532 results about "Multiple language" patented technology

The invention discloses a cross-border e-commerce compliance intelligent auditing platform and a multi-language contract analysis method, and relates to the field of cross-border contract compliance auditing. In the multi-modal data access step, customs codes, laws and regulations and other multi-source data are collected, and 18 kinds of language contract texts are analyzed; in the cross-language semantic alignment step, a knowledge graph is constructed, and multi-language legal concept mapping is achieved; in the compliance risk reasoning step, a rule engine and an agent cooperatively check a contract, and the compliance conclusion confidence is calculated; the dynamic risk assessment step adopts an LSTM network to analyze historical data and predict a risk trend; in the multi-language report generation step, a multi-format bilingual or multilingual report is generated based on a template engine, encrypted and archived. According to the invention, cross-border contracts are audited efficiently and intelligently, dynamic adaptation laws and regulations are analyzed in multiple languages, compliance risks are identified accurately, and a multi-language report is generated quickly; therefore, the checking efficiency is improved, the manual workload is reduced, the compliance risk is reduced, and the enterprise cross-border business competitiveness and the risk response capability are enhanced.

Aspects of the present disclosure relate to detecting hallucinations in language model outputs. Embodiments include receiving a user query. Embodiments further include prompting a language processing machine learning model to generate responses to the user query in each language of a set of multiple languages. Embodiments further include receiving the responses from the language processing machine learning model in response to the prompting. Embodiments further include creating embedding representations of the responses. Embodiments further include calculating, based on the embedding representations, a degree of semantic similarity between the responses. Embodiments further include determining that a response of the responses contains a model hallucination based on comparing the degree of semantic similarity between the responses to a threshold.

In various examples, synthesizing speech in multiple languages in conversational AI systems and applications is described herein. Systems and methods are disclosed that use one or more models to synthesize speech from a first language spoken by a speaker to a second, target language selected by the speaker. In some examples, to perform the translation, the model(s) may disentangle one or more attributes associated with speech from speakers, such as speakers' identities, speakers' accents, and text associated with the speech. Additionally, the model(s) may allow for fine-grained control of additional attributes associated with output speech, such as one or more frequencies, one or more energies, and one or more phoneme durations. Furthermore, the model(s) may be configured to use the accent associated with the target language when generating text, such as when aligning text encodings with one or more phonemes.

The invention belongs to the technical field of document content extraction, and particularly discloses a document content extraction method and system based on multimodal model collaboration, a terminal and a medium. Comprising the following steps: identifying the type of an input to-be-processed document, and judging the document type; on the basis of the type identification result, calling a multi-modal model to analyze the document content, and outputting space coordinates, visual features and semantic features of document elements; generating a content sequence according with a reading habit through a semantic sequence reconstruction algorithm; paragraph boundary detection, paragraph recombination and semantic association modeling of charts and texts are completed based on the multilayer attention network and the graph neural network; grammar error correction, format optimization and title hierarchy generation are carried out by using a large language model and a hierarchical classification network; and converting the identification result into a structured output file. According to the method, the processing requirements of different types of documents can be considered, and high-precision analysis and efficient output are realized under the scenes of complex layouts, multiple languages and formula tables.

The invention discloses a multi-language knowledge graph construction method, system and device and a medium. The method comprises the steps of obtaining text data, voice data and image data; performing standardization processing on the text data, the voice data and the image data to obtain multi-source heterogeneous data after format unification; mapping the multi-source heterogeneous data with the unified format into a unified semantic representation space, and performing semantic annotation to generate a corresponding semantic annotation result; the embedding representations of different languages are aligned based on a comparative learning technology, the embedding of multiple languages is mapped to a unified semantic space, and a semantic annotation result is updated; and constructing a knowledge graph based on the updated semantic annotation result, and updating the knowledge graph in real time based on an incremental updating algorithm. According to the technical scheme, precise analysis of complex semantics and efficient annotation of unseen knowledge can be achieved, dynamic real-time updating of the knowledge graph is achieved, and the response speed and knowledge updating efficiency of a system are remarkably improved.

