127 results about "Sentence generation" patented technology

A system and method for machine translation are disclosed. Source sentences are received. For each source sentence, a target sentence comprising target words is generated. A plurality of translation neighbors of the target sentence is generated. Phrase alignments are computed between the source sentence and the translation neighbor. Translation neighbors are scored with a translation scoring model, based on the phrase alignment. Translation neighbors are ranked, based on the scores. In training the model, parameters of the model are updated based on an external ranking of the ranked translation neighbors. The generating of translation neighbors, scoring, ranking, and, in the case of training, updating the parameters, are iterated with one of the translation neighbors as the target sentence. In the case of decoding, one of the translation neighbors is output as a translation. The system and method may be at least partially implemented with a computer processor.

By inputting words of source language as a keyword (31), a translation pairs are extracted (50) from a parallel corpus database including source language and target one. From the partially corresponding information on the translation sentence, a corresponding phrase group table formed by the corresponding phase of the target language corresponding to the source language phrase including a keyword phrase f the source language is stored (60). Text generator (70) assumes a relationship between the phrases of different language contained in the corresponding phrase group table and generates a text sentence candidate (32) of the target language.

By inputting words of source language as a keyword (31), a translation pairs are extracted (50) from a parallel corpus database including source language and target one. From the partially corresponding information on the translation sentence, a corresponding phrase group table formed by the corresponding phase of the target language corresponding to the source language phrase including a keyword phrase f the source language is stored (60). Text generator (70) assumes a relationship between the phrases of different language contained in the corresponding phrase group table and generates a text sentence candidate (32) of the target language.

A method and computer system for translating sentences between languages from an intermediate language-independent semantic representation is provided. On the basis of comprehensive understanding about languages and semantics, exhaustive linguistic descriptions are used to analyze sentences, to build syntactic structures and language independent semantic structures and representations, and to synthesize one or more sentences in a natural or artificial language. A computer system is also provided to analyze and synthesize various linguistic structures and to perform translation of a wide spectrum of various sentence types. As result, a generalized data structure, such as a semantic structure, is generated from a sentence of an input language and can be transformed into a natural sentence expressing its meaning correctly in an output language. The method and computer system can be applied to in automated abstracting, machine translation, natural language processing, control systems, Internet information retrieval, etc.

An information transmission method in which in a process of transmitting information the transmission is conducted so as to increase uncertainty has not been established. Information terminals have a plurality of fixed phrases for sentence generation, and sentences are generated by inserting words, which are inputted into the information terminals or selected in relation to an event created on the information terminals, into a fixed phrase for sentence generation selected from the plurality of fixed phrases for sentence generation. For a sentence received at another information terminal via a communication channel, a determination is made as to whether or not to vary that sentence, and when variation is to be carried out, a new sentence is generated by replacing a fixed phrase for sentence generation contained in the received sentence with a different associated fixed phrase for sentence generation or by replacing a word contained in the received sentence with another word or symbol, and forwarding the generated new sentence via the communication channel to the other information terminal.

A platform to implement a Chinese input method consists of the following components: 1. A Keypad. 2. A cascade Multi-Window, 3. A Sentence Editing Buffer. 4. An Attribute Viewing Window. 5. A Text Accumulation Window. 6. A two-level phrase set refining control window. The keypads are designed based on the lexical structures of the Zhu-Yin and Pin-Yin phonetic systems. Efficient mouse operations have been designed to enter phonetic symbol strings. The Multi-Window can present a great many candidate words and phrases. It allows a user to browse on its multi pages without mouse clicking. A two-phase sentence generation procedure relieves users from the burden of sentence segmentation and creates the possibility of harvesting system supplied longer generalized phrases.

