CN112328751A - Method and device for processing text - Google Patents
Method and device for processing text Download PDFInfo
- Publication number
- CN112328751A CN112328751A CN202011396279.1A CN202011396279A CN112328751A CN 112328751 A CN112328751 A CN 112328751A CN 202011396279 A CN202011396279 A CN 202011396279A CN 112328751 A CN112328751 A CN 112328751A
- Authority
- CN
- China
- Prior art keywords
- text
- sentence
- processed
- matching
- vector
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 74
- 238000012545 processing Methods 0.000 title claims abstract description 53
- 239000013598 vector Substances 0.000 claims abstract description 267
- 238000000605 extraction Methods 0.000 claims description 55
- 238000013528 artificial neural network Methods 0.000 claims description 18
- 230000000306 recurrent effect Effects 0.000 claims description 13
- 238000004590 computer program Methods 0.000 claims description 10
- 230000008569 process Effects 0.000 description 13
- 238000010586 diagram Methods 0.000 description 11
- 230000011218 segmentation Effects 0.000 description 8
- 230000002457 bidirectional effect Effects 0.000 description 7
- 230000006870 function Effects 0.000 description 6
- 230000015654 memory Effects 0.000 description 6
- 238000004364 calculation method Methods 0.000 description 5
- 238000004891 communication Methods 0.000 description 5
- 230000007246 mechanism Effects 0.000 description 5
- 238000012549 training Methods 0.000 description 5
- 239000003795 chemical substances by application Substances 0.000 description 4
- 238000003058 natural language processing Methods 0.000 description 4
- 230000003287 optical effect Effects 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 3
- 238000013519 translation Methods 0.000 description 3
- 210000004556 brain Anatomy 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000000644 propagated effect Effects 0.000 description 2
- 230000006403 short-term memory Effects 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/3331—Query processing
- G06F16/334—Query execution
- G06F16/3344—Query execution using natural language analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/284—Lexical analysis, e.g. tokenisation or collocates
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Molecular Biology (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Evolutionary Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Databases & Information Systems (AREA)
- Machine Translation (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The embodiment of the disclosure discloses a method and a device for processing texts. One embodiment of the method comprises: acquiring a text to be processed; determining a text to be processed as a target text, and executing the following generation steps: respectively obtaining the characteristic vector of each sentence in the target text; processing the feature vectors of sentences in the target text by using a pre-trained text vector generation model to generate the feature vectors of the target text, wherein the text vector generation model comprises a first attention model; and selecting matching data from the matching data set as a matching result of the text to be processed according to the feature vector of the text to be processed. The embodiment can realize text matching according to the semantics of the text.
Description
Technical Field
The embodiment of the disclosure relates to the technical field of computers, in particular to a method and a device for processing texts.
Background
Information Retrieval (Information Retrieval) is one of the main ways users query and retrieve Information. Generally, information retrieval refers to a process of processing, organizing, and storing information in a certain manner, and then accurately finding out relevant information according to the needs of users. Currently, the commonly used information retrieval methods include various methods based on text matching.
Text matching is one of the important basic questions in current natural language processing, and can be applied not only to information retrieval, but also to many natural language processing tasks such as question-answering systems, dialogue systems, machine translation, and the like.
The specific process of the common text matching method generally includes extracting keywords of a text, storing the text and the corresponding keywords in an associated manner, and then retrieving a text matched with the text input by a user by matching the text input by the user with the respective keywords stored in advance. Or, an index is established in advance for the stored text, and then the text input by the user is matched based on the index.
These existing text matching methods generally belong to exact matching, i.e. the text input by the user must contain text that is exactly identical to the stored keywords. Correspondingly, a new text matching method is gradually appeared as fuzzy matching. Currently, a text matching method based on fuzzy matching is generally implemented by performing synonym or near synonym replacement on some keywords in the text.
Disclosure of Invention
The embodiment of the disclosure provides a method and a device for processing texts.
In a first aspect, an embodiment of the present disclosure provides a method for processing text, the method including: acquiring a text to be processed; determining a text to be processed as a target text, and executing the following generation steps: respectively obtaining the characteristic vector of each sentence in the target text; processing the feature vectors of sentences in the target text by using a pre-trained text vector generation model to generate the feature vectors of the target text, wherein the text vector generation model comprises a first attention model; and selecting matching data from the matching data set as a matching result of the text to be processed according to the feature vector of the text to be processed.
In some embodiments, separately obtaining a feature vector of each sentence in the target text comprises: for a sentence in the target text, word vectors of words in the sentence are respectively obtained, and the word vectors of the words in the sentence are processed by utilizing a pre-trained sentence vector generation model to generate a feature vector of the sentence, wherein the sentence vector generation model comprises a second attention model.
In some embodiments, the matching dataset is a matching text set; and selecting matching data from the matching data set as a matching result of the text to be processed according to the feature vector of the text to be processed, wherein the matching result comprises the following steps: respectively obtaining a feature vector of each matched text in the matched text set, wherein the feature vector of the matched text is obtained by determining the matched text as a target text in advance and executing the generation step; determining the similarity between the feature vectors of the texts to be processed and the feature vectors of each matched text in the matched text set; and selecting the matching text from the matching text set as a matching result of the text to be processed according to the corresponding similarity.
In some embodiments, the text vector generation model further comprises a first feature extraction network, wherein an output of the first feature extraction network is an input to the first attention model.
In some embodiments, the sentence vector generation model further comprises a second feature extraction network, wherein an output of the second feature extraction network is an input of the second attention model.
In some embodiments, the first feature extraction network is a recurrent neural network.