A method of leveraging improved conversational bots in a contact center system according to an embodiment includes performing knowledge distillation through machine learning to teach a student artificial intelligence model based on a teacher artificial intelligence model and reduce a size of an initial multilingual vocabulary, wherein the student artificial intelligence model includes fewer machine learning embedding layers than the teacher artificial intelligence model, removing tokens from the initial multilingual vocabulary based on a grouping of languages with linguistic similarities to reduce the size of the initial multilingual vocabulary, parsing user text from a human user into one or more tokens, identifying token indexes associated with the respective one or more tokens in a reduced multilingual vocabulary, determining embedding values associated with the identified token indexes, and generating a multilingual embedding output for the user text indicative of user intent based on the embedding values using machine learning.

Systems and methods for generating speaker video and audio in multiple languages for videoconferencing are provided. For example, a computing device can access a speaker speech audio signal that includes a speaker speech in a first language, a video of the speaker and a translated speech audio signal of the speaker speech in a second language. The computing device generates, based on the translated speech audio signal, a converted translated speech audio signal that includes a speech in the second language having voice characteristics in the speaker speech. The computing device further generates a lip-synched speaker video based on the video of the speaker and the converted translated speech audio signal. Lip movements in the lip-synched speaker video correspond to the converted translated speech audio signal. The converted translated speech audio signal and the lip-synched speaker video are transmitted to a video conference provider configured to host the video conference.

Disclosed are apparatuses, systems, and techniques that leverage one or more language models (LMs)—such as large language models (LLMs—for efficient multi-speaker speech recognition. The techniques include processing, using a speaker diarization model, an audio feature to generate a first association of the audio feature with one or more prospective speakers, the audio feature being representative of one or more spoken words. The techniques further include providing, to an LM, a first prompt requesting the LM to identify a second association of the one or more spoken words with the one or more prospective speakers and receiving, from the LM, a first response identifying the second association of the one or more spoken words with the one or more prospective speakers. The techniques further include determining, using the first association and the second association, one or more speakers that produced the one or more spoken words.

Methods, systems, and computer storage media for providing speech synthesis using a code-mixed speech engine in a speech synthesis system. A code-mixed speech engine supports generating natural and intelligible speech in a target speaker voice—for code-mixed-text of two or more languages—based on a code-mixed speech model that supports both code-mixing and cross-locale voice transfer scenarios. In operation, code-mixed training data associated with a plurality of different languages is accessed. A code-mixed speech model—associated with a training engine and an inference engine that support generating code-mixed synthesized speech—is generated. The code-mixed speech model is deployed. A request being received for synthesized speech of a speech synthesis service. An instance of code-mixed synthesized speech is generated. The instance of code-mixed synthesized speech is generated using the code-mixed speech model. The instance of code-mixed synthesized speech is communicated for output on an interface associated with the speech synthesis service.

To provide a system that extracts useful information from text data and efficiently performs translation and information retrieval between multiple languages.SOLUTION: A system according to the embodiment comprises an analysis unit, an extraction unit, a translation unit, an information search unit, and a provision unit. The analysis unit extracts information from text data, and performs topic modeling and emotion analysis. The extraction unit summarizes the information extracted by the analysis unit, and extracts main points. The translation unit translates the text data extracted by the extraction unit between multiple languages. The information search unit searches for and acquires the text data translated by the translation unit in response to input from a user. The provision unit provides the information acquired by the information search unit.SELECTED DRAWING: Figure 1

One example method includes receiving an audio stream comprising speech; generating, by automatic speech recognition (“ASR”) software, a plurality of hypotheses, each hypothesis comprising a transcription of a first portion of the speech; rescoring, using a first trained language model, each hypothesis of the plurality of hypotheses; and responsive to a first hypothesis not satisfying a threshold, generating and outputting, using a trained large language model (“LLM”), a final transcription based on the plurality of hypotheses.

Embodiments are provided for building a configurable multilingual model. A computing system obtains a plurality of language-specific automatic speech recognition modules and a universal automatic speech recognition module trained on a multi-language training dataset comprising training data corresponding to each of the plurality of different languages. The computing system then compiles the universal automatic speech recognition module with the plurality of language-specific automatic speech recognition modules to generate a configurable multilingual model that is configured to selectively and dynamically utilize a sub-set of the plurality of language-specific automatic speech recognition modules with the universal automatic speech recognition module to process audio content in response to user input identifying one or more target languages associated with the audio content.

