927 results about "Word embedding" patented technology

Disclosed systems, methods, and computer readable media can detect an association between semantic entities and generate semantic information between entities. For example, semantic entities and associated semantic collections present in knowledge bases can be identified. A time period can be determined and divided into time slices. For each time slice, word embeddings for the identified semantic entities can be generated; a first semantic association strength between a first semantic entity input and a second semantic entity input can be determined; and a second semantic association strength between the first semantic entity input and semantic entities associated with a semantic collection that is associated with the second semantic entity can be determined. An output can be provided based on the first and second semantic association strengths.

The invention provides a Chinese sentiment analysis method based on BERT, LSTM and CNN fusion. The Chinese sentiment analysis method comprises the steps: performing text preprocessing on a plurality of Chinese corpora in a Chinese corpus data set to obtain a plurality of sequences corresponding to the plurality of Chinese corpora; extracting word embedding of each sequence by using a BERT model; performing feature extraction on each sequence by adopting BERT, LSTM and CNN to obtain text deep semantic features corresponding to each sequence; and classifying the obtained text deep semantic features by using a softmax classifier to train and test the model so as to realize sentiment polarity prediction analysis. According to the technology, the defects in the prior art can be overcome, and the accuracy of Chinese text sentiment analysis is improved.

The invention discloses a natural language processing-based multi-language analysis method and device. The method comprises the following steps of: selecting to input a natural language text information language category through a language detection training model; obtaining word embedding expression information of corresponding words which can be recognized by a computer through a trained word vector model, and extracting a keyword of the obtained word embedding expression information through a TF-IDF manner; calculating an article vector and a category vector of each preset category according to the keyword and a keyword weight, and calculating a similarity between an article of natural language text information and each preset category so as to determine a text classification result ofthe natural language text information; and inputting the word embedding expression information of the natural language text information into a trained convolutional neural network and a parallel-framework text emotion analysis model of a bidirectional gate circulation unit, and obtaining a final emotion tendency value through calculation. According to the method and device, the problem that traditional multi-language analysis method needs to know domain knowledges of related linguistics and needs plenty of manpower to carry out operation is solved.

Systems and processes for operating a digital assistant are provided. In accordance with one or more examples, a method includes, receiving training data for a data-driven learning network. The training data include a plurality of word sequences. The method further includes obtaining representations of an initial set of semantic categories associated with the words included in the training data; and training the data-driven learning network based on the plurality of word sequences included in the training data and based on the representations of the initial set of semantic categories. The training is performed using the word sequences in their entirety. The method further includes obtaining, based on the trained data-driven learning network, representations of a set of semantic embeddings of the words included in the training data; and providing the representations of the set of semantic embeddings to at least one of a plurality of different natural language processing tasks.

A natural language understanding method begins with a radiological report text containing clinical findings. Errors in the text are corrected by analyzing character-level optical transformation costs weighted by a frequency analysis over a corpus corresponding to the report text. For each word within the report text, a word embedding is obtained, character-level embeddings are determined, and the word and character-level embeddings are concatenated to a neural network which generates a plurality of NER tagged spans for the report text. A set of linked relationships are calculated for the NER tagged spans by generating masked text sequences based on the report text and determined pairs of potentially linked NER spans. A dense adjacency matrix is calculated based on attention weights obtained from providing the one or more masked text sequences to a Transformer deep learning network, and graph convolutions are then performed over the calculated dense adjacency matrix.

A natural language processing (“NLP”) manager is provided that manages NLP model training. An unlabeled corpus of multilingual documents is provided that span a plurality of target languages. A multilingual embedding is trained on the corpus of multilingual documents as input training data, the multilingual embedding being generalized across the target languages by modifying the input training data and/or transforming multilingual dictionaries into constraints in an underlying optimization problem. An NLP model is trained on training data for a first language of the target languages, using word embeddings of the trained multilingual embedding as features. The trained NLP model is applied for data from a second of the target languages, the first and second languages being different.

The invention relates to the technical field of artificial intelligence, and discloses a text sentiment analysis method, a text sentiment analysis system and a computer readable storage medium, for improving the accuracy of text sentiment analysis. The method comprises the steps of inputting a word vector corresponding to any sentence into a preset LSTM network model, thus acquiring a hiding layervector of each word; tagging the part-of-speech of an acquired text word set, training the text word set carrying part-of-speech tagging information, and splitting a part-of-speech vector matrix generated by training by using words as units, thus acquiring the part-of-speech vector corresponding to each word; using sentences as units, performing word embedding weighted summation attention analysis on the hiding layer vector and the part-of-speech vector corresponding to each word in the sentence to acquire a sentence vector carrying attention information of each sentence, and using the sentence vector carrying the attention information to serve as the input of a sentiment classification model, thus acquiring a sentiment classification result of each sentence and/or a classification resultof the original text.

