96 results about "Natural language sentence" patented technology

The present invention relates to an information generation and retrieval apparatus based on a standardized format of sentence structure and semantic structure and a method thereof and a computer readable recording medium for recording a program for implementing the method. The method for generating and retrieving information for use in an apparatus for generating and retrieving information based on standardized formats of sentence structure and semantic structure, comprises a first step of transforming a natural language sentence (information and knowledge) described by a information provider to a conceptual graph depending on standardized formats of sentence structure and semantic structure and indexing the conceptual graph; and a second step of transforming a natural language query sentence inputted from a user to a conceptual graph depending on the standardized formats of sentence structure and semantic structure and searching information relevant to the requirement of the user among the indexed information.

The present invention relates to an information generation and retrieval apparatus based on a standardized format of sentence structure and semantic structure and a method thereof and a computer readable recording medium for recording a program for implementing the method. The method for generating and retrieving information for use in an apparatus for generating and retrieving information based on standardized formats of sentence structure and semantic structure, comprises a first step of transforming a natural language sentence (information and knowledge) described by a information provider to a conceptual graph depending on standardized formats of sentence structure and semantic structure and indexing the conceptual graph; and a second step of transforming a natural language query sentence inputted from a user to a conceptual graph depending on the standardized formats of sentence structure and semantic structure and searching information relevant to the requirement of the user among the indexed information.

An improved system and method for revising natural language parse trees is provided. A revision dependency parser may learn a set of transformation rules that may be applied to dependency parse trees generated by a base parser for revising the dependency parse trees. A corpus of natural language sentences and a set of correct dependency parse trees may be used to train a revision dependency parser to correct dependency parse trees generated by the base parser. A revision engine may compare the dependency parse trees produced by the base parser with the correct ones present in the training data to produce an observation-rule pair for each dependency. A rule may specify a transformation on the predicted dependency parse tree generated by the base parser to replace an incorrect dependency with a corrected dependency or may change the type of dependency expressed for the grammatical function of the dependent word.

Systems and methods for classifying natural language (NL) sentences using a combination of NL algorithms or techniques is disclosed. Each NL algorithm or technique may identify a different similarity trait between two or more sentences, and each may help compare the meaning of the sentences. By combining the various similarity factors, preferably by various weighting factors, a distance metric can be computed. The distance metric provides a measure of the overall similarity between sentences, and can be used to assign a sentences to an appropriate sentence category.

Related free-text documents, a corpus, are used to empirically derive a semantic knowledge base through a method in which documents are segmented into unique sentences, and then used to define sentential propositions which are arranged in a knowledge hierarchy. The method takes compound natural language sentences and transforms them to simple sentences by a process that is a part of the invention. A knowledge editor enables a domain expert using the methods of the invention to map the sentences in the corpus to sentential proposition(s). The resulting knowledge base can be used to semantically analyze documents in data mining and decision support applications, and can assist word processors or speech recognition devices. The invention is illustrated in connection with radiology reports, but it has wide applicability.

One embodiment of the present method, program, and apparatus for natural language generation enables a language generation system to generate a grammatically correct natural language sentence by retrieving and adapting one or more stored sentences having semantic features similar to semantic features of a system-generated semantic representation. The retrieval of stored sentences is guided at least in part by the ease with which the stored sentences may be adapted to produce a grammatically correct sentence, enabling rapid generation of accurate output.

A structured natural language query and knowledge system is provided to allow a user who lacks programming skills to enter a database query or a rule in the form of a structured natural language sentence. The scope of the sentence is preferably defined by an improved object relational query language, an object relational algebra, or both. Command and conditions that appear in natural language form are defined with corresponding formal query texts. A user is prompted to compose a structured natural language sentence using the defined commands and conditions. The user-selected command and its arguments appear as the verb phrase of a structured natural language sentence. The user-selected conditions and their parameters appear as the adjective phrases of the sentence. The sentence is parsed and changed into a translated formal query text for formal database query and rule processing.

A method and apparatus are presented for using directed reasoning to provide natural language responses to questions and commands in natural language. The directed reasoning process uses concepts in the natural language sentence to identify relevant response concepts in a knowledge base, by creating a continuous chain of logic from the input concept to the response concept and limiting the creation of the logic chains to only those consistent with the context of the natural language sentence and information in the knowledge base. The invention allows the identification of relevant concepts that elude traditional concept searching based on pattern matching, while focusing computational resources on only those logic calculations relevant to the natural language sentence and knowledge base. The invention will automatically identify the relevant source documents to use and create a knowledge base during run time, without requiring preprogrammed ontologies or knowledge bases.