The embodiment of the invention discloses a voice recognition method and device, electronic equipment and a storage medium. The method comprises the steps that to-be-recognized voice is acquired; recognizing the voice to be recognized through the mixed voice recognition model to obtain a voice recognition result; the hybrid speech recognition model comprises a target coding module, a hybrid expert model MoE module and a target decoding module, the MoE module comprises a routing network and a plurality of expert network groups in one-to-one correspondence with multiple language types, and the routing network is used for carrying out target coding according to initial speech characterization; and determining a target expert network group corresponding to the initial voice representation from the plurality of expert network groups, and processing the initial voice representation through the target expert network group to obtain a target voice representation. According to the embodiment of the invention, the method improves the recognition accuracy of the mixed voice containing a plurality of languages.

The invention provides a double-screen different-display real-time translation machine implementation method and device, and the method comprises the following steps: 1, obtaining and starting a double-screen different-display real-time translation machine, and initializing an SOC SSD2386 chip, two touch screens, a plurality of microphones distributed near the two touch screens and a storage unit of the double-screen different-display real-time translation machine; by adopting a double-screen different display technology, the target language text information can be respectively and clearly displayed on the touch screen of the opposite user, so that the situation of image disorder caused by mixed display of multiple languages on the same screen is avoided, the information reading speed and accuracy of the user are greatly improved, and the communication fluency and efficiency are ensured; according to the method, the target language text information is individually set according to the display preference of the user, the font size, the color and the like are adjusted, and only the translation result required by the opposite user is displayed, so that the interference of irrelevant information on the user is effectively avoided, the user experience is improved, and the smooth communication is promoted.

According to the feature fusion speech synthesis method and device based on the VITS, the expressive force and naturalness of speech can be improved, high-quality speech which is more natural, smoother and rich in expressive force is generated, and the expression is particularly outstanding in speech synthesis of multiple languages, multiple emotion expressions and complex contexts. The method comprises the following steps: (1) generating rhythm features matched with input text information according to the input text information; (2) establishing a deep learning model based on Transform; (3) comprehensively processing the rhythm features and the text features by adopting a multi-modal feature fusion method; (4) inputting the comprehensive feature vector into a speech synthesis network based on a variational auto-encoder for processing, wherein the speech synthesis network is trained by adopting a generative adversarial network and a training method based on maximum likelihood; and (5) adopting a neural network model of a random time length predictor, and performing input data transformation through a multi-layer normalized flow structure.

The invention relates to a method for improving the internal accuracy of a response from an artificial intelligence (AI) trained in multiple languages, in particular a generative AI trained in multiple languages. The invention further relates to a data processing system for carrying out the method.

A document information extraction service can utilize an LLM to provide support for extracting information from documents in multiple languages. A trained first machine learning model can map data extracted from a first master document of a first document type in a first language. The mappings can be corrected via user input to obtain ground truth data for the master document. The ground truth data can be translated into a second language and optionally corrected to obtain translated ground truth data. An LLM can generate a training dataset of fake documents of the first document type that contain text in the second language based at least in part on the translated ground truth data. The trained first machine learning model can be trained further with the training dataset and deployed to extract data from documents of the first document type that contain text in the second language.

To provide a multilingual voice distribution method capable of translating a voice uttered by a speaker into other languages, minimizing the occurrence of a time lag accompanying the translation, and transmitting the voice to a plurality of recipients without delay. [Solution] Speech uttered by a speaker is input to a speech terminal 1 and transmitted with a PTT (Press to Talk) button held down. The speaker's speech is input to a translation engine 3 and converted to text by a text conversion unit 3a. The text converted from the speech is translated into multiple languages ​​by a multilingual translation unit 3b and converted into speech in each language by a speech synthesis engine unit 3c for each language. The converted multilingual speech is input to a speech distribution server 4, and the multilingual speech is distributed in a preselected language by a speech distribution unit 4a, and the translated and distributed speech is received by a listening terminal 5 and transmitted to the listener.

