650 results about "Phrase" patented technology

Various embodiments discussed herein are directed to improving existing technologies by detecting a likelihood of hallucination arising from one-shot, few-shot, or outside knowledge contexts. For example, regarding the one-shot or few-shot contexts, some embodiments determine a set of tokens in a language model output that are not found in target content but are found in at least one example. When such phrases are not very common words, this is highly indicative that the model is hallucinating because these phrases should be located in the target content but are not, but are instead located in the examples.

Techniques for document recommendation based on conversational log for real time assistance are described. A first machine learning module identifies key phrases of a conversational log in real time. The first machine learning module executes multiple machine learning models trained to determine a probability that a portion of a conversation includes a key phrase. A second machine learning module identifies assistance pertaining to the identified key phrases of the conversational log. The second machine learning module executes a machine learning model trained to identify semantic similarity and word matching features of embedding representations of the key phrases and a knowledge base of assistance. The assistance is provided to a user during a conversation in real time.

Systems and methods are described for obtaining accurate responses from large language models (LLMs) and chatbots, including for question and answering, exposition, and summarization. These systems and methods accomplish these objectives via use of noun phrase avoiding processes such as a noun phrase collision detection process, a query splitting process, and a topical splitting process as well as by use of formatted facts, formatted fact model correction interfaces (FF MCIs), bounded-scope deterministic (BSD) neural networks, processes and methods, and intelligent storage and retrieval (ISAR) systems and methods. These systems and methods avoid and bypass noun phrase collisions and correct for errors caused by noun phrase collisions so that hallucinations are eliminated from LLM responses.

The invention relates to the technical field of artificial intelligence and the field of financial science and technology, and discloses a method for adjusting the playing speed of voice content in a video, and the method comprises the steps: obtaining a video file containing to-be-adjusted voice content, carrying out the voice recognition of an audio track of the video file, and converting the audio data of the voice content into text data; dividing the text data into a plurality of sentences or phrases; identifying keywords in each sentence or phrase, performing emotion content classification on the sentences or phrases containing the keywords, and determining the importance level of each sentence or phrase; according to the emotion content and the importance level, adjusting the playing speed of the corresponding voice content, and generating adjusted audio data; aligning the adjusted audio data with the video image to generate synchronous audio and video data; and a final video is generated. Through automatic voice recognition, sentiment analysis and playing speed adjustment, the steps of manual participation are reduced, and the video editing efficiency is greatly improved.

A system and method for generating question-answer pairs is disclosed. The system and method can receive a document. A sentence and / or a further sentence in the document may be identified. A syntactic map for the sentence and / or the further sentence may be generated. Noun phrases and prepositional phrases may be identified based on the syntactic map. Sentence level questions may be generated based on phrases identified using natural language processing (NLP) techniques. Document level questions can also be generated based on syntactic maps generated and NLP techniques.

A system is configured to: (a) analyze data records using word embeddings to produce vectors indicative of similarity scores for words or phrases in the data records relative to electronically stored reference data; (b) compute shapeley additive explanation (SHAP) values to analyze and rank contributions of the words or phrases to the similarity scores from the word embeddings; (c) use a language model trained on a vocabulary used in the data records to identify contextual insights that are not explicitly disclosed in the data to improve accuracy of the word embeddings; and (d) iteratively and dynamically adjust or update the model using human-in-the-loop based reinforcement learning from human feedback by: receiving an input representative of a modification to at least one of the contributions determined by the SHAP computation, and responsive to receiving the input, adjusting at least one of the ranks of the word embeddings in the model.

Disclosed herein are apparatus, system, method, and computer-readable medium aspects for extracting themes from textual data. Textual data is initially cleansed from its submitted form into a simplified form for improved accuracy of topic extraction. From the cleansed text, phrases are extracted. Embeddings of the phrases are then determined so that similarities can be identified between different phrases within the text. Using these embodiments, clustering is performed on the embeddings to reveal the topics included within the text submission, as well as their frequency and relationship to one another. This clustering processing can be repeated at multiple levels of granularity for improved accuracy. Based on an analysis of the resulting clusters, a graphical representation of the clusters at the various levels is generated to provide an easy-to-understand indication of the body of text and the topics and themes included therein.