In some embodiments, the second feature extraction network is a recurrent neural network.
In a second aspect, an embodiment of the present disclosure provides an apparatus for processing text, the apparatus including: an acquisition unit configured to acquire a text to be processed; a generating unit configured to determine a text to be processed as a target text, and execute the following generating steps: respectively obtaining the characteristic vector of each sentence in the target text; processing the feature vectors of sentences in the target text by using a pre-trained text vector generation model to generate the feature vectors of the target text, wherein the text vector generation model comprises a first attention model; and the selecting unit is configured to select the matching data from the matching data set as a matching result of the text to be processed according to the feature vector of the text to be processed.
In some embodiments, the generating unit is further configured to: for a sentence in the target text, word vectors of words in the sentence are respectively obtained, and the word vectors of the words in the sentence are processed by utilizing a pre-trained sentence vector generation model to generate a feature vector of the sentence, wherein the sentence vector generation model comprises a second attention model.
In some embodiments, the matching dataset is a matching text set; and the selecting unit is further configured to: respectively obtaining a feature vector of each matched text in the matched text set, wherein the feature vector of the matched text is obtained by determining the matched text as a target text in advance and executing the generation step; determining the similarity between the feature vectors of the texts to be processed and the feature vectors of each matched text in the matched text set; and selecting the matching text from the matching text set as a matching result of the text to be processed according to the corresponding similarity.
In some embodiments, the text vector generation model further comprises a first feature extraction network, wherein an output of the first feature extraction network is an input to the first attention model.
In some embodiments, the sentence vector generation model further comprises a second feature extraction network, wherein an output of the second feature extraction network is an input of the second attention model.
In some embodiments, the first feature extraction network is a recurrent neural network.
In some embodiments, the second feature extraction network is a recurrent neural network.
In a third aspect, an embodiment of the present disclosure provides an electronic device, including: one or more processors; storage means for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors are caused to implement the method as described in any implementation of the first aspect.
In a fourth aspect, embodiments of the present disclosure provide a computer-readable medium on which a computer program is stored, which computer program, when executed by a processor, implements the method as described in any of the implementations of the first aspect.
The method and the device for processing the text provided by the embodiment of the disclosure utilize the text vector generation model comprising the attention model to process the feature vector of each sentence in the text to be processed so as to generate the feature vector of the text to be processed, so that the generated feature vector of the text to be processed contains semantic information of each sentence in the text to be processed, namely, the semantic of different sentences to be processed respectively has different importance, and the same sentence respectively has semantic problems of different semantics in contexts formed by different texts, and the like, so that the feature vector of the text to be processed is utilized to complete the matching of the text to be processed, and the accuracy of the matching result of the text to be processed can be improved.
Drawings
Other features, objects and advantages of the disclosure will become more apparent upon reading of the following detailed description of non-limiting embodiments thereof, made with reference to the accompanying drawings in which:
FIG. 1 is an exemplary system architecture diagram in which one embodiment of the present disclosure may be applied;
FIG. 2 is a flow diagram for one embodiment of a method for processing text, according to the present disclosure;
FIG. 3 is a flow diagram of yet another embodiment of a method for processing text in accordance with the present disclosure;
FIG. 4 is a schematic diagram of one application scenario of a method for processing text in accordance with an embodiment of the present disclosure;
FIG. 5 is one exemplary network architecture of a sentence vector generation model and a text vector generation model in accordance with embodiments of the present disclosure;
FIG. 6 is a schematic block diagram illustrating one embodiment of an apparatus for processing text according to the present disclosure;
FIG. 7 is a schematic structural diagram of an electronic device suitable for use in implementing embodiments of the present disclosure.
Detailed Description
The present disclosure is described in further detail below with reference to the accompanying drawings and examples. It is to be understood that the specific embodiments described herein are merely illustrative of the relevant invention and not restrictive of the invention. It should be noted that, for convenience of description, only the portions related to the related invention are shown in the drawings.
It should be noted that, in the present disclosure, the embodiments and features of the embodiments may be combined with each other without conflict. The present disclosure will be described in detail below with reference to the accompanying drawings in conjunction with embodiments.
Fig. 1 illustrates an exemplary architecture 100 to which embodiments of a method for processing text or an apparatus for processing text of the present disclosure may be applied.
As shown in fig. 1, the system architecture 100 may include terminal devices 101, 102, 103, a network 104, and a server 105. The network 104 serves as a medium for providing communication links between the terminal devices 101, 102, 103 and the server 105. Network 104 may include various connection types, such as wired, wireless communication links, or fiber optic cables, to name a few.
The terminal devices 101, 102, 103 interact with a server 105 via a network 104 to receive or send messages or the like. Various client applications may be installed on the terminal devices 101, 102, 103. Such as search-type applications, browser-type applications, instant messaging tools, social platform software, shopping-type applications, information flow-type applications, and so forth.
The terminal apparatuses 101, 102, and 103 may be hardware or software. When the terminal devices 101, 102, 103 are hardware, they may be various electronic devices having a display screen, including but not limited to smart phones, tablet computers, e-book readers, laptop portable computers, desktop computers, and the like. When the terminal apparatuses 101, 102, 103 are software, they can be installed in the electronic apparatuses listed above. It may be implemented as multiple pieces of software or software modules (e.g., multiple pieces of software or software modules to provide distributed services) or as a single piece of software or software module. And is not particularly limited herein.
The server 105 may be a server providing various services, such as a backend server providing backend support for the terminal devices 101, 102, 103. The server 105 may receive the text to be processed from the terminal devices 101, 102, 103, generate a feature vector of the text to be processed, and then obtain a matching result of the text to be processed by using the feature vector of the text to be processed. Further, the server 105 may also return the matching result of the text to be processed to the terminal devices 101, 102, 103.