The invention discloses a text classification method combining dynamic word embedding with part-of-speech tagging, and provides the text classification method based on a deep neural network through combining dynamic word embedding with part-of-speech tagging. The method can fully utilize the advantages that a large-scale corpus can provide more accurate grammar and semantic information, and can also adjust word embedding by combining with the features of the corpus during a model training process, and thus the features of the corpus can be better learned. Meanwhile, classification accuracy can be further improved by combining with part of speech information of words in sentences. The invention also comprehensively utilizes the advantages of LSTM in the aspect of learning context information of words and part of speech in the sentences, and the advantages of CNN in the aspect of learning text local features. The classification model provided by the invention has the advantages of high accuracy and strong universality, and achieves good effect in some famous public corpuses including IMDB corpus, Movie Review and TREC.

The invention discloses a fine-grained emotion polarity prediction method based on a hybrid attention network, and aims to overcome the problems of lack of flexibility, insufficient precision, difficulty in obtaining global structure information, low training speed, single attention information and the like in the prior art. The method comprises the following steps: 1, determining a text context sequence and a specific aspect target word sequence according to a comment text sentence; 2, mapping the sequence into two multi-dimensional continuous word vector matrixes through log word embedding;3, performing multiple different linear transformations on the two matrixes to obtain corresponding transformation matrixes; 4, calculating a text context self-attention matrix and a specific aspect target word vector attention matrix by using the transformation matrix, and splicing the two matrixes to obtain a double-attention matrix; 5, splicing the double attention matrixes subjected to different times of linear change, and then performing linear change again to obtain a final attention representation matrix; and 6, through an average pooling operation, inputting the emotion polarity into asoftmax classifier through full connection layer thickness to obtain an emotion polarity prediction result.

The invention discloses a text abstract automatic generation method based on a self-attention network. The text abstract automatic generation method comprises the following steps: 1) carrying out wordsegmentation on an input text to obtain a word sequence; 2) performing word embedding on the word sequence to generate a corresponding word vector sequence; 3) coding the word vector sequence by using a self-attention network encoder, and 4) decoding an input text coding vector by using a self-attention network decoder to generate a text abstract. The text abstract automatic generation method hasthe advantages of high model calculation speed, high training efficiency, high generated abstract quality, good generalization performance of the model and the like.

The invention discloses a deep learning model-based image Chinese description method and belongs to the field of computer vision and natural language processing. The method comprises the steps of preparing an ImageNet image data set and an AI Challenger image Chinese description data set; pre-training the ImageNet image data set by utilizing a DCNN to obtain a pre-trained DCNN model; performing image feature extraction and image feature mapping on the AI Challenger image Chinese description data set, and transmitting image features to a GRU threshold recursive network recurrent neural network;performing word coding matrix construction on an AI Challenger image mark set in the AI Challenger image Chinese description data set; extracting word embedding features by utilizing an NNLM, and finishing text feature mapping; taking the GRU threshold recursive network recurrent neural network as a language generation model, and finishing image description model building; and generating a Chinese description statement. According to the method, the blank of image Chinese description is filled up; a function of automatically generating the image Chinese description is realized; the accuracy ofdescription contents is well improved; and a foundation is laid for development of Chinese NLP and computer vision.

The invention discloses an automatic question generation method based on deep learning. The method comprises the following steps: constructing a training set (articles, answers and questions), a verification set (articles, answers and questions) and a prediction set (articles and answers); creating Encoder-based construction by using deep learning framework tensorflow A sequence of decoders to a sequence neural network model; Carrying out word segmentation, word list making and word embedding operations on sentences in the data set; Wherein the data set comprises a training set, a verificationset and a prediction set; The training set is used for training the model, the verification set is used for detecting whether the currently trained model is over-fitted or not, and if over-fitting isachieved, stopping training; Otherwise, continuing training; And decoding the prediction set by using the trained model to generate a problem. The method is good in generalization effect and low in labor cost, the generated questions are better matched with articles and answers, and the method can be widely applied to the fields of intelligent teaching, intelligent questions and answers, knowledge question and answer games and the like.

The invention provides an image description generating method based on a neural network and image focuses. The method is characterized in that an original one-layer word embedding structure is replaced by a two-layer word embedding structure, and accordingly word expression can be learned effectively; image feature expression is directly used as the input of an m-RNN model, and accordingly the capacity of a circulating layer can be fully utilized, and the small-dimension circulating layer can be used; by a decision soft attention mechanism, the attention of an image salient region is reflected and used as one input of a multi-modal layer. By the method, the focus relation between targets or scenes is utilized effectively, and the semantic features of the image is described in a targeted manner.