The invention relates to the technical field of artificial intelligence, provides a multi-language task execution method, device and equipment and a medium, is applied to financial and medical health care service scenes, and can acquire text data and visual action data of multiple languages according to an execution instruction and perform preprocessing to realize standardized processing of multi-modal data; constructing a cross-language word vector semantic space based on adversarial training according to the multi-language features to realize preliminary word vector alignment; performing multi-language semantic alignment on the multi-language features according to a cross-language word vector semantic space, further breaking language barriers, and realizing depth mapping and alignment among different language semantics; a multi-language culture knowledge graph is constructed, so that culture knowledge is introduced, and the culture perception ability is improved; fusion is carried out through a gating fusion mechanism, fusion features with language attributes and cultural attributes can be obtained, and therefore multi-language tasks can be executed more accurately by integrating multi-language information and cultural knowledge.

To provide a system capable of highly accurate translation in real time and coping with a plurality of languages.SOLUTION: A data processing system includes a data processing device 12 and a smart device 14. In specific processing by a specific processing section 290 of the data processing device 12, when a user plays a moving image, a microphone acquires voice data of the moving image in real time, the acquired voice data is transmitted to a voice recognition module, which converts the data into text data, the converted text data is transmitted to a translation module, the translation module acquires data on a necessary language from a language database to translate the data into a designated language, the translated text data is output to a display to present the text data to the user.SELECTED DRAWING: Figure 2

The invention discloses an automatic test input generation method based on dynamic page splitting, and belongs to the technical field of dynamic webpages. In order to effectively verify the semantic consistency of a page in a complex Web application supported by multiple languages, the method comprises the following steps: acquiring a dynamic webpage to be tested, and analyzing a document object model structure of the dynamic webpage based on a Cypress framework; the monitored dynamic content elements are classified and marked in real time by using the real-time DOM operation capability of Cypress; on the basis of the structural information of the page components, in combination with integration of Cypress and GraphQL API, dynamically obtaining test data from a background, and automatically generating a test input template matched with the current state of the page; the method is applied to dynamic page testing in real time; generating a test report; a natural language processing module is introduced, and text content in the page is analyzed; and optimizing a page splitting strategy, an input generation rule and GraphQL query configuration.

A system and method for classification. A method includes identifying candidate entities among text data by applying at least one entity identification rule to the text data. Inputs are constructed based on the identified candidate entities, where each input includes a first portion of text indicating a candidate entity and at least one second portion of text and where the at least one second portion of text of each input is adjacent to the first portion of text of the input. Multiple language models are applied to the inputs, where each language model is trained to identify a respective set of entities and where outputs of the language models include at least one portion of entity-indicating text for each input. Based on the outputs of the language models, at least one named entity in the text data is determined.

Various embodiments disclosed relate to a method of producing an analysis document using generative artificial intelligence, the method comprising: maintaining a database containing multiple language learning models and a collection of prompts; generating a patent prosecution tool user interface integrated with a generative artificial intelligence program; receiving a user input comprising a matter associated with at least one matter dataset, where the matter dataset includes a plurality of patent documents associated with the matter; selecting a language learning model from the database; retrieving a plurality of patent documents for display on the patent prosecution tool user interface; loading at least one of the plurality of patent documents into the artificial intelligence program; displaying more than one prompt from the collection of prompts; initiating a script in the generative artificial intelligence program based on a user selected prompt and the at least one patent document; analyzing the patent document according to the script to produce a preliminary analysis; producing the analysis document according to the preliminary analysis; saving the analysis document.

The invention particularly relates to a code writing, checking and modifying method based on a large model. According to the code writing, checking and modifying method based on the large model, a fine-tuning large language model is adopted to analyze user requirements, and multi-language initial codes are generated; executing deep code review through a static analysis engine, wherein the deep code review comprises syntax tree verification, security vulnerability mode matching and API dependency chain detection; executing the automatically generated test case set in an isolation environment, monitoring operation indexes in real time, and generating a defect thermodynamic diagram; and driving the large model to perform iterative code reconstruction based on a defect analysis result until it is confirmed that a preset quality standard is met. The code writing, checking and modifying method based on the large model is suitable for multiple languages, the defect rate of codes is reduced, the software development period is prolonged, the labor cost is reduced, and the vulnerability detection rate is greatly increased.