The invention relates to the technical field of natural language processing, in particular to an AI reading management and control method and system based on artificial intelligence, and the method comprises the following steps: carrying out text word segmentation processing on an input original text, segmenting the text into independent sentences, recognizing basic word units in each sentence, analyzing semantic adjacency relationships and syntactic structure features among vocabularies, and carrying out word segmentation processing on the word units; screening and extracting potential phrases representing paragraph meanings, and establishing a candidate semantic unit set; according to the method, through word segmentation, syntactic structure recognition and semantic adjacency analysis of the original text, potential phrases capable of representing paragraph significance are extracted, the candidate semantic unit set is constructed, and modeling of the semantic structure in the text is achieved. And associating the set with the reading fixation duration and the playback action of the user sentence by sentence to obtain a reading behavior response of each semantic fragment, and executing semantic weighting and reading behavior cross analysis according to the reading behavior response. Through the linkage mode, the semantic focus actually focused by the user at present can be recognized.

There is provided an automated system for extracting data from a set of documents corresponding to a vendor, contractor, or product. Documents may be converted to text-based documents and text may be extracted therefrom based on keywords and phrases. Extracted sentences may be mapped to requirements for an assessment. The documents may be processed in batches to determine whether the extracted text is sufficient to perform the assessment. The assessment may be performed and output a score indicating the degree to which the vendor, contractor or product complies with requirements. Processing of documents, as well as processing of evaluations, and processing of multiple assessments, may be performed in parallel.

A method includes preparing key actions in a process design document (PDD) from historical PDD documents by tokenizing the PDD with part-of-speech (POS) grammar against key verbs and phrases, extracting automation unit pairs from the historical process documents to map the automation unit pairs with the key verbs and phrases, constructing a set of automation tasks against one or more of the key verbs and phrases, coupling different process models with domain verbs for the phrases to mark one or more of the automation unit pairs as actions against one or more of the key verbs, constructing a tree-like mapping where each key action points to different actions in a target specific automation language, and defining a vocabulary interface of a domain specific language (DSL) by using the key actions and phrases.

Electronic discovery using predictive filtering is disclosed herein. An example method includes receiving a plurality of documents. A selection of a filter value is received, where the filter value comprises a field value or a set of field values for the plurality of documents. A prediction of responsive phrases, responsive concepts, or other meta-data is generated. The plurality of documents is evaluated based on a new filter value. Then, a prediction of other responsive phrases or other responsive concepts is generated. A predictive value is generated. Filter criteria is generated based on the predictive value for each of the other responsive phrases, the other predicted responsive concepts, or other predictive meta-data. A selection is then received of the filter criteria. A filter is built and applied based on the selection, and documents are generated from at least a sub-portion of the received plurality of documents.

Systems and methods for natural language processing are described. One or more aspects of a method, apparatus, and non-transitory computer readable medium include receiving a text phrase; encoding the text phrase using an encoder to obtain a hidden representation of the text phrase, wherein the encoder is trained during a first training phrase using self-supervised learning based on a first contrastive loss and during a second training phrase using supervised learning based on a second contrastive learning loss; identifying an intent of the text phrase from a predetermined set of intent labels using a classification network, wherein the classification network is jointly trained with the encoder in the second training phase; and generating a response to the text phrase based on the intent.