The invention discloses a video description method based on multi-feature fusion. The method is characterized in that (1) deep spatial-temporal features of a video are extracted by fusing traditional CNN features and SIFT flow features; (2) an S2VT sentence generation model with added average pooling features being video overall features is adopted to generate corresponding sentence description according to the deep spatial-temporal features extracted in the step (1); and (3) a word2vec word vector method is adopted to replace one-hot vector word representation to optimize the sentence generation model in the step (2). The method has the advantages that more robust spatial-temporal features can be better extracted through multi-feature fusion; meanwhile, the average pooling features are added into the sentence generation model, so that more relations are established between visual information and words; finally, the word2vec word vector method is adopted to replace one-hot vector word representation, so that more relations are established between words, and video description performance is effectively improved.

The invention provides an SQL sentence generation method and a system. The SQL sentence generation method comprises the following steps: setting a descriptive table entry of the fields and tables; filling in the descriptive information corresponding to the descriptive table entry to generate a data dictionary; and generating the SQL sentences according to the data dictionary and the structure query language (SQL) sentence generation rules. In the embodiment of the invention, unified descriptive entry is set firstly, and descriptive information is filled, thus providing a possibility for realizing the purpose of the invention; and then according to the SQL sentence generation rules, the application data dictionary is set into a template of the SQL sentence, thus realizing automatic and rapid generation of the SQL sentence in a large scale, and improving the compiling efficiency of the SQL sentence.

A toy having a speech recognition function for two-way conversation with a dialogue partner. Digital speech processing circuitry incorporates first and second measures together with speech recognition components, to store speech compression data and recognition data, for controlling sentence generation in response to speech recovered from the partner.

The invention discloses an RNN-based automatic picture description generation method. A deep web which is well trained in advance is firstly used for image feature extraction; non-noun and non-verb components are removed for words in the sentence; an LSTM network is finally used for joint training on the image features and lexical features; during the sentence generation process, a sentence formed by nouns and verbs is generated through the inputted image and the well-trained LSTM network; and then, through large corpus on the network, the final outputted sentence is generated. Automatic recognition can be realized, a digital image uploaded by the user is understood, and a natural sentence understood by a human being is generated.

A method of, and system for, generating a sentence from a semantic representation maps the semantic representation to an unordered set of syntactic nodes. Simplified generation grammar rules and statistical goodness measure values from a corresponding analysis grammar are then used to create a tree structure to order the syntactic nodes. The sentence is then generated from the tree structure. The generation grammar is a simplified (context free) version of a corresponding full (context sensitive) analysis grammar. In the generation grammar, conditions on each rule are ignored except those directly related to the semantic representation. The statistical goodness measure values, which are calculated through an analysis training phase in which a corpus of example sentences is processed using the full analysis grammar, are used to guide the generation choice to prefer substructures most commonly found in a particular syntactic / semantic context during analysis.

The embodiment of the invention provides a method and device for generating webpage text summarization. The method comprises the steps that a key entry in a webpage text is determined; the initial weight of each entry in the webpage text is calculated according to the occurrence frequency of each entry in the webpage text; when any entry in the webpage text is the key entry, the initial weight of the key entry is increased to obtain the target weight of the key entry; when any entry in the webpage text is a non-key entry, the initial weight of the non-key entry is adopted as the target weight of the non-key entry; the target weight of each entry in the webpage text is utilized for obtaining the sentence weight of each sentence through calculation; summarization sentences are selected according to the sentence weight, and the webpage text summarization is generated by the summarization sentences. According to the method and device, the accuracy of the webpage text summarization is improved.

Provided is a route guidance system that provides a route in premises in order to perform service of route guidance in the premises, which includes: a route guidance apparatus, which is comprised of visible region calculation means that calculates a range where a user can see, signpost extraction means that extracts a signpost to be guided from a visible region, route search means that searches the route from a starting point to a destination, route information generation means that comprises guidance map generation means that generates a map for guidance and guidance sentence generation means that generates the guidance sentence, route leading means that obtains current positional information to send appropriate guidance information to a portable terminal, and position determination means that performs coordinate conversion to information from a position detection apparatus of the user; a portable terminal apparatus including a user interface, which displays an image or a route guidance sentence; a position detection apparatus that obtains current positional information of the user; a map DB that stores data / signpost information regarding the route; a guidance sentence DB that stores basic data for generating the guidance sentence; and route information data that stores guidance data output from the route information generation means.