It should be noted that the to-be-processed text may also be directly stored locally in the server 105, and the server 105 may directly extract and process the to-be-processed text stored locally, in this case, the terminal devices 101, 102, and 103 and the network 104 may not be present.
It should be noted that the method for processing text provided by the embodiment of the present disclosure is generally performed by the server 105, and accordingly, the apparatus for processing text is generally disposed in the server 105.
It should be further noted that the terminal devices 101, 102, and 103 may also have text processing applications installed therein, and the terminal devices 101, 102, and 103 may also process the text to be processed based on the text processing applications, in this case, the method for processing the text may also be executed by the terminal devices 101, 102, and 103, and accordingly, the device for processing the text may also be installed in the terminal devices 101, 102, and 103. At this point, the exemplary system architecture 100 may not have the server 105 and the network 104.
The server 105 may be hardware or software. When the server 105 is hardware, it may be implemented as a distributed server cluster composed of a plurality of servers, or may be implemented as a single server. When the server 105 is software, it may be implemented as multiple pieces of software or software modules (e.g., multiple pieces of software or software modules used to provide distributed services), or as a single piece of software or software module. And is not particularly limited herein.
It should be understood that the number of terminal devices, networks, and servers in fig. 1 is merely illustrative. There may be any number of terminal devices, networks, and servers, as desired for implementation.
With continued reference to FIG. 2, a flow 200 of one embodiment of a method for processing text in accordance with the present disclosure is shown. The method for processing text comprises the following steps:
In the present embodiment, the text may be any type and content of text. The text to be processed may refer to any text. For example, the text to be processed may be an article, a description, or the like. In general, the text to be processed may be composed of several sentences. Each sentence may be composed of several words and/or phrases.
An executing agent (such as a server shown in fig. 1) of the method for processing the text can acquire the text to be processed from a storage device (such as terminal devices 101, 102, 103 shown in fig. 1, a connected database, a third-party data platform and the like) local to the executing agent or connected with the executing agent.
As an example, the user may input a text through a terminal device used by the user, and transmit the input text to the execution main body. At this time, the execution body may receive, as the text to be processed, a text transmitted by the user using the terminal device used by the user.
Step 202, determining the text to be processed as the target text, and executing the following generating steps 2021 and 2022:
In this embodiment, the feature vector of a sentence may refer to a vector obtained by encoding (embedding) the sentence and used for representing the sentence. The executing agent may obtain the feature vectors of the sentences in the target text from a local or other storage device.
Specifically, for the target text, the feature vector of each sentence can be obtained through the following steps:
step one, sentence splitting processing is carried out on a target text to obtain a sentence set.
In this step, the target text may be sentence-divided by using various sentence-dividing methods existing in natural language processing. For example, a sentence segmentation method based on a regular expression is adopted to segment the target text, and some sentence segmentation tools in an open source are utilized to segment the target text, and so on.
And step two, for each sentence in the sentence set, coding the sentence to obtain a feature vector of the sentence.
In this step, each Sentence in the Sentence set may be encoded by using various existing Sentence encoding (sequence Embedding) methods, and a feature vector that can represent each Sentence is obtained.
For example, a representation of learning each sentence can be derived based on a pre-trained sentence representation learning model. The sentence representation learning model may be some existing sentence learning representation models such as PV-DM (Paragraph Vector: a Distributed memory model), PV-DBOW (Paragraph Vector with word representation: Distributed bag of words), or may be a model constructed based on these existing sentence learning representation models.
The feature vectors of the sentences in the target text can be generated in parallel to improve the processing speed.
It should be noted that the feature vector of each sentence in the target text may be generated by other electronic devices by executing the above steps, or may be generated by the execution subject by executing the above steps.
In this embodiment, a feature vector of a text may refer to a vector obtained by encoding (embedding) the text and may be used to represent the text. The text vector generation model is used for generating a feature vector for representing the text according to the feature vectors respectively corresponding to the sentences in the text.
The text vector generation model may include a first attention model. Wherein the first Attention model may be an Attention model implemented based on an Attention Mechanism (Attention Mechanism). The attention mechanism is a data processing method in machine learning, and is applied to various different types of machine learning tasks such as natural language processing, image recognition, voice recognition and the like. The attention mechanism is a simulation of the attention behavior of a human in reading and hearing. In particular, the simulated human brain differs in the concentration distribution at each location within the scene in the reading or listening scene, and the human brain preferentially captures information from the scene that it deems useful, while ignoring or discarding low value information.
Taking text as an example, each sentence in the text may have a different importance to the semantics of the text. For example, the presence or absence of some sentences has little or no semantic effect on the text. For another example, some sentences, if deleted, may directly affect the understanding of the text. In addition, each sentence in the text typically has a context, and the sentence has a semantic association with the sentences of which the context is. For example, the semantics of some sentences need to be understood in conjunction with the semantics of the first few sentences or the last few sentences.
Based on this, the first attention model may be used to apply different weights to the sentences in the target text to strengthen the sentences having a stronger influence on the semantics of the target text and weaken the sentences having a weaker influence on the semantics of the target text, so as to learn the dependency relationship between the sentence structure of the target text and each sentence, so that the generated feature vector of the target text can more accurately express the semantics of the target text.
The first attention model may be various attention models. For example, the first attention model may be an attention model used by an encoder and a decoder in a common translation model, an attention model used in various pre-training models (e.g., BERT, etc.), and so on. For another example, the first attention model may be a model constructed by a technician based on an existing attention model.
