Abstract: Computer-based processes are disclosed for analyzing and improving document readability. Document readability is improved by using rules and associated logic to automatically detect various types of writing problems and to make and/or suggest edits for eliminating such problems. Many of the rules seek to generate more concise formulations of the analyzed sentences, such as by eliminating unnecessary words, rearranging words and phrases, and making various other types of edits. Proposed edits can be conveyed, e.g., through a word processing platform, by changing the visual appearance of text to indicate how the text would appear with (or with and without) the edit.

Abstract: An information processing apparatus includes an acquiring unit, a detecting unit, and a voice command unit. The acquiring unit acquires voice information of a speaker. The detecting unit detects operation related to speech by the speaker. The voice command unit performs a voice command in accordance with the voice information acquired by the acquiring unit after the detecting unit detects the operation.

Abstract: A system and a method for obtaining a photo-realistic video from a text. The method includes: providing the text and an image of a talking person; synthesizing a speech audio from the text; extracting an acoustic feature from the speech audio by an acoustic feature extractor; and generating the photo-realistic video from the acoustic feature and the image by a video generation neural network. The video generating neural network is pre-trained by: providing a training video and a training image; extracting a training acoustic feature from training audio of the training video by the acoustic feature extractor; generating video frames from the training image and the training acoustic feature by the video generation neural network; and comparing the generated video frames with ground truth video frames using generative adversarial network (GAN). The ground truth video frames correspond to the training video frames.

Abstract: The subject matter of this specification can be embodied in, among other things, a method that includes receiving two or more data sets each representing speech of a corresponding individual attending an internet-based social networking video conference session, decoding the received data sets to produce corresponding text for each individual attending the internet-based social networking video conference, and detecting characteristics of the session from a coalesced transcript produced from the decoded text of the attending individuals for providing context to the internet-based social networking video conference session.

Abstract: Utterances of at least two speakers in a speech signal may be distinguished and the associated speaker identified by use of diarization together with automatic speech recognition of identifying words and phrases commonly in the speech signal. The diarization process clusters turns of the conversation while recognized special form phrases and entity names identify the speakers. A trained probabilistic model deduces which entity name(s) correspond to the clusters.

Abstract: Disclosed herein are example techniques for voice detection by multiple NMDs. An example implementation may involve one or more servers receiving, via a network interface, data representing multiple audio recordings of a voice input spoken by a given user, each audio recording recorded by a respective NMD of the multiple NMDs, wherein the voice input comprises a detected wake-word. Based on respective sound pressure levels of the multiple audio recordings of the voice input, the servers (i) select a particular NMD of the multiple NMDs and (ii) forego selection of other NMDs of the multiple NMDs. The servers send, via the network interface to the particular NMD, data representing a playback command that corresponds to a voice command in the voice input represented in the multiple audio recordings, wherein the data representing the playback command causes the particular NMD to play back audio content according to the playback command.

Abstract: A speech recognition method comprises: generating, based on a preset speech knowledge source, a search space comprising preset client information and for decoding a speech signal; extracting a characteristic vector sequence of a to-be-recognized speech signal; calculating a probability at which the characteristic vector corresponds to each basic unit of the search space; and executing a decoding operation in the search space by using the probability as an input to obtain a word sequence corresponding to the characteristic vector sequence.

Abstract: Systems and methods for using machine learning and rules-based algorithms to create a patent specification based on human-provided patent claims such that the patent specification is created without human intervention are disclosed. Exemplary implementations may: obtain a claim set; obtain a first data structure representing the claim set; obtain a second data structure; obtain a third data structure; and determine one or more sections of the patent specification based on the first data structure, the second data structure, and the third data structure.

Abstract: A method for recovering a current frame of an audio stream includes detecting that a current packet is lost, the current packet including an audio signal; splitting one or more frames into respective high-band signals and respective low-band signals, the one or more frames precede the current frame in the audio stream; inferring a current low-band signal of the current frame using, as inputs to a machine-learning model, the respective low-band signals; combining the inferred current low-band signal with the respective high-band signals to obtain the current frame; and adding the current frame to a playout buffer.

Abstract: Systems and methods of voice activated thread management in a voice activated data packet based environment are provided. A natural language processor (“NLP”) component can receive and parse data packets comprising a first input audio signal to identify a first request and a first trigger keyword. A direct action application programming interface (“API”) can generate a first action data structure with a parameter defining a first action. The NLP component can receive and parse a second input audio signal to identify a second request and a second trigger keyword, and can generate a second action data structure with a parameter defining a second action. A pooling component can generate the first and second action data structures into a pooled data structure, and can transmit the pooled data structure to a service provider computing device to cause it device to perform an operation defined by the pooled data structure.

Abstract: An appliance control apparatus according to an embodiment of the present invention includes a recognition device including a communication unit connected to a speech processing device or a control command input device to receive a control command for an appliance from the speech processing device or the control command input device, and a recognition module configured to recognize a setting environment for the appliance based on the received control command and generate control information for the appliance based on the recognized setting environment, and a control device configured to receive the control information from the recognition device, generate a control signal based on the received control information, and transmit the generated control signal to the appliance or a control command input device connected to the appliance.

