CN110019742B - Method and device for processing information - Google Patents
Method and device for processing information Download PDFInfo
- Publication number
- CN110019742B CN110019742B CN201810630153.2A CN201810630153A CN110019742B CN 110019742 B CN110019742 B CN 110019742B CN 201810630153 A CN201810630153 A CN 201810630153A CN 110019742 B CN110019742 B CN 110019742B
- Authority
- CN
- China
- Prior art keywords
- text
- target
- determining
- word
- words
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 57
- 238000012545 processing Methods 0.000 title claims abstract description 42
- 230000006399 behavior Effects 0.000 claims abstract description 98
- 239000013598 vector Substances 0.000 claims description 57
- 230000004044 response Effects 0.000 claims description 41
- 238000013145 classification model Methods 0.000 claims description 14
- 230000011218 segmentation Effects 0.000 claims description 13
- 238000006243 chemical reaction Methods 0.000 claims description 10
- 238000004590 computer program Methods 0.000 claims description 9
- 238000012549 training Methods 0.000 description 24
- 230000003993 interaction Effects 0.000 description 10
- 238000010801 machine learning Methods 0.000 description 10
- 238000010586 diagram Methods 0.000 description 9
- 230000006870 function Effects 0.000 description 9
- 238000013528 artificial neural network Methods 0.000 description 8
- 238000004891 communication Methods 0.000 description 8
- 230000015654 memory Effects 0.000 description 6
- 230000008569 process Effects 0.000 description 5
- 230000003287 optical effect Effects 0.000 description 4
- 238000013527 convolutional neural network Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000000306 recurrent effect Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 239000000835 fiber Substances 0.000 description 2
- 230000000644 propagated effect Effects 0.000 description 2
- 239000004065 semiconductor Substances 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 230000003278 mimic effect Effects 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
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
- G06F16/332—Query formulation
- G06F16/3329—Natural language query formulation or dialogue systems
-
- 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
- G06F16/35—Clustering; Classification
- G06F16/355—Class or cluster creation or modification
-
- 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
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Animal Behavior & Ethology (AREA)
- Artificial Intelligence (AREA)
- Human Computer Interaction (AREA)
- Machine Translation (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The embodiment of the application discloses a method and a device for processing information. One embodiment of the method comprises the following steps: acquiring a text to be processed, and determining a business category corresponding to the text; determining a target word in the text based on the business category; acquiring historical target words, and determining target system behaviors based on the business category, the target words and the historical target words; and executing the target system behavior. This embodiment increases the flexibility of making decisions on the target system behavior.
Description
Technical Field
The embodiment of the application relates to the technical field of computers, in particular to a method and a device for processing information.
Background
With the development of computer technology, man-machine interaction technology is increasingly applied. Human-computer interaction systems may be systems built to mimic human-to-human interaction patterns, and may be generally referred to as dialog systems.
The conventional dialogue system usually makes a written dialogue rule by a person, and matches corresponding system behaviors (such as feeding back a sentence or searching for information) or fed back information based on the dialogue rule after analyzing information sent by a user. Heretofore, it has been necessary to manually think of a plurality of sentence patterns in advance for information matching. Meanwhile, each time a business scenario is added, a rule needs to be rewritten.
Disclosure of Invention
The embodiment of the application provides a method and a device for processing information.
In a first aspect, an embodiment of the present application provides a method for processing information, the method including: acquiring a text to be processed, and determining a business category corresponding to the text; determining target words in the text based on the business category; acquiring historical target words, and determining target system behaviors based on business categories, the target words and the historical target words; and executing the target system behavior.
In some embodiments, obtaining text to be processed includes: receiving voice information of a user; and performing voice text conversion on the voice information to generate a text to be processed.
In some embodiments, determining the traffic class corresponding to the text includes: and inputting the text into a pre-trained business classification model to determine a business category corresponding to the text, wherein the classification model is used for representing the corresponding relation between the text and the business category.
