Abstract: A computer-implemented interaction control method includes determining, based on a first requirement, a first category from one or more categories estimated from each of a plurality of pieces of information input by a user, selecting, based on the determined first category, a first conversation topic for interaction with the user from conversation topics, executing, by using the first conversation topic, the interaction with the user via a user interface, determining, when detecting that the first conversation topic is inappropriate in accordance with a result of the interaction executed by using the first conversation topic, based on a second requirement, a second category from the one or more categories, selecting, based on the determined second category, a second conversation topic for the interaction with the user from the conversation topics, executing, by using the second conversation topic, the interaction with the user via the user interface.

Abstract: An electronic device is provided. The electronic device includes a first memory configured to operate at a first speed and store compressed vectors corresponding to words, and scaling factors corresponding to the compressed vectors; a second memory that is faster than the first memory and is configured to store a first group of the compressed vectors, and store a first group of the scaling factors; and a processor configured to obtain a first compressed vector and a first scaling factor corresponding to an input word from the first memory or the second memory and process the obtained first compressed vector and the obtained first scaling factor by using a neural network.

Abstract: In some implementations, a language proficiency of a user of a client device is determined by one or more computers. The one or more computers then determines a text segment for output by a text-to-speech module based on the determined language proficiency of the user. After determining the text segment for output, the one or more computers generates audio data including a synthesized utterance of the text segment. The audio data including the synthesized utterance of the text segment is then provided to the client device for output.

Abstract: A method for training a neural network, includes: training a super network to obtain a network parameter of the super network, wherein each network layer of the super network includes multiple candidate network sub-structures in parallel; for each network layer of the super network, selecting, from the multiple candidate network sub-structures, a candidate network sub-structure to be a target network sub-structure; constructing a sub-network based on target network sub-structures each selected in a respective network layer of the super network; and training the sub-network, by taking the network parameter inherited from the super network as an initial parameter of the sub-network, to obtain a network parameter of the sub-network.

Abstract: A method for personalizing a message between a sender and a receiver is provided. The method comprises semantically analyzing a communication history to form a knowledge graph, deriving formality level values using a first trained ML model, analyzing parameter values of replies to determine receiver impact score, and training a second ML system to generate a model to predict the receiver impact score value. The method also comprises selecting a linguistic expression in a message being drafted, determining an expression intent, modifying the linguistic expression based on the formality level and the expression intent to generate a modified linguistic expression, and testing whether the modified linguistic expression has an increased likelihood of a higher receiver impact score. The method also comprises repeating selecting the linguistic expression, determining the expression intent, modifying the linguistic expression, and testing until a stop criterion is met.

Abstract: An editor service receives a textual input. The editor service provides the textual input to a rule-based grammar checker to obtain a grammar checker result. The editor service also provides the textual input to a machine learning (ML) fluency model that checks the textual input for errors and provides a ML model result. The editor service aggregates the grammar checker result and the ML model result and generates an editor service output based upon the aggregated results. A representation of the editor service result is provided to the client computing system for surfacing through a user interface.

Abstract: A set of candidate edits for a word of a sentence is obtained. Each of the set of candidate edits includes an edit word. Two or more surrounding words that each have a dependency relationship with the edit word are identified in the sentence. At least one of the two or more surrounding words is identified irrespective of their proximity to the edit word. The dependency relationship between the edit word and each of the surrounding words and the set of candidate edits is provided as input to a grammar accuracy prediction model. One or more outputs of the grammar accuracy prediction model are obtained. The one or more outputs indicate grammatical accuracy of each candidate edit from the set in the sentence in view of the dependency relationship with surrounding words. The candidate edit with highest accuracy is selected from the candidate edit set for the sentence.

Abstract: A method and system for performing real-time sentiment modulation in conversation systems is disclosed. The method includes generating an impact table comprising a plurality of sentiment vectors and a plurality of emotion vectors associated with the plurality of sentences. The method further includes generating for each of the plurality of sentences, a dependency vector based on the associated sentiment vector and the associated emotion vector. The method further includes stacking the dependency vector generated to generate a waveform representing variance in sentiment and emotions across words within the plurality of sentences. The method further includes altering at least one portion of the waveform based on a desired emotional output to generate a reshaped waveform. The method further includes generating a set of rephrased sentences associated with the at least one portion, based on the reshaped waveform, the set of sentences, a user defined sentiment output.

Abstract: The present disclosure relates to a method and device for generating a finite state automata for recognizing a chemical name in a text, and a recognition method. According to an embodiment of the present disclosure, the method comprises substituting representation constants of categories of character segments appearing in an organic compound name set into the organic compound name set to obtain a conversion name set; updating the conversion name set based on a conversion name segment which repeatedly appears in the conversion name set; and generating the finite state automata based on the updated conversion name set.

Abstract: A speech-based system includes a local device in a user premises and a network-based control service that directs the local device to perform actions for a user. The control service may specify a first action that is to be performed upon detection by the local device of a stimulus. In some cases, performing the first action may rely on the availability of network communications with the control service or with another service. In these cases, the control service also specifies a second, fallback action that does not rely upon network communications. Upon detecting the stimulus, the local device performs the first action if network communications are available. If network communications are not available, the local device performs the second, fallback action.

