Abstract: Embodiments of the present invention provide an approach for compressing data, and more particularly, to large-scale text data encoding and compression using absolute overfitting on pre-trained language models. Large-scale data is parsed into sentences. A unique token is generated for each sentence to form a token list. A generative (or compression) model is trained from the tokens in the token list to produce the corresponding sentence of each token through absolute overfitting of a pre-trained language model. The compressed text data is stored as the token list and generative model, resulting in a storage space savings.

Abstract: The disclosure relates to systems, methods, and computer-readable media for an artificial intelligence (AI) conversational agent configured for enhanced user interaction in multi-channel contact centers. An example method includes receiving a query regarding a particular property of an asset during a real-time conversation, classifying an intent of the query into a predefined category, triggering, based on the classifying of the intent, a finetuned large language model for the predefined category to answer the query, generating, by the finetuned large language model, a response to the query based on at least one of an information data model and a knowledge graph data store, and outputting the response to the query.

Abstract: Some disclosed embodiments are directed to obtaining a decoded audio data including a spoken language utterance recognized in audio data and identifying a disfluency in the decoded audio data. Upon determining that correcting the disfluency would improve a readability score of the decoded audio data, the system generates a particular correction to correct the disfluency and applies the particular correction to the decoded audio data. Then, an updated decoded audio data is generated which reflects the particular correction. The updated decoded audio data has improved readability over the decoded audio data.

Abstract: The disclosure relates to an electronic device and a control method thereof. The electronic device includes a memory, and a processor configured to: obtain first feature data for estimating a waveform by inputting acoustic data of a first quality to a first encoder model; and obtain waveform data of a second quality that is a higher quality than the first quality by inputting the first feature data to a decoder model to.

Abstract: A conversion table generation device includes a similar word extraction unit and a conversion table generation unit. The similar word extraction unit is configured to extract similar words similar to first words, for each of the first words included in a word group used in a dialogue. The conversion table generation unit is configured to associate any one of the first words with the extracted similar words, that are similar to the plurality of first words, as second words, on the basis of the priority, and generates a conversion table for voice recognition with the second word as a conversion source and the first word as a conversion destination.

Abstract: A second wake word detector, at a media-playback device, that plays audio (or other) content to a device, such as a voice-enabled device, detects false wake words in the audio content. The second wake word detector analyzes the audio stream to determine if the audio stream contains any audio that sounds like the wake word. If so, the second wake word detector can generate one of a plurality of instructions that describes the time period, within the audio content, in which the false wake word was encountered. The instruction can cause a first wake word detector to assume one of a plurality of configurations. The media-playback device can then instruct or inform the voice-enabled device of the presence of the false wake word. In this way, the wake word detector, at the voice-enabled device, is not activated to receive the false wake word or ignores the wake word.

Abstract: As part of a dialog session between a user and an automated assistant, implementations can process, using a streaming ASR model, a stream of audio data to generate ASR output, process, using an NLU model, the ASR output to generate NLU output, and generate, based on the NLU output, a stream of fulfillment data. Further, implementations can further determine, based on processing the stream of audio data, audio-based characteristics associated with spoken utterance(s) captured in the stream of audio data. Based on a current state of the stream of NLU output, the stream of fulfillment data, and the audio-based characteristics, implementations can determine whether a next interaction state to be implemented is: (i) causing fulfillment output to be implemented; (ii) causing natural conversation output to be audibly rendered; or (iii) refrain from causing any interaction to be implemented, can cause the next interaction state to be implemented.

Abstract: Technologies relating to construction of a product ontology based upon queries in a search engine log are described. Candidate phrases are extracted from queries in the search engine log. The candidate phrases are partitioned to generate candidate product classes, where the candidate phrases are partitioned by sequentially expanding the ending phrases of the candidate product phrases. Using such an ending phrase-based approach naturally forms a hierarchy of candidate product classes. Embeddings are generated for the candidate product classes, and the candidate product classes are clustered based upon the embeddings such that semantically equivalent product classes are merged.

Abstract: Devices and techniques are generally described for generating user satisfaction data in a natural language processing system. In various examples, first user input data may be received by a natural language processing system. Behavioral data related to the first user input data may be determined. Natural language processing error data related to the first user input data may be determined. First response data corresponding to the first user input data may be determined. First response characteristic data related to the first response data may be determined. First user satisfaction data may be determined based at least in part on the first response characteristic data, the behavioral data, and the natural language processing error data.

Abstract: An entity-based platform stores a database of entity mentions and related, entity-based information obtained by processing text in a corpus of documents such as content from a public network or content from a document management system. The platform may analyze changes to a previously-processed document in order to align mentions in the modified document to mentions in a prior document and/or to identify new mentions not present in the previously-processed document. As the underlying documents in the corpus change over time, this advantageously permits sustained use of knowledge graphs and other entity-based representations or analysis linked to specific entity mentions.

Abstract: Techniques are disclosed for predicting a table column using machine learning. For example, a system can include at least one processing device including a processor coupled to a memory, the processing device being configured to implement the following: determining a local word density for words in a table, the local word density measuring a count of other words in a first region surrounding the words; determining a local numeric density for the words, the local numeric density measuring a proportion of digits in a second region surrounding the words; determining semantic associations for the words by processing the words using an ML-based semantic association model trained based on surrounding words in nearby table columns and rows; and predicting a table column index for the words by processing the table using an ML-based table column model trained based on the local word density, local numeric density, and semantic association.