In some embodiments, determining the target word in the text based on the business category includes: extracting feature information from the text in response to determining that the traffic class matches a preset traffic class, wherein the feature information comprises a character vector sequence, a word vector sequence and a context feature information sequence, the character vector sequence is composed of character vectors of characters forming the text, the word vector sequence is composed of word vectors of words obtained after word segmentation of the text, and the context feature information sequence is composed of feature information of contexts of the words obtained after word segmentation of the text; and inputting the characteristic information extracted from the text into a pre-trained target word determining model to obtain target words in the text, wherein the target word determining model is used for determining the target words in the text.
In some embodiments, determining the target word in the text based on the business category includes: determining the character string length of the text in response to determining that the service class does not match the preset service class; and in response to determining that the character string length is smaller than the preset length, matching the text with preset target words in the preset target word set, and determining words in the text, which are matched with any preset target word, as target words of the text.
In some embodiments, the method further comprises: the historical dialogue state information, the target words, the business categories and the system behavior information for indicating the last system behavior are stored in advance, and the collected information is stored as the current dialogue state information so as to update the historical dialogue state information.
In some embodiments, obtaining historical target words, determining target system behavior based on business categories, target words, and historical target words, includes: and taking the business category, the target word and the historical target word as input information, and inputting the input information into a pre-trained decision model to obtain target system behaviors, wherein the decision model is used for representing the corresponding relation between the input information and the system behaviors.
In some embodiments, performing target system behavior includes: searching candidate answers matched with the target words from a preset knowledge graph in response to determining that the target system behavior is feedback voice information; determining the similarity between the converted text and the candidate answer; and outputting the candidate answer as a target answer in response to determining that the similarity is greater than a preset value.
In some embodiments, performing the target system behavior further comprises: responding to the fact that the similarity is not larger than a preset value, taking the candidate answer as a first candidate answer, and inputting the text into a pre-trained end-to-end model to obtain a second candidate answer, wherein the end-to-end model is used for representing the corresponding relation between the text and the answer; inputting the first candidate answer and the second candidate answer into a pre-trained ranking model to obtain a ranking result, wherein the ranking model is used for ranking the candidate answers; and determining a target answer based on the sorting result, and outputting the target answer.
In some embodiments, performing target system behavior includes: and in response to determining that the target system behavior is a query knowledge graph, retrieving knowledge point information associated with the target word from the preset knowledge graph, and outputting the knowledge point information.
In a second aspect, an embodiment of the present application provides an apparatus for processing information, the apparatus including: an acquisition unit configured to acquire a text to be processed, and determine a service class corresponding to the text; a first determining unit configured to determine a target word in a text based on a traffic class; a second determining unit configured to acquire a history target word, determine a target system behavior based on the traffic class, the target word, and the history target word; and an execution unit configured to execute the target system behavior.
In some embodiments, the acquisition unit comprises: a receiving module configured to receive voice information of a user; the conversion module is configured to perform voice text conversion on the voice information and generate a text to be processed.
In some embodiments, the acquisition unit is further configured to: and inputting the text into a pre-trained business classification model to determine a business category corresponding to the text, wherein the classification model is used for representing the corresponding relation between the text and the business category.
In some embodiments, the first determining unit comprises: an extraction module configured to extract feature information from a text in response to determining that a traffic class matches a preset traffic class, wherein the feature information includes a character vector sequence, a word vector sequence, and a context feature information sequence, the character vector sequence being composed of character vectors constituting characters of the text, the word vector sequence being composed of word vectors of words obtained after word segmentation of the text, the context feature information sequence being composed of feature information of contexts of words obtained after word segmentation of the text; and the first input module is configured to input the characteristic information extracted from the text into a pre-trained target word determining model to obtain target words in the text, wherein the target word determining model is used for determining the target words in the text.
In some embodiments, the first determining unit comprises: a first determining module configured to determine a string length of the text in response to determining that the traffic class does not match the preset traffic class; and the matching module is configured to match the text with preset target words in the preset target word set in response to determining that the length of the character string is smaller than the preset length, and determine words in the text, which are matched with any preset target word, as target words of the text.
In some embodiments, the apparatus further comprises: and a storage unit configured to aggregate the previously stored historical dialogue state information, target words, business categories, and system behavior information indicating the last system behavior performed, and store the aggregated information as current dialogue state information to update the historical dialogue state information.