A method for processing natural language Chinese sentences can transform a Chinese sentence into a Triple representation using shallow parsing techniques. The method is concerned with parsing Chinese sentences by employing lexical and syntactical information to extract more prominent entities in a Chinese sentence, and the sentence is then transformed into a Triple representation by employing the Triple rules referring to elemental Chinese syntax—SVO (subject, verb, and object in order). The lexical and syntactical information in our method is referring a lexicon possessed of part-of-speech (POS) information and phrase-level syntax in Chinese respectively. The Triple representation consists of three elements which are agent, predicate, and patient in a sentence.

Systems and method for automated contextual dialog generation for cognitive conversations include embedding a natural language sentence input by a user into a corresponding sentence vector using a sentence embedder. A context array is generated using a contextual sentence embedder to embed the sentence vector and previous sentence vectors of a conversation history into a context array. Response words are predicted from the sentence vector by performing sequence-to-sequence dialog prediction with a dialog prediction network. Context of the input sentence is quantified by extracting context features from the context array using a situation quantification network. A response dialog is generated in natural language to display to a user, the response dialog responding to the input sentence with a response generator by determining a dialog state including the response words and the quantified context and optimizing the response dialog with reinforcement learning corresponding to the dialog state.

The present invention provides a translation memory system including: a memory which stores plural pairs of a natural language sentence written in a first language and an interlingua representation of the natural language sentence; an analysis unit which performs a syntactic and semantic analysis on a natural language sentence written in a second language and translates the natural language sentence into an interlingua representation on the basis of the analysis result; a search unit which searches the memory to identify an interlingua representation which corresponds to or has a predetermined level of similarity to the interlingua representation obtained by the analysis unit, and which extracts a natural language sentence written in the first language paired with the identified interlingua representation; and an output unit which outputs the natural language sentence extracted by the search unit as a translation result.

To enable a user to perform, by using an ontology, an information search utilizing a meaning given to information on a network by Semantic Web without being required to perform complicated operations for forming an inquiry sentence in conformity with the ontology. In a stage before a search engine for performing an information search by using an ontology is provided an information search supporting system having a morpheme analysis section, a syntactic and semantic analysis section, and a conversion execution section which converts a natural language sentence on which syntactic analysis and semantic analysis have been performed into an inquiry sentence described in an ontology description language by referring to a case frame ontology dictionary in which are stored information indicating to which property in an ontology does the relationship among a predicate, a subject and an object in the natural language sentence correspond and the case frame of the natural language sentence in the property. A natural language sentence is input and an inquiry sentence described in the ontology language is formed on the basis of the results of syntactic analysis and semantic analysis on the natural language sentence and is provided to the search engine.

Classification of sequences, such as the translation of natural language sentences, is carried out using an independence assumption. The independence assumption is an assumption that the probability of a correct translation of a source sentence word into a particular target sentence word is independent of the translation of other words in the sentence. Although this assumption is not a correct one, a high level of word translation accuracy is nonetheless achieved. In particular, discriminative training is used to develop models for each target vocabulary word based on a set of features of the corresponding source word in training sentences, with at least one of those features relating to the context of the source word. Each model comprises a weight vector for the corresponding target vocabulary word. The weights comprising the vectors are associated with respective ones of the features; each weight is a measure of the extent to which the presence of that feature for the source word makes it more probable that the target word in question is the correct one.

The invention discloses a semantic character labeling method of a natural language sentence, which is characterized in that Chinese syntax analysis and semantic character label are simultaneously realized by adopting a combined learning model. The invention can simultaneously output the syntax analysis result of one sentence and gives the semantic role labeling result of a predicative by using a combined model. Because semantic information is increased in a syntax analysis model in the combined learning model, a model trained is particularly suitable for a semantic role labeling task. Therefore, the semantic role label output by the model has high performance. Meanwhile, the performance between the result output by a single syntax analysis model and the syntax analysis result output by the combined model is not large. Particularly, the syntax analysis performance can also be improved by adding semantic information.

The invention discloses a method for constructing a semantic relation tree for natural language sentences, which is characterized in that the construction method comprises the following steps: a shortest path enclosure tree is firstly constructed for two entities required to establish the semantic relation tree in a sentence; if the structure type between the two entities is not of a predicative connecting structure, the shortest path enclosure tree is the semantic relation tree required; and if the structure between the two entities is the predicative connecting structure, a path enclosure tree related to a context is obtained through context information related to expansion and predicative to be taken as the semantic relation tree. The semantic relation tree obtained by using the method can not only cover key structural syntax information, but also reduce unnecessary noise and improve the performance of relation extraction on the premise that the relation extraction speed is almost not influenced. The result is universal through being compared by adopting the semantic relation tree.