Abstract: A speech-based system includes a local device in a user premises and a network-based control service that directs the local device to perform actions for a user. The control service may specify a first action that is to be performed upon detection by the local device of a stimulus. In some cases, performing the first action may rely on the availability of network communications with the control service or with another service. In these cases, the control service also specifies a second, fallback action that does not rely upon network communications. Upon detecting the stimulus, the local device performs the first action if network communications are available. If network communications are not available, the local device performs the second, fallback action.

Abstract: An apparatus links an entity in a first knowledge-graph with a word in a text. The apparatus, based on a number of first-edges coupled to each of first-nodes serving as a transition-source and a number of second-edges coupled to each of second-nodes serving as a transition-destination in the first knowledge-graph, identifies a third-edge to be deleted from edges coupled to a third-node among the second-nodes which has a preset input-order indicating a number of edges that transition to the third-node, and generates a second knowledge-graph by deleting the third-edge from the first knowledge-graph. The apparatus couples first and second nodes which have been coupled to each other by the third-edge in the first knowledge-graph, via a fourth-node to which the first and second nodes are coupled by edges in the second knowledge-graph, and provides the word in the text and the entity linked with the word to a user.

Abstract: Symbol prediction can be implemented using a multi-task system trained for different tasks. The tasks may include a single symbol prediction, symbol category prediction, and symbol subcategory prediction. Categories of symbols can be generated by clustering sets of training data using a clustering scheme.

Abstract: There is disclosed a computer implemented method that includes accessing a dataset having (1) a first set of questions including at least one pair of relational questions that correspond respectively with a pair of binary answers and (2) a second set of questions including at least another pair of relational questions that correspond respectively with a binary answer and a scalar answer. A question answering network is used to compute both a relational loss for the at least one pair of relational questions, and a relational loss for the at least another pair of relational questions. Both the relational loss for the at least one pair of relational questions and the relational loss for the at least another pair of relational questions are optimized, and a neural network model is trained with the optimized relational losses.

Abstract: Techniques are disclosed for determining a target language for a communication session and configuring a language mode of an input method editor (IME) to the target language. An example methodology implementing the techniques includes, by a computing device, detecting a communication to a recipient via a software application running on the computing device, determining a target language for the communication, and configuring a language mode of an input method editor to the target language. The target language may be determined based on an attribute or attributes of the recipient of the communication. In some cases, the target language may be determined based on an attribute or attributes of a contents of a prior communication.

Abstract: Conversational image editing and enhancement techniques are described. For example, an indication of a digital image is received from a user. Aesthetic attribute scores for multiple aesthetic attributes of the image are generated. A computing device then conducts a natural language conversation with the user to edit the digital image. The computing device receives inputs from the user to refine the digital image as the natural language conversation progresses. The computing device generates natural language suggestions to edit the digital image based on the aesthetic attribute scores as part of the natural language conversation. The computing device provides feedback to the user that includes edits to the digital image based on the series of inputs. The computing device also includes as feedback natural language outputs indicating options for additional edits to the digital image based on the series of inputs and the previous edits to the digital image.

Abstract: At least one processor may obtain a document comprising text tokens. The at least one processor may determine, based on a pre-trained language model, word embeddings corresponding to the text tokens. The at least one processor may determine, based on the word embeddings, named entities corresponding to the text tokens; and one or more accuracy predictions corresponding to the named entities. The at least one processor may compare the one or more accuracy predictions with at least one threshold. The at least one processor may associate, based on the comparing, the named entities with one or more confidence levels. The at last one processor may deliver the named entities and the one or more confidence levels.

Abstract: Disclosed is a solution for diagnosing problems from logs used in an application development environment. A random sample of log statements is collected. The log statements can be completely unstructured and/or do not conform to any natural language. The log statements are tagged with predefined classifications. A natural language processing (NLP) classifier model is trained utilizing the log statements tagged with the predefined classification. New log statements can be classified into the plurality of predefined classifications utilizing the trained NLP classifier model. From the log statements thus classified, statements having a problem classification can be identified and presented through a dashboard running in a browser. Outputs from the trained NLP classifier model can be provided as input to another trained model for automatically and quickly identifying a type of problem associated with the statements, eliminating a need to manually sift through tens or hundreds of thousands of lines of logs.

Abstract: A method and a system for agitation detection and response for a programming language are provided. The method includes collecting software code and activity data pertaining to one or more activities performed by a developer that is using a segment of a programming language. The method also includes evaluating the activity data to generate an agitation level of the developer when using the segment of the programming language. The method can also include generating a developer context by evaluating the software code. The developer context can include insights into the operation of features in the programming language by the developer. The activity and developer context can be provided to a software development provider for independent analysis.

Abstract: There is provided a computer implemented method of aligning an automatically generated transcription of an audio recording to a manually generated transcription of the audio recording comprising: identifying non-aligned text fragments, each located between respective two non-continuous aligned text-fragments of the automatically generated transcription, each aligned text-fragment matching words of the manually generated transcription, for each respective non-aligned text fragment: mapping a target keyword of the manually generated transcription to phonemes, mapping the respective non-aligned text fragment to a corresponding audio-fragment of the audio recording, mapping the audio-fragment to phonemes, identifying at least some of the phonemes of the audio-fragment that correspond to the phonemes of the target keyword, and mapping the identified at least some of the phonemes of the audio-fragment to a corresponding word of the automatically generated transcript, wherein the corresponding word is an incorrect a