The invention belongs to the technical field of large model application, and particularly relates to a semantic enhancement text matching method and system based on an achievement transformation large model. The semantic enhancement text matching method based on the achievement transformation large model comprises five parts, namely preparation of data and knowledge, enhancement of multi-level semantic knowledge, learning of semantic difference, intelligent weight distribution of semantic information and prediction of text similarity. The preparation of data and knowledge is used for constructing basic data and knowledge resources required by a text matching task through text preprocessing and knowledge base construction. And enhancement of multi-level semantic knowledge: enriching semantic representation of the text by injecting external knowledge at three levels of word level, phrase level and sentence pattern level. And semantic difference learning: learning fine-grained semantic features by comparing a learning framework, and improving the recognition capability of the model on the text semantic difference. Intelligent weight distribution of semantic information is carried out, and dynamic balance and optimal combination of different source information are realized by evaluating feature importance and dynamically adjusting weights. And predicting the text similarity, calculating the similarity score of the text pair by integrating the feature information of each stage, and providing an interpretable prediction result. The complete text matching method and system are designed based on the achievement transformation large model, the problems of insufficient knowledge utilization, incomplete semantic representation, fixed feature weight and the like of the large model in a text matching task are effectively solved, the text matching effect of the large model in an actual application scene is improved, and the text matching efficiency is improved. And a feasible solution is provided for achievement transformation of a large model technology.

A method and related system operations for summarizing information includes determining a plurality of local explainability parameters for a set of inputs using a local explainability model for a prediction model. The method further includes generating a plurality of feature effect indicators for a target prediction and generating a plurality of phrases based on the plurality of feature effect indicators by using a text generation model. The method further includes generating a summarization using a large language model based on the plurality of phrases.

To effectively resolve the issue of inappropriate handovers during communication sessions, methods, systems, and machine-readable mediums which utilize speech features captured by microphones in the original wireless peripheral device and / or the wireless peripheral device to which the communication session is to be handed over to determine if a handover should proceed or should be reversed. Speech features refer to the composite attributes of spoken language that encompass both acoustic and linguistic features. Acoustic features characterize the sound properties of speech and include, but are not limited to, timbre, pitch, intonation, speaking rate, articulation, prosody, melody, spectral features, formant frequencies, and the like. Linguistic features pertain to the actual content conveyed, comprising words, phrases, syntax, and semantics. This includes the analysis of words, phrases, syntax, and semantics to understand the context and continuity of the conversation.

Systems and methods are described for obtaining accurate responses from large language models (LLMs) and chatbots, including for question and answering, exposition, and summarization. These systems and methods accomplish these objectives via use of noun phrase avoiding processes such as a noun phrase collision detection process, a query splitting process, and a topical splitting process as well as by use of formatted facts, formatted fact model correction interfaces (FF MCIs), bounded-scope deterministic (BSD) neural networks, processes and methods, and intelligent storage and retrieval (ISAR) systems and methods. These systems and methods avoid and bypass noun phrase collisions and correct for errors caused by noun phrase collisions so that hallucinations are eliminated from LLM responses.

The present invention relates to a method to provide comprehensive key vocabulary for search involving the steps of receiving (101) a user-specified term or phrases with text cleaning (101A), identifying (102) domain specific vocabulary, extracting (103) relationship between plurality of key phrases in a text document, deriving (104) a knowledge graph that involves in establishing relationship between two phrases and a connection to form edge relationship thereby deriving a lemma, improving (105) the quality of the key phrases, storing (106) phrases, document ids and statistical details about phrases, ordering (107) the phrases identified using the ranking criteria based on the phrases statistics, selecting (108) and returning the best key phrases as search results and showing (109) the search results as word cloud for the users to quickly identify the key vocabulary.

For semantic data leak prevention, one or more semantic interpretations of an input text are obtained with a generative pre-trained transformer, GPT, configured for interpreting the input text with reference to a target context, and it is determined whether the input text comprises a hidden data leakage by checking the semantic interpretations of the input text with a data leak prevention, DLP, system configured for detecting a data leakage based on a collection of words and phrases of interest. As a result, enabled is a prevention of hidden data leaks when a sensitive information is hidden within a text, for example, by using advanced language techniques.