In some embodiments, the second determining unit is further configured to: and taking the business category, the target word and the historical target word as input information, and inputting the input information into a pre-trained decision model to obtain target system behaviors, wherein the decision model is used for representing the corresponding relation between the input information and the system behaviors.
In some embodiments, the execution unit comprises: the retrieval module is configured to respond to the fact that the target system behavior is determined to be feedback voice information, and candidate answers matched with the target words are retrieved from a preset knowledge graph; a second determination module configured to determine a similarity of the converted text to the candidate answer; and the first output module is configured to output the candidate answer as the target answer in response to determining that the similarity is greater than a preset value.
In some embodiments, the execution unit further comprises: the second input module is configured to respond to the fact that the similarity is not larger than a preset value, the candidate answers are used as first candidate answers, texts are input into a pre-trained end-to-end model to obtain second candidate answers, and the end-to-end model is used for representing the corresponding relation between the texts and the answers; the ranking module is configured to input the first candidate answer and the second candidate answer into a pre-trained ranking model to obtain a ranking result, wherein the ranking model is used for ranking the candidate answers; and the second output module is configured to determine a target answer based on the sorting result and output the target answer.
In some embodiments, the execution unit is further configured to: and in response to determining that the target system behavior is a query knowledge graph, retrieving knowledge point information associated with the target word from the preset knowledge graph, and outputting the knowledge point information.
In a third aspect, embodiments of the present application provide one or more processors; a storage device having one or more programs stored thereon, which when executed by one or more processors, cause the one or more processors to implement a method as in any of the embodiments of a method for processing information.
In a fourth aspect, embodiments of the present application provide a computer-readable medium having stored thereon a computer program which, when executed by a processor, implements a method as any of the embodiments of the method for processing information.
According to the method and the device for processing information, provided by the embodiment of the application, the business category corresponding to the text to be processed is determined, then the target word in the text is determined based on the business category, the target word and the acquired historical target word, the target system behavior is determined, and finally the target system behavior is executed, so that the dialogue rule does not need to be manually formulated for matching, the next system behavior can be automatically executed based on the business category and the target word, and the flexibility of decision making of the target system behavior is improved.
Drawings
Other features, objects and advantages of the present application will become more apparent upon reading of the detailed description of non-limiting embodiments, made with reference to the accompanying drawings in which:
FIG. 1 is an exemplary system architecture diagram in which an embodiment of the present application may be applied;
FIG. 2 is a flow chart of one embodiment of a method for processing information in accordance with the present application;
FIG. 3 is a schematic illustration of one application scenario of a method for processing information according to the present application;
FIG. 4 is a flow chart of yet another embodiment of a method for processing information in accordance with the present application;
FIG. 5 is a schematic diagram of an embodiment of an apparatus for processing information in accordance with the present application;
FIG. 6 is a schematic diagram of a computer system suitable for use with a server implementing an embodiment of the application.
Detailed Description
The application is described in further detail below with reference to the drawings and examples. It is to be understood that the specific embodiments described herein are merely illustrative of the application and are not limiting of the application. It should be noted that, for convenience of description, only the portions related to the present application are shown in the drawings.
It should be noted that, without conflict, the embodiments of the present application and features of the embodiments may be combined with each other. The application will be described in detail below with reference to the drawings in connection with embodiments.
Fig. 1 shows an exemplary system architecture 100 to which the method for processing information or the apparatus for processing information of the present application may be applied.
As shown in fig. 1, a system architecture 100 may include terminal devices 101, 102, 103, a network 104, and a server 105. The network 104 is used as a medium to provide communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connections, such as wire, wireless communication links, or fiber optic cables, among others.
The user may interact with the server 105 via the network 104 using the terminal devices 101, 102, 103 to receive or send messages or the like. Various communication client applications, such as a voice interaction type application, a shopping type application, a search type application, an instant messaging tool, a mailbox client, social platform software, etc., may be installed on the terminal devices 101, 102, 103.
