desire to detect the events and take action to address them. The events may be detected by processing customer support sessions during a test window and computing test counts, where each test count corresponds to a topic and a customer parameter. The topics may be determined from the customer support sessions and the customer parameters may relate to information about the customer (e.g., services received by the customer). Baseline counts may also be computed that correspond to typical or expected behavior when no event is occurring. Event detection scores may be computed by processing the test counts and baseline counts and used to determine if an event has occurred. The process may be repeated for subsequent test windows.

Abstract: A third-party company may assist other companies in providing customer support to their customers. The third-party company may provide software to a computer of a customer service representative to present a user interface to assist the customer service representative in responding to customer requests. Third-party company may also send update data to the computer of the customer service representative to cause a portion of the user interface to be updated, where the update data is determined using an intent of a message received from a customer. A message received from the customer may be processed to determine the intent of the message, a template may be obtained using the intent, and the update data may be generated by rendering the selected template. The update data may then be transmitted to the computer of the customer service representative to cause a portion of the user interface to be updated.

Abstract: During a conversation between a customer and a customer support representative, suggestions may be presented to the customer support representative to upsell a product to the customer. Information about the customer and/or information about the conversation may be processed by a computer to determine when to suggest the upsell to the customer support representative and the one or more products to be upsold. The determination may be performed by computing features from the information about the customer and the information about the conversation, and processing the features with one or more classifiers.

Abstract: Mathematical models and a state vector may be used to respond to a customer and perform actions for the customer, such as transmitting an API request to a server to update the customer's address. The mathematical models may include a message encoding model to generate a representation of a message received from a customer, a message generation model to generate a message to transmit to a customer, an API request model to generate an API request to transmit to a server, an API response encoding model to generate a representation of an API response, a state update model to update the state vector using another model output, and an action model to select an action to be performed. A mathematical model may process the state vector in determining an output, and an output of the mathematical model may also be used to update the state vector.

Abstract: A user may respond to a request of another user by entering text, such as a customer service representative responding to a customer. Suggestions of responses may be provided to the responding user so that the responding user may use a suggested response instead of entering text to provide a response. Previous messages between the two users and other information may be used to determine an appropriate suggested response. A conversation feature vector may be determined from previous messages, and response feature vectors may be determined from the conversation feature vector. The response feature vectors may be used to determine an appropriate suggested response.

Abstract: A user may respond to a request of another user by entering text, such as a customer service representative responding to a customer. As the responding user enters text, a completion of the text may be suggested to the responding user so that the responding user may select the completion instead of continuing to enter text. Previous messages between the two users and other information may be used to determine an appropriate completion to the entered text. A topic vector may be determined from previous messages, and a feature vector may be determined from the entered text. The topic vector and the feature vector may be used to identify characters that may follow the entered text.

Abstract: A user may request assistance or information using natural language, such as by speaking or entering a text. A response may be automatically determined to the request of the user be performing semantic processing to understand the request and formulate an appropriate response. A response may be automatically determined, be computing features from the request, using the features to select a node of a graph corresponding to the request, and then causing an action to be performed to provide a response to the user. For example, the action may include providing information to the user, requesting further information from the user, or connecting the user with a customer service representative.