The text vector generation model may perform different attention calculations on the feature vectors of the sentences in the target text by using the first attention model, and then generate the feature vectors of the target text by using the feature vectors corresponding to the sentences output by the first attention model.
The specific method for generating the feature vector of the target text can be flexibly set according to the actual application requirements. For example, feature vectors respectively corresponding to sentences output by the first attention model may be directly combined in sequence to generate a feature vector of the target text. For another example, feature vectors corresponding to respective sentences output by the first attention model may be weighted and calculated to generate feature vectors of the target text.
And 203, selecting matching data from the matching data set as a matching result of the text to be processed according to the feature vector of the text to be processed.
In this embodiment, the matching data set may be composed of at least one matching data. The matching data may be data of various types and contents. For example, matching data includes, but is not limited to, text, images, video, audio, and the like.
The matching result of the text to be processed may refer to data matched with the text to be processed. According to different application requirements and application scenes, the specific matching tasks can be flexibly set. For example, the matching task may be to pick data similar to the text to be processed. For another example, the matching task may be to select data that differs greatly from the text to be processed.
Specifically, the matching data can be selected from the matching data set as the matching result of the text to be processed according to the specific matching task by calculating the similarity between the feature vector of the text to be processed and the feature vector of each matching data in the matching data set.
Wherein the feature vector of each matching data in the matching data set may be preset. According to different types of the matching data, different methods can be adopted to obtain the feature vector of each matching data in the matching data set in advance. For example, the feature vectors of the matching data may be extracted by using various existing feature extraction methods (e.g., a feature extraction method based on a feature extraction network, etc.).
Optionally, for the matching data in the matching data set, a text corresponding to the matching data may be determined first, then the text corresponding to the matching data is determined as a target text, and the feature vector of the text corresponding to the matching data is generated by using the generating step.
For example, when the matching data is a text, the feature vector of the matching data may be generated by directly using the above-described generation step. Further, the similarity between the feature vector of the text to be processed and the feature vector of the matching data may be calculated as the similarity between the two texts.
For another example, when the matching data is audio, the text corresponding to the audio may be recognized by using various existing speech recognition methods, and then the feature vector of the text corresponding to the audio may be obtained by using the above-described generation step. Specifically, feature vectors of sentences in the text corresponding to the audio are respectively obtained, and then the feature vectors of the sentences in the text corresponding to the audio are processed by utilizing a pre-trained text vector generation model to generate the feature vectors of the text corresponding to the audio. Further, the similarity between the feature vector of the text to be processed and the feature vector of the text corresponding to the audio can be calculated, and the obtained similarity is determined as the similarity between the text to be processed and the audio.
The similarity between the feature vectors can be determined by using various existing similarity calculation methods. For example, a calculation method based on cosine similarity, a calculation method based on euclidean distance, and the like.
In some alternative implementations of the present embodiment, the matching data set may be a matching text set. I.e. the matching data set consists of several texts. At this time, after the feature vector of the text to be processed is obtained, matching data can be selected from the matching text set as a matching result of the text to be processed through the following steps:
step one, respectively obtaining the characteristic vector of each matched text in the matched text set.
In this step, the feature vector of each matching text in the matching text set may be obtained and stored as a target text in advance by performing the above-described generating step. Specifically, for each matching text in the matching text set, the feature vector of each sentence in the matching text may be obtained, and then the feature vector of each sentence in the matching text is processed by using a pre-trained text vector generation model to generate the feature vector of the matching text.
And step two, determining the similarity between the feature vectors of the texts to be processed and the feature vectors of the matched texts in the matched text set.
And step three, selecting the matching texts from the matching text set as matching results of the texts to be processed according to the corresponding similarity.
In this step, the matching text may be selected from the matching text set as the matching result of the text to be processed according to the actual matching task. For example, if the matching task is to select a text similar to the text to be processed, several matching texts may be selected from the matching text set as the matching result of the text to be processed according to the sequence of similarity from large to small.
It should be noted that the feature vectors of the respective matching data in the matching text set can be generated in parallel, thereby increasing the text processing speed.
In this way, the feature vector capable of accurately expressing the semantics of the text to be processed can be generated by using the text vector generation model, so that the text which is matched with or not matched with the semantics of the text to be processed can be accurately retrieved from the matched text set by calculating the similarity between the feature vectors of the text.
In some optional implementations of this implementation, the text vector generation model may further include a first feature extraction network. The first feature extraction network may be various networks for feature extraction. The output of the first feature extraction network may be an input to the first attention model.
At this time, the first feature extraction network may further perform feature extraction on the feature vector of each sentence in the input target text to obtain a feature vector corresponding to each sentence after further feature extraction, and then the first attention model may process the feature vectors corresponding to each sentence after further feature extraction to generate the feature vector of the target text.
Alternatively, the first feature extraction Network may be a Recurrent Neural Network (RNN). For example, the first feature extraction network may be a unidirectional or bidirectional LSTM (Long Short-Term Memory), a unidirectional or bidirectional GRU (Gated recursive Unit), or the like.
Therefore, the network structure of the text vector generation model can be flexibly set according to the actual text matching task so as to more pertinently process the text to obtain the feature vector which can more fully and comprehensively express the semantics of the text and is convenient for the subsequent text matching,
at present, common text matching methods are usually realized based on keyword matching, and semantic information contained in a text cannot be considered by the methods, so that a matching result can be directly influenced. For the text, the method provided by the embodiment of the disclosure generates the feature vector of the text according to the feature vector of each sentence in the text by using the attention model, so that the feature vector of the text contains semantic information of the text, and semantic association between the sentence structure containing the text and the sentence of the text, and further performs text matching based on the feature vector of the text, thereby improving the accuracy of the matching result.