The terminal devices 101, 102, 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 and supporting network communications, including but not limited to smartphones, tablets, laptop and desktop computers, and the like. When the terminal devices 101, 102, 103 are software, they can be installed in the above-listed electronic devices. Which may be implemented as multiple software or software modules (e.g., to provide distributed services), or as a single software or software module. The present invention is not particularly limited herein.
The server 105 may be a server providing various services, such as a background server providing support for voice interaction type applications installed on the terminal devices 101, 102, 103. The background server may perform processing such as voice-text conversion on the received voice information, and perform processing such as analysis on the converted text, so as to generate a processing result (for example, determine a target system behavior, and execute the target system behavior).
The server may be hardware or software. When the server is hardware, the server may be implemented as a distributed server cluster formed by a plurality of servers, or may be implemented as a single server. When the server is software, it may be implemented as a plurality of software or software modules (e.g., to provide distributed services), or as a single software or software module. The present invention is not particularly limited herein.
It should be noted that, the method for processing information provided by the embodiment of the present application is generally performed by the server 105, and accordingly, the device for processing information is generally disposed in the server 105.
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 information in accordance with the present application is shown. The method for processing information comprises the following steps:
Step 201, obtaining a text to be processed, and determining a service class corresponding to the text.
In the present embodiment, an execution subject of a method for processing information (e.g., the server 105 shown in fig. 1) may first acquire text to be processed. The text may be transmitted to the execution subject by a terminal (e.g., terminal devices 101, 102, 103 shown in fig. 1) used by the user through a wired connection or a wireless connection. It should be noted that the wireless connection may include, but is not limited to, 3G/4G connection, wiFi connection, bluetooth connection, wiMAX connection, zigbee connection, UWB (ultra wideband) connection, and other now known or later developed wireless connection.
After receiving the text to be processed, the executing body may determine a service class corresponding to the text. The traffic classes described above may include, but are not limited to: product query class, after-sales service class, order query class, offer query class, specific product offer query class, page jump class, other classes. Here, the traffic class may be determined in various ways. As an example, the execution body may store a correspondence table between keywords and business categories in advance. The execution body may match the acquired text with the keywords in the correspondence table, and determine the service class corresponding to the matched keywords as the service class corresponding to the text to be processed. For example, if the matched keyword is "i want to buy," the business category may be a product query category. If the matched keyword is "return," the business category may be an after-market service category. If the matched keyword is "order," the business category may be an order query category. If the matched keyword is "preferential," the business category may be a preferential query category, and so on.
Step 202, determining target words in the text based on the business category.
In this embodiment, the execution body may determine the target word in the text based on the determined service class. Wherein the target word may be one or more of the following: product words, brand words, attribute words (e.g., price, memory, color, etc.). The target word may be another word, and is not limited to the above list. Here, the target word in the above text may be determined in various ways.
In some alternative implementations of the present embodiment, each business category may correspond to a pre-trained target word determination model. Wherein the target word determining model can be used for determining target words in the text. The execution subject may determine and extract the target word in the text using a target word determination model corresponding to the determined business category. The method can be concretely implemented by referring to the following steps:
First, extracting characteristic information from the text. Wherein the feature information may be a word vector sequence composed of word vectors constituting words of the text; the word vector sequence may be a word vector sequence composed of word vectors of words obtained by word segmentation of the text. The feature information may include both the word vector sequence and the word vector sequence. The feature information may include other features, such as a sequence of context feature information constituted by a context of a word obtained by segmenting the text (e.g., a character string constituted by a word preceding the word, and a word following the word) (e.g., a sequence constituted by word vectors of respective words in the context).
Note that the word vector may be a vector representing characteristics of a word. The value of each dimension of the word vector may represent a feature having some semantic and grammatical interpretation. A feature may be, among other things, various information used to characterize the basic elements (e.g., components, radicals, strokes, meanings, etc.) of a word. As an example, the above-mentioned execution body may prestore correspondence tables of 21886 chinese characters and graphic symbols contained in the chinese character inner code extension specification (CHINESE INTERNAL Code Specification, GBK) and word vectors, and each of the word vectors may have the same dimension. For each word in the text, the execution body may find a word vector corresponding to the word from the correspondence table. Here, the word vector of each word and graphic symbol may be trained in advance by performing supervised training of the neural network using a machine learning method, or preset by a technician based on a large amount of data statistics.