This disclosure relates to extraction of tasks from documents based on a weakly supervised classification technique, wherein extraction of tasks is identification of mentions of tasks in a document. There are several prior arts addressing the problem of extraction of events, however due to crucial distinctions between events-tasks, task extraction stands as a separate problem. The disclosure explicitly defines specific characteristics of tasks, creates labelled data at a word-level based on a plurality of linguistic rules to train a word-level weakly supervised model for task extraction. The labelled data is created based on the plurality of linguistic rules for a non-negation aspect, a volitionality aspect, an expertise aspect and a plurality of generic aspects. Further the disclosure also includes a phrase expansion technique to capture the complete meaning expressed by the task instead of merely mentioning the task that may not capture the entire meaning of the sentence.

The invention belongs to the technical field of natural language information processing, and discloses a text compression method based on natural language features and self-information, and the specific technical scheme is as follows: step 1, carrying out data preprocessing, useless information deletion and multi-document retrieval simulation on a Google question and answer data set in an open source corpus to obtain a data set oriented to long context compression, obtaining a multi-document long text; 2, performing self-information evaluation on the multi-document long text to obtain phrase granularity and text self-information distribution, extracting natural language features of part-of-speech analysis, named entity analysis and dependency relationship analysis, and optimizing distribution based on the natural language features to obtain a compressed text after distribution optimization; and step 3, compressed text prompt setting, target question and answer model construction, question and answer performance index obtaining and application to downstream tasks, the method is very excellent in long text compression accuracy, and the text compression performance is improved.

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.

The invention discloses a question answer determination method and device and electronic equipment. The method comprises the steps that a to-be-retrieved question input by a target object is obtained, a target document associated with the to-be-retrieved question is obtained, and the target document comprises an internal document in an internal database and real-time knowledge in an external knowledge base; determining a first semantic vector and a first keyword corresponding to the to-be-retrieved question, and determining a second semantic vector and a second keyword corresponding to the target document; according to the first semantic vector, the second semantic vector, the first keyword and the second keyword, determining a candidate phrase with the highest association degree with the to-be-retrieved question from the target document; and analyzing the candidate phrases by adopting a generative language model to obtain an answer to the to-be-retrieved question. According to the method and the device, the technical problems of low retrieval efficiency and inaccurate semantic matching in problem retrieval in related technologies are solved.

The invention discloses a large language model reasoning acceleration method and device based on two-stage speculative decoding and a storage medium, and the method comprises the steps: constructing and initializing a Trie tree, and inserting a historical corpus, and phrase sequences in a document library or a code library into the Trie tree one by one; in the reasoning process, longest prefix matching is carried out based on a Trie tree, and a candidate draft sequence is generated by adopting branch backtracking and recursive search; performing confidence evaluation on the candidate draft sequence, calculating a joint confidence score of the sequence through probability multiplication and a Top-K screening mechanism, and judging whether the joint confidence score reaches a confidence threshold; if the accumulated confidence of the candidate sequence reaches a threshold value, skipping a small model generation stage, and directly entering large model verification; otherwise, entering a small model draft completion stage; and the final large model takes the replaced and updated draft sequence as final output. According to the method, adaptive acceleration of the decoding process can be realized, and the long text reasoning delay of the large language model is remarkably reduced while the generation quality is ensured.

To provide a sentence generation device, a sentence generation system, a sentence generation method, and a program capable of acquiring a sentence desired by a user.SOLUTION: A sentence generation device acquires an output sentence regarding theme information to be provided to an objective person. The sentence generation device includes: a persona information setting unit which sets the persona information indicating an attribute of the objective person; an input sentence generation unit which generates an advice sentence being a sentence regarding the objective person on the basis of the persona information, by not using the theme information, but by using a learning model, and extracts a key phrase from the advice sentence; and an output sentence acquisition unit which acquires an output sentence generated by using the learning model on the basis of the key phrase extracted by the input sentence generation unit and the theme information.SELECTED DRAWING: Figure 2

In embodiment, a method of displaying an answer of a question-answer pair in response to a natural language search query includes receiving, from a Bidirectional Encoder Representations from Transformers (BERT) model, an array of attention matrices for the question-answer pair, where each attention matrix of the array of attention matrices includes an array of attribution values, generating a total attribution value for each word of an answer of the question-answer pair from the array of attention matrices, and displaying the answer on an electronic display, wherein one or more words of the answer is highlighted based on the total attribution values for each word.