With further reference to FIG. 3, a flow 300 of yet another embodiment of a method for processing text is shown. The flow 300 of the method for processing text comprises the steps of:
In this embodiment, a word vector of a word may refer to a vector obtained by encoding the word and used for representing the word. Specifically, the sentence may be segmented by using various existing segmentation methods (e.g., a dictionary-based segmentation method, a statistical-based segmentation method, a deep learning-based segmentation method, etc.), so as to obtain a word set corresponding to the sentence. Alternatively, each word in the sentence may be taken as a word to perform word segmentation processing, so as to obtain a word set corresponding to the sentence.
For each sentence in the target text, the word vector of each word in the word set corresponding to the sentence can be obtained by utilizing various existing methods for generating the word vector. For example, a word vector for each word may be generated based on one-hot encoding. As another example, a Word vector for each Word may be generated using the Word2Vec model.
Alternatively, for each sentence in the target text, word vectors for the individual words in the sentence may be generated in parallel to increase processing speed.
The sentence vector generation model may include a second attention model. Wherein the second attention model may be an attention model implemented based on an attention mechanism. The second attention model may be used to apply different weights to the words in a sentence to strengthen the words having stronger influence on the semantics of the sentence and weaken the words having weaker influence on the semantics of the sentence, so as to learn the dependency relationship between the sentence structure of the sentence and the words constituting the sentence, so that the generated feature vector of the sentence can more accurately express the semantics of the sentence.
The second attention model may be various attention models. For example, the second attention model may be an attention model used by an encoder and a decoder in a common translation model, an attention model used in various pre-training models (e.g., BERT, etc.), and so on. For another example, the second attention model may be a model constructed by a technician based on an existing attention model. The first attention model and the second attention model may have the same model structure or different model structures.
The sentence vector generation model may perform different attention calculations on word vectors of each word in each sentence by using the second attention model, and then generate a feature vector of the sentence by using word vectors corresponding to each word output by the second attention model.
The specific method for generating the feature vector of the sentence can be flexibly set according to the actual application requirement. For example, word vectors respectively corresponding to words in the sentence output by the second attention model may be directly combined in sequence to generate a feature vector of the sentence. For another example, word vectors corresponding to respective words in the sentence output by the second attention model may be weighted and calculated to generate a feature vector of the sentence.
It should be noted that, in the present disclosure, for convenience of describing the two attention models, the two attention models are named as a first attention model and a second attention model, respectively. It will be understood by those skilled in the art that the first and second do not constitute a particular limitation of the attention model.
For a sentence, each word in the sentence may have a different importance to the semantics of the sentence. For example, the presence or absence of some words has a weak or even no semantic effect on the sentence. For another example, if some words are deleted, the understanding of the sentence will be directly affected. In addition, semantic relevance is generally provided between the words in the sentence. For example, the semantics of some words in the sentence need to be understood in conjunction with the semantics of the first or last words of the word.
Therefore, the feature vector of each sentence in the text to be processed is processed by using the attention model to generate the feature vector of each sentence, so that the feature vector of each sentence contains semantic information of each word forming the sentence, the influence of different words in one sentence on the semantics of the sentence is considered, the problems that one word has different semantics in the contexts formed by different sentences and the like are solved, the accuracy of the feature vector of the sentence is improved, the feature vector of the text to be processed is generated by using the generated feature vector of each sentence in the text to be processed, the matching of the text to be processed is completed by using the feature vector of the text to be processed, and the accuracy of the matching result of the text to be processed can also be improved.
And step 3022, inputting the feature vector of each sentence in the target text into a pre-trained text vector generation model to obtain the feature vector of the target text.
Step 303, determining similarity between the feature vector of the text to be processed and the feature vectors of the matching texts in the matching text set.
In this embodiment, the feature vector of each matching text in the matching text set may be generated in advance and stored. Moreover, the generation method of the feature vector of each matching text in the matching text set may be the same as the generation method of the feature vector of the text to be processed. That is, each matching text in the matching text set can be used as a target text, and the feature vector of each matching text is obtained by executing the steps 3021 and 3022.
And 304, selecting the matching text from the matching text set as a matching result of the text to be processed according to the corresponding similarity.
In some optional implementations of this implementation, the sentence vector generation model described above may further include a second feature extraction network. The second feature extraction network may be various networks for feature extraction. The output of the second feature extraction network may be used as an input to the second attention model.
At this time, the second feature extraction network may further perform feature extraction on the word vector of each word in each sentence in the input target text to obtain a word vector corresponding to each word after the further feature extraction, and then the second attention model may process the word vectors corresponding to each word after the further feature extraction to generate a feature vector of each sentence in the target text.
Alternatively, the second feature extraction Network may be a Recurrent Neural Network (RNN). For example, the second feature extraction network may be a unidirectional or bidirectional LSTM (Long Short-Term Memory), a unidirectional or bidirectional GRU (Gated recursive Unit), or the like.
The network structures of the first feature extraction network and the second feature extraction network may be the same or different. It should be noted that, in the present disclosure, for convenience of describing the two feature extraction networks, the two feature extraction networks are named as a first feature extraction network and a second feature extraction network, respectively. Those skilled in the art will appreciate that the first and second do not constitute a specific limitation on the feature extraction network.