Note that the word vector may be a vector representing features of a word. The value of each dimension of the word vector may represent a feature having some semantic and grammatical interpretation. The feature herein may be various information for characterizing a basic element (e.g., meaning, etc.) of a word. The execution body may generate the word vector of each term by using various word vector generation methods, for example, may be generated using an existing word vector generation tool (e.g., word2vec, etc.), or may be generated by using a training neural network. It should be noted that the above-mentioned word vector generation method and similarity calculation method are well known techniques widely studied and applied at present, and will not be described here.
And secondly, inputting the characteristic information extracted from the text into a target word determining model corresponding to the determined business category to obtain the target word in the text. It should be noted that, the target word determining model may be obtained by performing supervised training on an existing model structure (for example, a Long Short-Term Memory (LSTM) or a model structure formed by combining LSTM and CRF (Conditional Random Field algorithm )) by using a machine learning method based on training samples. Here, each training sample may include a text and labels for indicating target words in the text. The machine learning method may be used to train the existing model structure used with text in the training sample as input and labels in the training sample as output, and determine the trained model as the target word determination model. It should be noted that, model training by using the machine learning method is a well-known technique widely studied and applied at present, and will not be described here.
In some optional implementations of this embodiment, the executing entity may first match the determined traffic class with a preset traffic class. By way of example, the preset business categories may include a product query class, an after-market service class, an order query class, a coupon query class, a specific product coupon query class, a page jump class. The execution body may first extract feature information from the text in response to determining that the traffic class matches a preset traffic class. The feature information may include a sequence of word vectors, and a sequence of contextual feature information, among others. The feature information extracted from the text may then be input to a pre-trained target word determination model, which may be used to determine target words in the text, to obtain target words in the text. The target word determination model herein may be used to perform target word extraction and determination of text for various business categories.
In some optional implementations of this embodiment, in response to determining that the traffic class does not match the preset traffic class (i.e., the traffic class is a class other than the preset traffic class), the executing entity may determine a string length (i.e., a number of characters) of the text. In response to determining that the string length is less than a preset length (e.g., 4), the text may be matched with preset target words in a set of preset target words, and words in the text that match any of the preset target words may be determined to be target words of the text. Here, the preset target word set may be summarized based on statistics of history data (e.g., statistics according to the frequency of occurrence).
In some optional implementations of this embodiment, the executing body may aggregate the pre-stored historical dialog state information, the target word, the service class, and the system behavior information for indicating the last system behavior executed, and store the aggregated information as the current dialog state information to update the historical dialog state information. The historical dialogue state information can comprise information such as historical target words, historical system behavior information and the like.
In some optional implementations of this embodiment, the execution body may further process the text before determining the target word. As an example, since a technician is required to label a target word in the process of generating a training sample, spot check can be performed on data labeled by the technician in advance, the discovered errors are summarized as rules, and a program is written by using the rules. At this time, the data correction can be performed on the text by using the program. As yet another example, the execution body may maintain a common mispronounced word list. The common misprinted word vocabulary contains common misprinted words. The execution body can utilize the common mispronounced word list to carry out mispronounced word recognition on the text and correct the mispronounced word into a correct word. As yet another example, because technicians are required to label target words during the generation of training samples, different technicians may label target words at different granularity. For example, for the text "I want to buy a pink dress," some technicians mark "pink" as an attribute word, and others mark "pink" as an attribute word. Therefore, a program can be written in advance, and granularity of data marked by a technician can be unified by using the program. At this time, the execution subject may process the text to be processed using this program.
Step 203, acquiring a historical target word, and determining a target system behavior based on the business category, the target word and the historical target word.
In this embodiment, the execution subject may acquire the history object word. The history target word may be a target word extracted from a history text transmitted by the user. The history text may be text within a predetermined period of time (for example, within 2 minutes), or may be text of a predetermined number of times (for example, approximately 3 times). The executing entity may then determine a target system behavior based on the business category, the target word, and the historical target word. The system behavior may be an operation performed by the human-computer interaction system, for example, feeding back a sentence of a user, calling an interface to perform information search, ending a dialogue, querying a knowledge graph, and the like. Here, the target system behavior may be determined in various ways. As an example, the execution body may store in advance a determination rule that matches different traffic categories and is formulated by a technician. The execution subject may determine the target system behavior based on the corresponding decision rule in combination with the category of the target word and the category of the history target word.