Abstract: A system for establishing a dental treatment environment, includes: a head-mounted device provided at a dental clinic to be mounted on a patient's head, the head-mounted device having an image display unit and an ear-mounted speaker, a microphone for converting a sound including the voice of the medical staff in charge of the patient into an electric signal; a voice recognition module for recognizing the voice of the medical staff in charge from the electric sound input from the microphone; a content module storing multiple image contents for relaxing the patient mentally physically; a user interface having a content selection unit configured such that the patient can select a play content provided to the image display unit from the multiple image contents; and an output signal generating module for generating an output signal that is output to the head-mounted device.

Abstract: A system for retrieving patient information using large language models including a computing device configured to receive a natural language query as a function of a user input, input the natural language query into a large language model communicatively connected to the least a processor, receive a computer language query comprising a plurality of nodes from the large language model, map the plurality of nodes to one or more entries in a patient database, receive a database response from the patient database as a function of the mapping, generate a final database query as a function of the database response. query the patient database using the final database query, receive a user response as a function of the final database query, and transmit the user response to a graphical user interface as a function of the final database query.

Abstract: An electronic apparatus includes an interface configured to connect with an external apparatus, and a processor. The processor is configured to, in response to a first user speech received by the electronic apparatus including a trigger word, identify a first noise level corresponding to the first user speech received by the electronic apparatus. The processor is configured to identify a first recognition apparatus among a plurality of recognition apparatuses having a highest first noise level corresponding to the first user speech. The plurality of recognition apparatuses identify the first user speech as the trigger word and include the electronic apparatus and the external apparatus. The processor is configured to perform a control operation to implement a function corresponding to a second user speech in response to identifying a second recognition apparatus as having a highest second noise level corresponding to the second user speech among the plurality of recognition apparatuses.

Abstract: A device receives document information associated with a document, and receives a request to identify insights in the document information. The device performs, based on the request, natural language processing on the document information to identify words, phrases, and sentences in the document information, and utilizes a first machine learning model with the words, the phrases, and the sentences to identify information indicating abstract insights, concrete insights, and non-insights in the document. The device utilizes a second machine learning model to match the abstract insights with particular concrete insights that are different than the concrete insights, and utilizes a third machine learning model to determine particular insights based on the non-insights. The device generates an insight document that includes the concrete insights, the abstract insights matched with the particular concrete insights, and the particular insights determined based on the non-insights.

Abstract: A meeting managing method, a server and a meeting managing system, and relates to the technical field of computers. The method according to the present application includes receiving data of a plurality of participants sent by a first client; binding data of each participant in the plurality of participants with a corresponding electronic-table-card identifier, wherein electronic table cards corresponding to each of the electronic-table-card identifiers correspond to microphones one to one; receiving the voice data sent by the target microphone; according to a target-electronic-table-card identifier corresponding to the target microphone, determining data of the target participant; and according to the voice data and the data of the target participant, generating a meeting document.

Abstract: According to an embodiment, an electronic apparatus includes a plurality of service engines corresponding to a plurality of speech recognition services, respectively; a communication module connected via wireless communication to a plurality of service servers for providing the plurality of speech recognition services, respectively; a mike configured to receive speech of a user; and a service controller configured to store at least one wakeup word corresponding to each of the plurality of speech recognition services in advance, determine a speech recognition service corresponding to a wakeup word received by the mike from among the plurality of speech recognition services based on the plurality of wakeup words stored in advance, send a user command received by the mike to a service engine corresponding to the determined speech recognition service among the plurality of service engines.

Abstract: A system and method for interacting with a voice-assisted member interface hosted by a provider backend server of a provider using a voice enabled-apparatus hosted by an apparatus vendor separate and distinct from the provider, the voice-enabled apparatus including a microphone unit, a speaker and a processor coupled to the microphone unit and the speaker, the processor configured to cause the voice-enabled apparatus to perform one or more functions in response to audio signals received at the microphone unit.

Abstract: According to various embodiments, provided is a control operation comprising: receiving a first utterance from a device; determining a specific category corresponding to the received first utterance from among a plurality of categories; determining a first domain for processing of the first utterance from among a plurality of domains related to the determined specific category; acquiring information enabling an operation corresponding to the first utterance to be performed based on the determined first domain; receiving a second utterance from the device; identifying whether processing of the second utterance is possible in the determined first domain; based on identifying that the processing of the second user utterance based on the first domain is not possible, checking a second domain among the plurality of domains, and identifying whether processing of the second utterance based on the second domain is possible; and based on identifying that the processing of the second utterance based on the second domai

Abstract: Systems and methods for pre-training a dialogue model with semantic information include: generating a dialogue-level abstract meaning representation (AMR) graph for an input dialogue associated with a speaker, learning core semantic units of the input dialogue based on nodes of the dialogue-level AMR graph, learning semantic relations between words of a sentence of the input dialogue based on edges of the dialogue-level AMR graph, learning an overall agreement of the input dialogue and the dialogue-level AMR graph, and training the dialogue model based on the learned core semantic units, semantic relations between words, and overall agreement.