The text vector generation model and the sentence vector generation model can be obtained by training based on a machine learning method. Specifically, the training process can be flexibly adjusted according to actual application requirements and application scenarios. For example, Adam can be used for parameter optimization, Dropout technology can be used for avoiding overfitting, and the like, so that the model training process is improved, and a text vector generation model and a sentence vector generation model which meet requirements are obtained.
The content that is not described in detail in the steps 301-304 can refer to the related description in the corresponding embodiment of fig. 2, and is not described herein again.
With continued reference to fig. 4, fig. 4 is an illustrative application scenario 400 of the method for processing text in accordance with the present embodiment. In the application scenario of fig. 4, a search text 401 input by a user may be received, and then sentence division processing is performed on the search text 401, so as to obtain a sentence set 402 corresponding to the search text. Then, the sentence vector generation model 4031 is used to generate sentence vectors corresponding to each sentence in the sentence set 402, so as to obtain a sentence vector set 404.
Specifically, as shown by reference numeral 403 in the figure, for each sentence in the sentence set 402, a word segmentation process is performed on the sentence to obtain a word set corresponding to the sentence. And then extracting the word vector of each word in the word set by using a word vector extraction method to obtain a word vector set. Then, the word vector set is input into the sentence vector generation model 4031 trained in advance, and the sentence vector of the sentence is obtained.
The sentence vector set 404 may then be input to a pre-trained text vector generation model 405, generating a text vector 406 of the search text 401. Thereafter, the similarity between the text vector 406 and the feature vector corresponding to each text in the pre-constructed text library 407 can be calculated respectively. The feature vector of each text in the text library 407 may be generated and stored in advance by using the sentence vector generation model 4031 and the text vector generation model 405.
Further, a text with a corresponding similarity greater than a preset similarity threshold may be selected from the text library 407 as a text related to the search text 401, and the text related to the search text 401 may be returned to the user as a search result 408.
Referring now to fig. 5, fig. 5 is an exemplary network architecture 500 of the sentence vector generation model and the text vector generation model according to the present embodiment.
As shown in fig. 5, the sentence vector generation model 501 may include a bidirectional recurrent neural network and an attention model. Specifically, word vectors corresponding to respective words of a sentence may be input to the sentence vector generation model 501, and a feature vector of the sentence, that is, a sentence vector may be generated.
The text vector generation model 502 may include a bidirectional recurrent Neural Network (ANN), an attention model, and an Artificial Neural Network (ANN) 5021. Specifically, sentence vectors corresponding to respective sentences of a text may be input to the text vector generation model 502, and feature vectors of the text, that is, text vectors, may be generated.
The text vector generation model 502 may further include various artificial neural networks 5021 according to actual application requirements, in addition to the bidirectional recurrent neural network and the attention model. For example, the text vector generative model 502 may also include an artificial neural network 5021 for dimension reduction, and the like.
It should be noted that fig. 5 shows only one exemplary network structure of the sentence vector generation model and the text vector generation model. Technicians can flexibly adjust the network structures of the sentence vector generation model 501 and the text vector generation model 502 according to actual application requirements or different application scenes.
The above-described embodiments of the present disclosure provide methods for generating not only feature vectors of text from feature vectors of respective sentences of text using an attention model, but also feature vectors of each sentence from word vectors of respective words of the sentence using an attention model. Therefore, the finally generated text feature vector can represent the semantic association between the sentence structure of the text and the sentences of the text, and can also represent the semantic association between the composition structure of each sentence and the words of the text, so that the semantic characteristics of the text can be more comprehensively represented, and the accuracy of the matching result obtained based on the feature vector of the text is further improved.
With further reference to fig. 6, as an implementation of the methods shown in the above figures, the present disclosure provides an embodiment of an apparatus for processing text, which corresponds to the method embodiment shown in fig. 2, and which is particularly applicable in various electronic devices.
As shown in fig. 6, the apparatus 600 for processing text provided by the present embodiment includes an obtaining unit 601, a generating unit 602, and a selecting unit 603. Wherein the obtaining unit 601 is configured to obtain a text to be processed; the generating unit 602 is configured to determine a text to be processed as a target text, and perform the following generating steps: respectively obtaining the characteristic vector of each sentence in the target text; processing the feature vectors of sentences in the target text by using a pre-trained text vector generation model to generate the feature vectors of the target text, wherein the text vector generation model comprises a first attention model; the selecting unit 603 is configured to select matching data from the matching data set as a matching result of the text to be processed according to the feature vector of the text to be processed.
In the present embodiment, in the apparatus 600 for processing text: the specific processing of the obtaining unit 601, the generating unit 602, and the selecting unit 603 and the technical effects thereof can refer to the related descriptions of step 201, step 202, and step 203 in the corresponding embodiment of fig. 2, which are not repeated herein.
In some optional implementations of the present embodiment, the generating unit 602 is further configured to: for a sentence in the target text, word vectors of words in the sentence are respectively obtained, and the word vectors of the words in the sentence are processed by utilizing a pre-trained sentence vector generation model to generate a feature vector of the sentence, wherein the sentence vector generation model comprises a second attention model.
In some optional implementations of this embodiment, the matching dataset is a matching text set; and the selecting unit 603 is further configured to: respectively obtaining a feature vector of each matched text in the matched text set, wherein the feature vector of the matched text is obtained by determining the matched text as a target text in advance and executing the generation step; determining the similarity between the feature vectors of the texts to be processed and the feature vectors of each matched text in the matched text set; and selecting the matching text from the matching text set as a matching result of the text to be processed according to the corresponding similarity.
In some optional implementations of this embodiment, the text vector generation model further includes a first feature extraction network, wherein an output of the first feature extraction network is an input of the first attention model.