A method and computer system for translating sentences between languages from an intermediate language-independent semantic representation is provided. On the basis of comprehensive understanding about languages and semantics, exhaustive linguistic descriptions are used to analyze sentences, to build syntactic structures and language independent semantic structures and representations, and to synthesize one or more sentences in a natural or artificial language. A computer system is also provided to analyze and synthesize various linguistic structures and to perform translation of a wide spectrum of various sentence types. As result, a generalized data structure, such as a semantic structure, is generated from a sentence of an input language and can be transformed into a natural sentence expressing its meaning correctly in an output language. The method and computer system can be applied to in automated abstracting, machine translation, natural language processing, control systems, Internet information retrieval, etc.

The invention belongs to the technical field of cross-media generation, and particularly relates to a video description generation system based on a graph convolution network. The video description generation system comprises a video feature extraction network, a graph convolution network, a visual attention network and a sentence description generation network. The video feature extraction network performs sampling processing on videos to obtain video features and outputs the video features to the graph convolution network. The graph convolution network recreates the video features accordingto semantic relations and inputs the video features into sentence descriptions to generate a recurrent neural network; and the sentence description generation network generates sentences according tofeatures of video reconstruction. The features of a frame-level sequence and a target-level sequence in the videos are reconstructed by adopting graph convolution, and the time sequence information and the semantic information in the videos are fully utilized when description statements are generated, so that the generation is more accurate. The invention is of great significance to video analysisand multi-modal information research, can improve the understanding ability of a model to video visual information, and has a wide application value.

The invention discloses a graph model text abstract generation method based on word frequency and semanteme. The method comprises the following steps of 1) performing word segmentation on sentences ina text, and performing part-of-speech tagging; 2) filtering the lexical items, and only reserving the lexical items with specific part-of-speech; and 3) training word vectors by using a Word2Vec model and a BM25 algorithm to form a feature word vector set, further representing sentences, and constructing a sentence-word text matrix; 4) constructing a text undirected graph model through the text matrix; and 5) performing iterative computation of sentence node weights by using a TextRank algorithm until convergence, and selecting TOP-K sentences to generate text abstracts. 6) experimental results show that the method is suitable for industrial production, compared with a traditional text automatic abstracting method considering a single word frequency characteristic of a text and based on atext semantic characteristic, according to the method, under the optimal adjustment factor combination, a higher Rouge value is obtained, it is proved that the method effectively integrates text wordfrequency and semantic features, and then the abstract generation accuracy is improved through a TextRank algorithm based on contextual information.

The invention provides an image natural language description generation method and device with cross-linguistic learning ability. The method comprises the steps that English description sentences are translated into target language description sentences through a machine; part of the target language description sentences are selected randomly to form a training sample set; a smooth sample set and a non-smooth sample set are used for training a sentence smooth degree model; the target language description sentences in a candidate data set are subjected to smooth degree evaluation through the sentence smooth degree model, and according to the smooth degree probability of each target language description sentence, a strategy for training an image description sentence generation model is set; according to the strategy, the image description sentence generation model is trained, and the trained image description sentence generation model is obtained. According to the image natural language description generation method and device with the cross-linguistic learning ability, the aim is achieved that the image description sentence generation model of a target language is guided, trained and generated dependent on smooth degree evaluation results, the influence of non-smooth target language description sentences on the training process is lowered, and the accuracy of the image description sentence generation model of the target language is improved.

The invention provides a query sentence optimization device which comprises a query location module used for obtaining an external query and a subquery in original inquiry sentence, a query combination module used for combining the external query and the subquery into a new external query and a query sentence generation module used for generating new query sentence according to the new external query. Correspondingly, a query sentence optimization method is further provided. By means of the technical scheme, field information of the queries can be compared according to the external query and the subquery to further judge which fields are redundant fields, filter the redundant fields and improve execution efficiency of structured query language (SQL) sentence. Simultaneously, the external query and the subquery are combined to be operated, nested subqueries are removed, and the aims of flattening and simplifying the SQL sentence are achieved.