A device for relation extraction in a natural language sentence having n words is suggested, the device comprising: a recurrent neural network for joint entity and relation extractions of entities and relations between the entities in the sentence, and an entity-relation table for storing entity labels for the entities and relation labels for the relations, wherein both the entity labels and the relation labels are defined as instances of binary relationships between certain words wi and wj in the sentence, with i∈[1, . . . , n], and j∈[1, . . . , n], wherein each of the entity labels is a first binary relationship for i=j, and wherein each of the relation labels is a second binary relationship for i≠j, wherein the recurrent neural network is configured to fill the cells of the entity-relation table and includes a forward neural network and a backward neural network.

The invention relates to a structured natural language query and knowledge system, which allows a user lack of programming experience to input database for query or input rules in the form of structured language sentences. The range of the sentences is defined by improved object-related query languages for preference, object-related algebra, or both the object-related query languages and the object-related algebra. Conditions which appear in the form of the natural language are defined by corresponding form query texts, the user is prompted to use defined instructions and conditions for arranging and structuring natural language sentences, the instructions selected by the user and the parameters thereof are used as verb phrases of structured natural language sentences to appear, and the conditions selected by the user and the parameters thereof are used as adjective phrases to appear. The sentences are divided and changed to query texts in the form of translation used for database query and rule treatment.

Disclosed herein is a computer implemented method and system for word sense disambiguation in a natural language sentence. The natural language sentence is parsed for identifying possible parts of speech for each term and identifying possible phrase structures. Terms comprising one or more linguistic roles are identified. The possible sense combinations for the terms with linguistic roles are identified. Emergent categories are applied to identify possible valid senses for each of the terms with identified linguistic roles. Linguistic role pairs are identified from among the terms identified with linguistic roles. The correspondence functions with the correspondence function types matching the identified linguistic role pairs are identified from an emergent categories database. The pair-wise senses for each term are compared with the identified linguistic roles to identify the possible sense combinations. The possible senses are inferred for each term with identified linguistic roles in the natural language sentence and previous sentences.

The invention discloses a deep learning-based structured information extraction method. The method comprises the following steps of: 1) constructing large-scale flag data by adoption of remote monitoring: constructing an extractor by using the existing Wikipedia to provide remote supervision, wherein a signature and a Wikipedia information box comprise entity-related structured facts, the facts are mentioned in a free text part of an entity webpage, and sentences which express the facts in the information box are used as training data; 2) integrating prior knowledge into a structured information extraction model, wherein the prior knowledge comprises type and phrase information, and labels of words in natural language sentences are related to the foregoing words and continuous words thereof; 3) utilizing the past and future input features by using a bidirectional hidden LSTM layer, and inputting each sentence into a bidirectional LSTM model in a sequence form; and 4) finally outputting a marker sequence, wherein the marker sequence comprises true or false markers and has a length same as the length of the input word sequence.

One embodiment of the present method and apparatus for natural language generation enables a language generation system to generate a grammatically correct natural language sentence by retrieving and adapting one or more stored sentences having semantic features similar to semantic features of a system-generated semantic representation. The retrieval of stored sentences is guided at least in part by the ease with which the stored sentences may be adapted to produce a grammatically correct sentence, enabling rapid generation of accurate output.

A method for analyzing a natural language sentence describing an action, to create an action structure to be used in creating an animation of the action, the method comprising: processing the natural language sentence to create a grammatical tree comprising an action word and its associated values; providing constructs for the action word, each of the constructs having parameter types for defining the action expressed by the action word; identifying from the constructs at least one construct wherein at least one of the parameter types can take on at least one of the associated values thereby defining a matching value; and recording the at least one of the parameter types from the at least one construct as well as the matching value, thereby creating the action structure.