In some optional implementations of this embodiment, the sentence vector generation model further comprises a second feature extraction network, wherein an output of the second feature extraction network is an input of the second attention model.
In some optional implementations of this embodiment, the first feature extraction network is a recurrent neural network.
In some optional implementations of this embodiment, the second feature extraction network is a recurrent neural network.
According to the device provided by the embodiment of the disclosure, the text to be processed is acquired through the acquisition unit; the generation unit determines a text to be processed as a target text, and executes the following generation steps: respectively obtaining the characteristic vector of each sentence in the target text; processing the feature vectors of sentences in the target text by using a pre-trained text vector generation model to generate the feature vectors of the target text, wherein the text vector generation model comprises a first attention model; and the selecting unit selects the matching data from the matching data set as the matching result of the text to be processed according to the feature vector of the text to be processed. Therefore, the feature vector of the text can contain semantic information of the text, and text matching is performed based on the feature vector of the text, so that the accuracy of a matching result can be improved.
Referring now to FIG. 7, a block diagram of an electronic device (e.g., the server of FIG. 1) 700 suitable for use in implementing embodiments of the present disclosure is shown. The server shown in fig. 7 is only an example, and should not bring any limitation to the functions and the scope of use of the embodiments of the present disclosure.
As shown in fig. 7, electronic device 700 may include a processing means (e.g., central processing unit, graphics processor, etc.) 701 that may perform various appropriate actions and processes in accordance with a program stored in a Read Only Memory (ROM)702 or a program loaded from storage 708 into a Random Access Memory (RAM) 703. In the RAM703, various programs and data necessary for the operation of the electronic apparatus 700 are also stored. The processing device 701, the ROM 702, and the RAM703 are connected to each other by a bus 704. An input/output (I/O) interface 705 is also connected to bus 704.
Generally, the following devices may be connected to the I/O interface 705: input devices 706 including, for example, a touch screen, touch pad, keyboard, mouse, camera, microphone, accelerometer, gyroscope, etc.; an output device 707 including, for example, a Liquid Crystal Display (LCD), a speaker, a vibrator, and the like; storage 708 including, for example, magnetic tape, hard disk, etc.; and a communication device 709. The communication means 709 may allow the electronic device 700 to communicate wirelessly or by wire with other devices to exchange data. While fig. 7 illustrates an electronic device 700 having various means, it is to be understood that not all illustrated means are required to be implemented or provided. More or fewer devices may alternatively be implemented or provided. Each block shown in fig. 7 may represent one device or may represent multiple devices as desired.
In particular, according to an embodiment of the present disclosure, the processes described above with reference to the flowcharts may be implemented as computer software programs. For example, embodiments of the present disclosure include a computer program product comprising a computer program embodied on a computer readable medium, the computer program comprising program code for performing the method illustrated in the flow chart. In such embodiments, the computer program may be downloaded and installed from a network via the communication means 709, or may be installed from the storage means 708, or may be installed from the ROM 702. The computer program, when executed by the processing device 701, performs the above-described functions defined in the methods of embodiments of the present disclosure.
It should be noted that the computer readable medium described in the embodiments of the present disclosure may be a computer readable signal medium or a computer readable storage medium or any combination of the two. A computer readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples of the computer readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In embodiments of the disclosure, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. In embodiments of the present disclosure, however, a computer readable signal medium may comprise a propagated data signal with computer readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take many forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to: electrical wires, optical cables, RF (radio frequency), etc., or any suitable combination of the foregoing.
The computer readable medium may be embodied in the electronic device; or may exist separately without being assembled into the electronic device. The computer readable medium carries one or more programs which, when executed by the electronic device, cause the electronic device to: acquiring a text to be processed; determining a text to be processed as a target text, and executing the following generation steps: respectively obtaining the characteristic vector of each sentence in the target text; processing the feature vectors of sentences in the target text by using a pre-trained text vector generation model to generate the feature vectors of the target text, wherein the text vector generation model comprises a first attention model; and selecting matching data from the matching data set as a matching result of the text to be processed according to the feature vector of the text to be processed.
Computer program code for carrying out operations for embodiments of the present disclosure may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C + +, and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet service provider).
The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
The units described in the embodiments of the present disclosure may be implemented by software or hardware. The described units may also be provided in a processor, and may be described as: a processor includes an acquisition unit, a generation unit, and a selection unit. The names of these units do not in some cases constitute a limitation on the unit itself, and for example, the acquiring unit may also be described as a "unit that acquires text to be processed".
The foregoing description is only exemplary of the preferred embodiments of the disclosure and is illustrative of the principles of the technology employed. It will be appreciated by those skilled in the art that the scope of the invention in the embodiments of the present disclosure is not limited to the specific combination of the above-mentioned features, but also encompasses other embodiments in which any combination of the above-mentioned features or their equivalents is made without departing from the inventive concept as defined above. For example, the above features and (but not limited to) technical features with similar functions disclosed in the embodiments of the present disclosure are mutually replaced to form the technical solution.
Claims (10)
1. A method for processing text, comprising:
acquiring a text to be processed;
determining the text to be processed as a target text, and executing the following generation steps: respectively obtaining the characteristic vector of each sentence in the target text; processing the feature vectors of sentences in a target text by using a pre-trained text vector generation model to generate the feature vectors of the target text, wherein the text vector generation model comprises a first attention model;
and selecting matching data from a matching data set as a matching result of the text to be processed according to the feature vector of the text to be processed.
2. The method of claim 1, wherein the separately obtaining feature vectors of sentences in the target text comprises:
for a sentence in a target text, respectively obtaining word vectors of words in the sentence, and processing the word vectors of the words in the sentence by using a pre-trained sentence vector generation model to generate a feature vector of the sentence, wherein the sentence vector generation model comprises a second attention model.