The present disclosure provides a method of extracting information from an image of a document in which the document image is properly aligned to be processed, the regions containing the desired information are detected and extracted from the document image, a text machine-learning model is performed in which the handwritten text in multiple languages may be extracted and stored, and a user may review, edit, and translate the extracted information to create a standardized digital format of the information contained in the document.

The invention provides a multilingual word alignment method based on an attention mechanism, and belongs to the technical field of natural language processing. The multilingual word alignment method based on the attention mechanism comprises the following steps: data collection: collecting a large amount of multilingual text data of Cantonese, Chinese and English; text representation: modeling the text; constructing a large model: adjusting and optimizing the large model; multilingual pre-training: pre-training the large model by using multilingual data; an attention mechanism: introducing a multilingual attention mechanism, and calculating attention scores among different language texts; according to the attention score, the alignment relation between the words is determined; by introducing a multilingual attention mechanism, the semantic relationship between different language texts can be better captured, so that the alignment accuracy is improved, a pre-trained large model capable of processing multiple languages of Cantonese, Chinese and English is realized, and multiple languages are processed at the same time.

The present invention facilitates communication between sign language users and machines by translating sign language and text using AI models, deep learning computer vision, and word embeddings. Users interact via sign language, captured and processed through deep learning and NLP modules. The system converts sign language videos into text, constructs coherent sentences, and generates contextually appropriate responses using a Retrieve and Generate (RAG) model. Responses are translated back into sign language videos, spelling out words not found in the dictionary. If requested, a human agent can respond. Key features include high-accuracy recognition, context-aware response generation, dynamic vocabulary updates, and optional human interaction. The method ensures efficient processing with LLM, embedding techniques, and deep learning, optimizing translation accuracy and user experience. The system adapts to multiple languages and dialects by training on specific sign languages, making it applicable globally.

The invention discloses an uncertainty guided few-sample harmful speech detection method (U-GIFT). According to the method, a pre-training language model is finely adjusted based on a small number of labeled samples, and a semi-supervised self-training and uncertainty guiding strategy is combined. Monte Carlo Dropout is started in the reasoning stage, multiple times of random forward propagation are carried out to obtain sample posterior distribution, prediction entropy and information gain are calculated, pseudo-label samples are sorted and screened, and only high-confidence samples are selected to be added into a training set. And in order to reduce the influence of a pseudo labeling error, designing a stability weighting mechanism, giving a sample weight according to a prediction variance, and constructing a joint loss function, so that the model preferentially learns a stable sample to improve the detection performance. According to the method, the semantic and attention mechanism of the pre-training model is utilized, the detection effect is remarkably improved under the conditions of few samples, imbalance, multiple languages and cross domains, models such as BERT, RoBERTa, XLM-R, LLaMA2 and DeepSeek-R1 are compatible, and the method is suitable for content auditing and risk prevention and control.

The invention relates to a memory retrieval method, a memory generation method and related devices. The method comprises the following steps: acquiring an input search word, and translating the input search word into a search word of a target language; performing memory retrieval by using the input retrieval word and the retrieval word of the target language according to the memory index to obtain a retrieval result; wherein the memory index is established according to at least one of an original and a copy of the memory information; the copy is obtained by translating the original into a target language. The method can match and support multiple languages.

The invention discloses a text dialogue method and device suitable for multiple languages, terminal equipment and a storage medium, after a source text input by a user in a current dialogue is obtained, a target language corresponding to the source text can be accurately recognized based on a phrase structure and a sentence structure of the source text, and therefore the user can select the target language based on the recognized target language. And a corresponding grammar rule can be selected to generate a subsequent reply text. In addition, the user intention and context information can be deeply understood based on the extracted local features, global features and dialogue features, so that the reply text conforming to the language expected by the user is generated, and the generated reply text is correct in grammar based on the language of the current input text and the grammar rule of the language. And the expression is natural and smooth, the stiff or incoherent condition is avoided, the accuracy and naturalness of the reply text in the dialogue process are improved, the multilingual text dialogue is realized, and the dialogue experience of the user is optimized.