The invention belongs to the field of computer natural language processing, and relates to a method for carrying out high-quality syntactic analysis on human natural language sentences, in particular to a syntactic analysis method based on sliding semantic string matching. The method is characterized in that when a rule base is built, hierarchic flattening transformation is firstly carried out on a usual phrase syntax tree, then semantic code labeling is carried out on the chunking information of each layer, and therefore the chunking rules of context-sensitive grammar of N elements are extracted; in the syntactic analysis process, the optimal chunking rules are matched to carry out stacking chunking type analysis through a sliding semantic string matching model; errors in the lower layer are found and corrected in the higher layer through an error correction template, and heuristic backtracking in the stacking chunking type syntactic analysis is achieved; template information is directly added to a semantic template index, and a machine can learn new syntactic rules immediately. The method solves the problems that a PCFG type syntactic analysis level can hardly be further improved and the correct chucking rules are hard to choose in the stacking chunking type syntactic analysis, and improves the existing syntactic analysis level.

The invention discloses a natural language analysis method, a natural language analysis device and an analyzer training method. The analysis method comprises the following steps: dividing the input natural language sentence into a plurality of sequences composed of the first type word and / or character which possibly is the constituent part of the second type word, wherein the first type word is the word except the second type word; computing the fraction of each sequence by a statistic probability model, wherein the model comprises the statistic the probability of the first type word in the context and the statistic probability of the character as the constituent part of the second type word; determining a candidate sequence according to the fraction; labeling the natural language sentence according to the candidate sequence; for the first type word which possibly exists in the candidate sequence, labeling word information obtained from the statistic probability model, and for the character which possibly exists except the possibly existent first type word, labeling character information obtained from the statistic probability model.

An anaphora analyzing apparatus is disclosed for automatically estimating an anaphora referential relation or an antecedent of a noun for use in a natural language sentence. A storage unit stores analyzed results outputted from an analyzer, and an antecedent candidate generator detects a target component required for anaphora analysis in accordance with the current analyzed results and the past analyzed results stored in the storage unit, and generates antecedent candidates corresponding to the target component. Then, a candidate rejecting section rejects unnecessary candidates having no potential for anaphora referential relation among the antecedent candidates by using a predetermined rejecting criterion, and outputs the remaining antecedent candidates. Further, a preference giving section calculates a predetermined estimated value for each of the remaining antecedent candidates by referring to an information table including predetermined estimation information obtained from a predetermined training tagged corpus, and gives the antecedent candidates preference in accordance with the calculated estimated value. Finally, a candidate deciding section decides a predetermined number of antecedent candidates based on a given preference in accordance with the preferenced antecedent candidates.

A method, computer program product, and apparatus for parsing consecutive sentences which includes tokenizing the words of the sentence and putting them through an iterative inductive processor. The processor has access to at least a first and second set of rules. The rules narrow the possible syntactic interpretations for the words in the sentence. After exhausting application of the first set of rules, the program moves to the second set of rules. The program reiterates back and forth between the sets of rules until no further reductions in the syntactic interpretation can be made. Thereafter, deductive token merging is performed if needed.

The invention discloses a method and a system for generating attention remote sensing image description based on high-low layer feature fusion. The method comprises the following steps: acquiring a to-be-detected remote sensing image; and obtaining natural language sentence description of the to-be-detected remote sensing image based on the to-be-detected remote sensing image and a pre-trained image description model, wherein the image description model is constructed by an encoder constructed by using a convolutional network, attention of high and low layer feature fusion and a decoder constructed by using a cyclic recursive network. According to the technical scheme provided by the invention, local detail information of image shallow features and global semantic information of high-levelfeatures are fully utilized, and global and local double consciousness is given to a traditional attention mechanism, so that semantic content expressed by the image and the mutual relation between the semantic content and the natural language sentence are understood more accurately, and the natural language sentence description with accurate content is generated for the remote sensing image.

The invention provides a generation method and device of a natural language sentence library. Through the generation method, according to a training data set, word information is obtained; a word vector of a preset dimension is adopted to convert the word information into a test set to be identified; and through a way that the test set to be identified in a RNN (Recurrent neural Network) model is trained to generate the natural language sentence library, the problems that a large quantity of manual intervention is required by a construction way of the natural language sentence library and an operation process is complex in the relevant art are solved so as to achieve effects that an identification rate is high, and the generation method and device is simple and easy in use and can meet the NLP (Natural Language Processing) business requirements of a questioning and answering system, a retrieval system, an expert system, on-line customer service, a mobile phone assistant, a private assistant and the like.

The invention discloses a dual-attention natural language inference method with context perception, for natural language sentence pairs with contextual information, using the method of multimodal learning and attention mechanism, take full advantage of that text, In order to understand the semantics of sentences and the interaction between two sentences under such semantics, and then to judge whether there is an inference relationship between two sentences, a method of sentence semantic representation is provided, which makes up for the shortcomings of the existing methods that do not considerthe situation of sentences.