The invention discloses an image description generation method based on an architectural short sentence constraint vector and a dual visual attention mechanism, A semantic model for automatically describing the visual content information of images is obtained by training a large number of labeled texts. The semantic model is composed of three parts: a short sentence generation model, a dual visualattention mechanism and a constrained language model. The text description can be generated automatically for any input test image. The invention can effectively establish the relationship between the word and the image in the text description, and has good performance for describing the salient objects or scenes of the image based on semantics.

The present invention provides sentence meaning expression creating method and device, machine translation method and system, word and expression interpreting method with meaning range automatically reduced based on the context and electronic dictionary with meaning range automatically reduced based on the context. The method of creating the sentence meaning expression for one sentence includes: editing semanteme unit expressing tree index library based on original language expression of semanteme unit expressing library; extracting semanteme unit expressing trees with each word in the sentence as real beginning based on the semanteme unit expressing tree index library; pruning the extracted semanteme unit expressing trees one word by one word; and finding the sentence meaning expression of the sentence based on the remained semanteme units.

A method for summarizing dialogs may include obtaining an agent text stream and a customer text stream, segmenting the agent text stream and customer text stream into sentences, and labeling sentences associated with the segmented agent text stream and the segmented customer text stream. The method may further include extracting sentences from the agent text stream and the customer text stream based upon frequencies of appearance of words and terms of interest; generating an agent summary paragraph based on the extracted sentences from the agent text stream, and generating a customer summary paragraph based on the extracted sentences from the customer text stream. The method may identify keywords associated with each of the agent summary paragraph and the customer summary paragraph.

A topic text sentence vector generation method comprises the following steps: S1, screening out all keywords according to topic text expression, adding a dictionary, carrying out dictionary word segmentation on sentences in a topic text, and marking the keywords appearing in the sentences at the same time; S2, on the basis of the word segmentation result and all screened keywords, after each sentence and the keywords contained in the sentence are coded, establishing an RNN model, and carrying out prediction training by adopting a method of randomly removing the keywords; and S3, generating a sentence vector for each sentence in the question text by utilizing the characteristics extracted by the trained model.

The invention discloses a medical report generating model based on a relation model, and a generating method thereof. The model comprises the components of a deep convolutional network which is used for classifying abnormal terminologies and extracting a visual characteristic, inputs the visual characteristic into a two-stage bidirectional long-and-short-period memory network layer for performingattention operation, and is connected with a full connecting layer after each stage of bidirectional long-and-short-period memory network layer for selecting a template and a word; a hierarchical recursive neural network which comprises two bidirectional long-and-short-period memory network layers, wherein the top bidirectional long-and-short-period memory network layer is operated through attention and is connected with the full connecting layer for selecting an accurate template, and the lower bidirectional long-and-short-period memory network layer receives the information of the top layerand is connected with the full connecting layer for selecting an accurate word; and a report generating module which comprises a template decoder for predicting an abnormal sentence and a word decoderfor generating a normal sentence, thereby selecting generating using template searching or sentence generation on the current sentence according to decision of the hierarchical recursive neural network, and finally connects all sentences which are obtained through searching or automatic generation for forming the medical report.

The present invention is provided with: a morpheme analysis unit 105 for analysis of text inputted by a user in natural language; an intent-inference processing unit 107 that, making reference to an intent inference model in which words and user intent inferred from the words are stored in associated form, infers the intent of the user from the result of the text analysis by the morpheme analysis unit 105; an unknown term extraction unit 108 that, in the event that the intent of the user cannot be uniquely identified by the intent-inference processing unit 107, extracts from the text analysis results an unknown term that is a word not stored in the intent inference model; and a response sentence generation unit 110 for generating a response sentence that includes the unknown term extracted by the unknown term extraction unit 108.