The invention discloses a text classification method and device, and relates to the technical field of artificial intelligence. The method comprises the steps of performing word segmentation processing on a to-be-classified text to obtain a word segmentation set corresponding to the to-be-classified text; conducting vectorization processing on the word segmentation set according to a preset word vector dictionary, acquiring a word vector set corresponding to the text to be classified and fursing the word vector dictionary with fast text vectors and word embedding vectors corresponding to segmented words; predicting a category label on a word vector set corresponding to the to-be-classified text through a preset label prediction model, acquiring the label prediction model through training according to the training set and a test set, and the test set is used for correcting error data in the training set; and obtaining a prediction result output by the label prediction model, wherein theprediction result corresponds to a text category corresponding to the to-be-classified text. The text classification accuracy can be greatly improved.

Systems, methods, and articles of manufacture to perform an operation comprising deriving, based on a corpus of electronic text, a machine learning data model that associates words with corresponding usage contexts over a window of time, according to a diffusion process, wherein the machine learning data model comprises a plurality of skip-gram models, wherein each skip-gram model comprises a word embedding vector and a context embedding vector for a respective time step associated with the respective skip-gram model, generating a smoothed model by applying a variational inference operation over the machine learning data model, and identifying, based on the smoothed model and the corpus of electronic text, a change in a semantic use of a word over at least a portion of the window of time.

The invention relates to the field of image understanding, discloses an image description generation system and method based on a weighing attention mechanism, and solves the problems that an existingimage description scheme lacks a polishing process, the training process and the testing process are inconsistent, and the generation description recognition degree is not high. The method comprisesthe following steps: a, processing a data set: extracting global features and local features of an image, constructing the data set, marking words in the data set, and generating corresponding word embedding vectors; B, training an image description generation model: generating rough image description by adopting a first layer of decoder based on a residual attention mechanism, and carrying out polishing on the generated image description by adopting a second layer of decoder based on the residual attention mechanism; And c, further training the model in combination with reinforcement learning: simulating a test process of the model in the training process, guiding the training of the model by generating a described CIDEr score, and adjusting the model in combination with reinforcement learning.

The invention provides a relation extraction method based on Bi-LSTM input information enhancement and belongs to the field of artificial intelligence natural language processing of computers. The method comprises the steps that by applying a strategy annotation dataset of an indeterminate label, a redundancy encoding technology is used for conducting character-level encoding on each word to generate a word form encoding vector; the word form encoding vector and a word embedding vector are spliced to generate a word vector used for capturing word form and word meaning information; Bi-LSTM of input information enhancement is used as a model encoding layer, the word vector is input to an encoding layer, and the encoding vector is output; the encoding vector is input into a decoding layer, and a decoding vector is obtained; by applying three layers of NN, an entity label, a relation type and entity number information are extracted from the decoding vector; finally, the gradient is calculated, the weight is updated, and a model is trained through a maximum target function. By means of the relation extraction method, the robustness of the system is improved, interference information caused by non-entity words is reduced, and the accuracy rate and recall rate of relation extraction are effectively increased.

The invention discloses a vulnerability utilization risk evaluation method based on multi-source word embedding and a knowledge map. The method comprises steps of building data source bodies associated with network security according to an information security knowledge map, then acquiring latest online information from different online data sources; preprocessing data and performing word vectorizatio by using word embedding; mapping data underwent word vectorization to a relation chart between entities and entities according to the knowledge map; and building a vulnerability utilization riskevaluation model by combining historical data sets and the knowledge map, evaluating vulnerability utilization risk, and delivering evaluation basis for vulnerability utilization risks. In the method,vulnerability utilization possibility is evaluated more comprehensively by combining network attack development logics, with the knowledge map as a reference, the method provides valuable reference for security defense, further, the method provides a reliability computing model, so that information really associated with vulnerability utilization prediction can be found in massive fragment information, and evaluation is more accurate.

A recommendation method includes the following steps: S100: converting text about a natural language form of a user and an item into numeric training data using a word embedding unit; S200: adopting an attention mechanism to increase the mutual influence between the user and the item in the training data; S300: Using the convolution neural network model based on attention mechanism to extract thelocal features and the core features of the training data, and finally obtaining the hidden features which can express the global features; S400: analyzing the hidden features by using a factoring machine to obtain the association between the user and the item, and the scoring prediction of the item by the user according to the association, and finally recommending the item to the user. Compared with the existing methods, this method improves the recommended accuracy and accuracy, and improves the data utilization rate.