As an example, it may be first determined whether the product word is included in the above-described target word and the above-described history target word. If not, the target system behavior can be determined to be a sentence of the feedback user, and the sentence is used for inquiring the product words required by the user. For example, "please ask you what products you can buy? ". If so, it may be determined whether brand words are included in the target words and the history target words. If the brand words are contained, the system behavior can be determined to call the search interface to search information. If no brand words are included, the target system behavior can be determined to be a sentence of the feedback user, and the target system behavior is used for inquiring the brand needed by the user. For example, "please ask you what brand? ".
It should be noted that other ways of determining the behavior of the target system may also be used. For example, the optimal system behavior may be solved using a dynamic programming approach, with the optimal system behavior determined to be the target system behavior.
Step 204, the target system behavior is performed.
In this embodiment, the execution subject may execute the target system behavior. For example, feedback a sentence of the user, call an interface for information search, end a dialogue, query a knowledge graph, and the like.
With continued reference to fig. 3, fig. 3 is a schematic diagram of an application scenario of the method for processing information according to the present embodiment. In the application scenario of fig. 3, a user performs man-machine interaction with a voice interaction system operated by the server 302 using a voice interaction class application installed in the terminal device 301. The user first enters voice information using the terminal device 301. After the terminal device 301 converts the voice information into text, the text is sent to the server 302 (e.g., "i want to buy a cell phone"). After the text is obtained, the server 302 first determines a business category (e.g., a product query category) corresponding to the text. The target word (e.g., the product word "cell phone") in the text is then determined based on the business category. And then acquiring a historical target word, and determining target system behaviors based on the business category, the target word and the historical target word. Here, it may be determined whether the target word and the historical target word include product words, and if not, the target system behavior is determined to be feedback of a user sentence (e.g., "please ask what brand you need"). Finally, the target system behavior is executed.
According to the method provided by the embodiment of the application, the business category corresponding to the text to be processed is determined, then the target word in the text is determined based on the business category, the target word and the acquired historical target word, then the target system behavior is determined, and finally the target system behavior is executed, so that the matching is carried out without manually making a dialogue rule, the next system behavior can be automatically executed based on the business category and the target word, and the flexibility of decision making of the target system behavior is improved.
With further reference to fig. 4, a flow 400 of yet another embodiment of a method for processing information is shown. The flow 400 of the method for processing information comprises the steps of:
Step 401, receiving voice information of a user.
In this embodiment, an execution subject (e.g., the server 105 shown in fig. 1) of the method for processing information may first receive the voice information of the user. Here, the user may transmit voice information using a terminal (e.g., terminal devices 101, 102, 103 shown in fig. 1).
Step 402, performing voice-text conversion on the voice information to generate a text to be processed.
In this embodiment, the execution body may perform voice-text conversion on the voice information to generate a text to be processed. Here, the voice information may be converted into text using various existing voice-to-text conversion methods and tools.
In step 403, the text is input to a pre-trained business classification model, and a business class corresponding to the text is determined.
In this embodiment, the execution subject may input the text into a pre-trained service classification model, and determine a service class corresponding to the text. The classification model is used for representing the corresponding relation between the text and the business category. As an example, the above-mentioned classification model may be a correspondence table that is pre-formulated by a technician based on a large amount of data statistics for characterizing the correspondence of texts to business categories. As yet another example, the traffic classification model may be obtained by performing supervised training on an existing model with classification function (e.g., convolutional neural network (Convolutional Neural Network, CNN), support vector machine (Support Vector Machine, SVM), etc.) using a machine learning method based on training samples. Here, each training sample may include a text and labels for characterizing the business category to which the text corresponds. The method can be used for training the existing model with the classification function by using a machine learning method, taking texts in training samples as input and labels in the training samples as output, and determining the trained model as a business classification model. It should be noted that, model training by using the machine learning method is a well-known technique widely studied and applied at present, and will not be described here.