A method (and system) of handling out-of-grammar utterances includes building a statistical language model for a dialog state using, generating sentences and semantic interpretations for the sentences using finite state grammar, building a statistical action classifier, receiving user input, carrying out recognition with the finite state grammar, carrying out recognition with the statistical language model, using the statistical action classifier to find semantic interpretations, comparing an output from the finite state grammar and an output from the statistical language model, deciding which output of the output from the finite state grammar and the output from the statistical language model to keep as a final recognition output, selecting the final recognition output, and outputting the final recognition result, wherein the statistical action classifier, the finite state grammar and the statistical language model are used in conjunction to carry out speech recognition and interpretation.

The invention provides a method for judging continuity of conversion contents. The method includes the steps that previous text sentences are input into a sentence generation model, so that next textsentences are generated; the similarity of each previous text sentence and a current sentence is calculated to construct a first similarity matrix; the similarity of each next text sentence and the current sentence is calculated to construct a second similarity matrix; the first similarity matrix and the second similarity matrix are separately input into a continuity judgment model to generate continuity feature parameters of the current sentence, and the continuity judgment model is constructed based on a convolutional neural network. The continuity judgment model and the sentence generationmodel are combined to realize the continuity of the conversion contents; the continuity of two sentences can be compared based on dimensions of semantics, and replies which are continuous and excellent in quality are pushed to users. The invention further provides a device for judging the continuity of the conversion contents and equipment.

The invention discloses a voice playing method and device and electronic equipment, and relates to the technical field of computers. The method comprises the steps of receiving a voice playing instruction sent by a target client for a target text; obtaining clause information of the target text, wherein the clause information comprises content of each sentence in the target text and a role corresponding to the sentence; obtaining current voice configuration information of the target text, wherein the current voice configuration information is used for representing the tone of the role; according to the clause information and the current voice configuration information, generating voice data of the tone of the role corresponding to the sentence for each sentence; and sending the file address of the multimedia file including the storage address of the voice data to a target client, so that the target client plays the voice data according to the file address. Thus, the tone of the voice data corresponding to the text can be adjusted in real time according to user requirements in the target client needing to play the voice file generated offline.

Embodiments of the present invention are directed to computer-implemented methods and systems for automatically generating a description of a task or expectation. The method comprises receiving, in the form of an electronic communication, a natural language sentence that expresses a call or commitment to action; generating, using a machine learning model, a description of a task or expectation for a user based on the natural language sentence; and storing the description of the task or expectation in a non-transitory computer-readable memory.

A machine-translation apparatus using multi-level verbal-phrase patterns includes: a simple sentence generation unit for generating an input simple sentence; a basic verbal-phrase pattern-matching unit for trying a match of a semantic code of each case component of the input simple sentence with basic verbal-phrase patterns; a default verbal-phrase pattern matching unit for trying a match of a size and case prepositions of the input simple sentence with default verbal-phrase patterns having a verb identical to that of the input simple sentence; a default word-order matching unit for trying a match of a word-order of the input simple sentence with default word-order verbal-phrase patterns having a case component structure identical to that of the input simple sentence; and a default preposition matching unit for generating a target sentence of an input sentence with default preposition patterns having a context identical to that of the input simple sentence.

The invention discloses a generation method and device of a synonymous sentence generation model, and a medium. The method comprises the steps of performing training by using a first set to obtain a generation model and a discrimination model; wherein the first set comprises a plurality of synonymous sentence groups, and each synonymous sentence group comprises at least two synonymous sentences; and performing iterative processing on the generation model and the discrimination model until the generation model converges. A large number of simple sentences are used, and the advantages of low cost and no need of manual annotation of the simple sentences are fully utilized. In the training process of the model, a large number of simple sentences and a reinforcement learning mode are combined,the semantic richness is greatly improved through the use of the simple sentences, the reinforcement learning mode can enable the model to be continuously optimized in the iteration process, and therefore a high-quality synonymous sentence generation model can be trained completely without depending on a large number of parallel corpora.