Abstract: A speaker recognition system for assessing the identity of a speaker through a speech signal based on speech uttered by said speaker is provided. The system includes a framing module that subdivides the speech signal over time into a set of frames, and a filtering module that analyzes the frames of the set to discard frames affected by noise and frames which do not comprise a speech, based on a spectral analysis of the frames. A feature extraction module extracts audio features from frames which have not been discarded, and a classification module processes the audio features extracted from the frames which have not been discarded for assessing the identity of the speaker.

Abstract: A system for separating audio based on sound producing objects includes a processor configured to receive video data and audio data. The processor is also configured to perform object detection using the video data to identify a number of sound producing objects in the video data and predict a separation for each sound producing object detected in the video data. The processor is also configured to generate separated audio data for each sound producing object using the separation and the audio data.

Abstract: Systems and methods are presented for the automatic placement of rules applied to topics in a logical hierarchy when conducting natural language processing. In some embodiments, a method includes: accessing, at a child node in a logical hierarchy, at least one rule associated with the child node; identifying a percolation criterion associated with a parent node to the child node, said percolation criterion indicating that the at least one rule associated with the child node is to be associated also with the parent node; associating the at least one rule with the parent node such that the at least one rule defines a second factor for determining whether the document is to also be classified into the parent node; accessing the document for natural language processing; and determining whether the document is to be classified into the parent node or the child node based on the at least one rule.

Abstract: Various embodiments of an apparatus, methods, systems and computer program products described herein are directed to a Topic Engine. The Topic Engine captures a plurality of content identifier sequences. Each respective sequence represents an order at which a corresponding user account accessed content. The Topic Engine generates a plurality of clusters. Each cluster is associated with respective content identifiers appearing within a proximity to each other across the plurality of content identifier sequences of different user accounts. The Topic Engine obtains one or more sample content identifiers from at least one cluster via sampling the cluster. The Topic Engine extracts keywords from content represented by the one or more sampled content identifiers. The Topic Engine identifies a topic for the cluster based on the one or more extracted keywords.

Abstract: A system for detection of at least one designated wake-up word for at least one speech-enabled application. The system comprises at least one microphone; and at least one computer hardware processor configured to perform: receiving an acoustic signal generated by the at least one microphone at least in part as a result of receiving an utterance spoken by a speaker; obtaining information indicative of the speaker's identity; interpreting the acoustic signal at least in part by determining, using the information indicative of the speaker's identity and automated speech recognition, whether the utterance spoken by the speaker includes the at least one designated wake-up word; and interacting with the speaker based, at least in part, on results of the interpreting.

Abstract: One aspect of this disclosure relates to presentation of a first effect on one or more presentation devices during an oral recitation of a first story. The first effect is associated with a first trigger point, first content, and/or first story. The first trigger point being one or more specific syllables from a word and/or phrase in the first story. A first transmission point associated with the first effect can be determined based on a latency of a presentation device and user speaking profile. The first transmission point being one or more specific syllables from a word and/or phrase before the first trigger point in the first story. Control signals for instructions to present the first content at the first trigger point are transmitted to the presentation device when a user recites the first transmission point such that first content is presented at the first trigger point.

Abstract: The disclosure provides a speech processing method and a device thereof. The method includes: acquiring a speech sampling signal frame in a mixed-excitation linear prediction (MELP) speech coding system and estimating signal quality of the speech sampling signal frame; determining, based on the signal quality, a specific linear prediction coding (LPC) order used by an LPC circuit; controlling the LPC circuit to convert the speech sampling signal frame into a line spectrum pair parameter based on the specific LPC order; replacing a speech signal spectrum of the speech sampling signal frame with the line spectrum pair parameter to generate a predicted speech signal; and performing a speech coding operation and a signal synthesizing operation of the MELP speech coding system based on the predicted speech signal.

Abstract: The present disclosure discloses a word vector retrofitting method. The method includes obtaining, by a computing device, a first model and a second model that are generated when original word vectors are trained, the first model being configured to predict a context according to an inputted word, and the second model being configured to predict a target word according to a context; inputting a corpus unit from a target corpus into the first model, inputting an output of the first model into the second model, and determining losses generated by the first model and the second model when the second model outputs the corpus unit; and retrofitting the first model and the second model according to the losses.

Abstract: Topics of potential interest to a user, useful for purposes such as targeted advertising and product recommendations, can be extracted from voice content produced by a user. A computing device can capture voice content, such as when a user speaks into or near the device. One or more sniffer algorithms or processes can attempt to identify trigger words in the voice content, which can indicate a level of interest of the user. For each identified potential trigger word, the device can capture adjacent audio that can be analyzed, on the device or remotely, to attempt to determine one or more keywords associated with that trigger word. The identified keywords can be stored and/or transmitted to an appropriate location accessible to entities such as advertisers or content providers who can use the keywords to attempt to select or customize content that is likely relevant to the user.