Computer-based generation of information-gathering questions in response to a user query can include parsing the user query using natural language processing and extracting from the user query one or more phrases corresponding to a predetermined category. A knowledge database can be accessed, and entities semantically related to each phrase can be extracted therefrom. A query sub-graph representing a relationship between each of the one or more phrases and the entities extracted from the knowledge database can be generated. An expanded user query can be generated by traversing the query sub-graph. Passages semantically related to the expanded user query can be retrieved from one or more passages databases and ranked. A neural question generator can generate a set of information-gathering questions based on the expanded user query and a group of select passages selected from the plurality of passages in accordance with each selected passage's ranking.

A server generates a user interface for allowing a user to rewrite portions of text for an electronic document to mitigate detected issues. For an input document, the server generates one or more indications over the one or more phrases in the sentence. An indication for a phrase may be generated based on a respective category associated with the phrase. Responsive to receiving an indication from the user to rewrite the sentence, the server generates a prompt to a machine-learned language model. The server receives a response generated by executing the machine-learned language model on the prompt. The server generates a pane user element to present the candidate sentence and an evaluation of the candidate sentence to the user, and responsive to receiving a selection of a candidate sentence, replacing the sentence in the editor with the selected sentence on the user interface.

The invention provides a retrieval method and device based on graph structure retrieval enhancement, electronic equipment and a storage medium, and the method comprises the steps: obtaining original data, screening the original data based on a preset screening rule to obtain a plurality of candidate phrases, calculating the similarity between the plurality of candidate phrases and a preset central semantic vector of each layer, the multiple candidate phrases are layered according to the similarity, feature vectors are constructed, levels and feature vectors corresponding to the multiple candidate phrases are obtained, graph entities are obtained based on the candidate phrases, context fragments of the graph entities in original data are extracted, and relation types between the graph entities are obtained based on the context fragments; according to the relationship type between the graph entities and the hierarchy of the graph entities, a multi-hierarchy knowledge graph is obtained, and retrieval is performed based on the multi-hierarchy knowledge graph and the corresponding associated information block, so that the technical problems of result redundancy and logic fracture of a retrieval enhancement generation method are solved through the method.

The invention relates to the technical field of data identification, in particular to a data security identification method covering physics, semantics and sensitive points, and the method comprises the following steps: processing an obtained first industry review document set and a second industry review document set to generate an industry semantic feature lexicon; performing semantic analysis on each sensitive point information sentence to generate a sensitive point information phrase sequence, taking the sensitive point information phrase sequence as input, performing training through a professional field document to generate a word vector model, and performing similarity calculation on the word vector model to generate a sensitive feature lexicon; based on the industry semantic feature lexicon and a word segmentation tool, keyword segmentation is carried out on the target document, and a semantic security identifier is generated; and performing word segmentation on a target sentence in the target document based on the inactive word bank and the industry semantic feature word bank, and generating a sensitive security identifier in combination with the sensitive feature word bank. By applying the method, effective identification can still be realized under the condition of data adjustment, and semantic representation of the body is enhanced.

A method for extracting key phrases based on extracting both informative and readable phrases from documents. The method includes separating documents into sentences to extract sentence-level phrases and separating the documents into segments of a fixed length to extract segment-level phrases. The method further includes using a PMI and entropy model to also extract readable phrases from the documents. The method includes comparing the sentence-level phrases, the segment-level phrases, and the readable phrases to generate key phrases for the documents.