3. The method of claim 1, wherein the matching data set is a matching text set; and
the selecting matching data from a matching data set as the matching result of the text to be processed according to the feature vector of the text to be processed comprises the following steps:
respectively obtaining the feature vectors of all the matched texts in the matched text set, wherein the feature vectors of the matched texts are obtained by determining the matched texts as target texts in advance and executing the generating step;
determining the similarity between the feature vectors of the texts to be processed and the feature vectors of the matched texts in the matched text set;
and selecting a matching text from the matching text set as a matching result of the text to be processed according to the corresponding similarity.
4. The method of one of claims 1-3, wherein the text vector generation model further comprises a first feature extraction network, wherein an output of the first feature extraction network is an input of the first attention model.
5. The method of claim 2, wherein the sentence vector generation model further comprises a second feature extraction network, wherein an output of the second feature extraction network is an input to the second attention model.
6. The method of claim 4, wherein the first feature extraction network is a recurrent neural network.
7. The method of claim 5, wherein the second feature extraction network is a recurrent neural network.
8. An apparatus for processing text, comprising:
an acquisition unit configured to acquire a text to be processed;
a generating unit configured to determine the text to be processed as a target text, and execute the following generating steps: respectively obtaining the characteristic vector of each sentence in the target text; processing the feature vectors of sentences in a target text by using a pre-trained text vector generation model to generate the feature vectors of the target text, wherein the text vector generation model comprises a first attention model;
and the selecting unit is configured to select matching data from the matching data set as a matching result of the text to be processed according to the feature vector of the text to be processed.
9. An electronic device, comprising:
one or more processors;
a storage device having one or more programs stored thereon;
when executed by the one or more processors, cause the one or more processors to implement the method of any one of claims 1-7.
10. A computer-readable medium, on which a computer program is stored which, when being executed by a processor, carries out the method according to any one of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011396279.1A CN112328751A (en) | 2020-12-03 | 2020-12-03 | Method and device for processing text |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011396279.1A CN112328751A (en) | 2020-12-03 | 2020-12-03 | Method and device for processing text |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112328751A true CN112328751A (en) | 2021-02-05 |
Family
ID=74301339
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011396279.1A Pending CN112328751A (en) | 2020-12-03 | 2020-12-03 | Method and device for processing text |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112328751A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113988027A (en) * | 2021-09-23 | 2022-01-28 | 阿里巴巴达摩院(杭州)科技有限公司 | Text generation method, device, equipment and storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111382563A (en) * | 2020-03-20 | 2020-07-07 | 腾讯科技（深圳）有限公司 | Text relevance determining method and device |
| CN111444700A (en) * | 2020-04-02 | 2020-07-24 | 山东山大鸥玛软件股份有限公司 | Text similarity measurement method based on semantic document expression |
| CN111858919A (en) * | 2019-04-30 | 2020-10-30 | 中移(苏州)软件技术有限公司 | Text classification method and device and computer readable storage medium |
-
2020
- 2020-12-03 CN CN202011396279.1A patent/CN112328751A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111858919A (en) * | 2019-04-30 | 2020-10-30 | 中移(苏州)软件技术有限公司 | Text classification method and device and computer readable storage medium |
| CN111382563A (en) * | 2020-03-20 | 2020-07-07 | 腾讯科技（深圳）有限公司 | Text relevance determining method and device |
| CN111444700A (en) * | 2020-04-02 | 2020-07-24 | 山东山大鸥玛软件股份有限公司 | Text similarity measurement method based on semantic document expression |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113988027A (en) * | 2021-09-23 | 2022-01-28 | 阿里巴巴达摩院(杭州)科技有限公司 | Text generation method, device, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107491547B (en) | Search method and device based on artificial intelligence | |
| CN107273503B (en) | Method and device for generating parallel text in same language | |
| CN107066449B (en) | Information pushing method and device | |
| CN109214386B (en) | Method and apparatus for generating image recognition model | |
| CN111428010B (en) | Man-machine intelligent question-answering method and device | |
| CN109543058B (en) | Method, electronic device, and computer-readable medium for detecting image | |
| CN110046254B (en) | Method and apparatus for generating a model | |
| CN107861954B (en) | Information output method and device based on artificial intelligence | |
| CN110019742B (en) | Method and device for processing information | |
| CN109241286B (en) | Method and device for generating text | |
| CN109858045B (en) | Machine translation method and device | |
| CN109740167B (en) | Method and apparatus for generating information | |
| CN109582825B (en) | Method and apparatus for generating information | |
| CN108121699B (en) | Method and apparatus for outputting information | |
| CN111582360A (en) | Method, apparatus, device and medium for labeling data | |
| CN109190123B (en) | Method and apparatus for outputting information | |
| CN116012481B (en) | Image generation processing method and device, electronic equipment and storage medium | |
| CN107766498B (en) | Method and apparatus for generating information | |
| CN112182255A (en) | Method and apparatus for storing media files and for retrieving media files | |
| US11036996B2 (en) | Method and apparatus for determining (raw) video materials for news | |
| CN110245334B (en) | Method and device for outputting information | |
| US11893990B2 (en) | Audio file annotation | |
| KR102434666B1 (en) | Method and computing device for generating text based on speech data using dictionary database | |
| CN112633004A (en) | Text punctuation deletion method and device, electronic equipment and storage medium | |
| CN112307738A (en) | Method and device for processing text |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210205 |
|
| RJ01 | Rejection of invention patent application after publication |