In response to determining that the traffic class matches the preset traffic class, feature information is extracted from the text, step 404.
In this embodiment, the executing body may first match the determined traffic class with a preset traffic class. By way of example, the preset business categories may include a product query class, an after-market service class, an order query class, a coupon query class, a specific product coupon query class, a page jump class. The execution body may extract feature information from the text in response to determining that the traffic class matches a preset traffic class. The feature information may include a sequence of word vectors, and a sequence of contextual feature information, among others. Here, the character vector sequence may be constituted by a character vector of a character constituting the text, the word vector sequence may be constituted by a word vector of a word obtained by word segmentation of the text, and the contextual characteristic information sequence may be constituted by characteristic information of a context of the word obtained by word segmentation of the text.
And step 405, inputting the characteristic information extracted from the text into a pre-trained target word determining model to obtain target words in the text.
In this embodiment, the execution subject may input the feature information extracted from the text into a pre-trained target word determining model, to obtain the target word in the text, where the target word determining model may be used to determine the target word in the text.
In step 406, in response to determining that the traffic class does not match the preset traffic class, a string length of the text is determined.
In this embodiment, in response to determining that the service class does not match the preset service class (i.e., the service class is a class other than the preset service class), the execution body may determine a character string length (i.e., the number of characters) of the text.
In step 407, in response to determining that the string length is less than the preset length, matching the text with preset target words in the set of preset target words, and determining the word in the text that matches any preset target word as the target word of the text.
In this embodiment, the execution body may, in response to determining that the string length is less than a preset length (e.g., 4), match the text with a preset target word in a set of preset target words, and determine a word in the text that matches any preset target word as a target word of the text. Here, the preset target word set may be summarized based on statistics of history data (e.g., statistics according to the frequency of occurrence).
In step 408, the business category, the target word and the history target word are used as input information, and the input information is input into a pre-trained decision model to obtain the target system behavior.
In this embodiment, the execution subject may use the service class, the target word, and the history target word as input information, and input the input information to a pre-trained decision model to obtain a target system behavior. The decision model may be used to characterize the correspondence between the input information and the system behavior. As an example, the above-mentioned decision model may be a correspondence table formulated by a technician based on a large amount of data statistics for characterizing the correspondence of input information to system behavior. As yet another example, the decision model may be a result of a supervised training of a pre-established neural network (e.g., a three-layer neural network consisting of an input layer, a hidden layer, and an output layer, which may use a sigmoid function or a softmax function, etc.) using a machine learning method based on training samples. Here, each training sample may include a set of input information (including business category, target word, historical target word) and system behavior annotations. Training samples may be derived from data generated by historical interactions of the user with the system. The execution subject may train the neural network by using a machine learning method, taking input information in a training sample as input, taking a label in the training sample as output, and determining the trained model as a decision model. It should be noted that, model training by using the machine learning method is a well-known technique widely studied and applied at present, and will not be described here.
Step 409, the target system behavior is performed.
In this embodiment, the execution subject may execute the target system behavior. For example, feedback a sentence of the user, call an interface for information search, end a dialogue, query a knowledge graph, and the like.
In some optional implementations of this embodiment, in response to determining that the target system behavior is feedback speech information, the executing entity may first retrieve candidate answers matching the target word from a preset knowledge graph. Then, the similarity between the converted text and the candidate answer is determined by various similarity calculation methods (such as Euclidean distance, cosine similarity, etc.). And outputting the candidate answer as a target answer in response to determining that the similarity is greater than a preset value.
In some optional implementations of this embodiment, in response to determining that the similarity is not greater than the preset value, the executing entity may first use the candidate answer as a first candidate answer, input the text into a pre-trained end-to-end model (sequence to sequence, seq2 seq) and obtain a second candidate answer. The end-to-end model may be used to characterize the correspondence between text and answers. In practice, the end-to-end model may consist of two parts, an encoder (encoder) and a decoder (decoder). The encoder may be a layer of a recurrent neural network (Recurrent neural Network, RNN). The decoder may be a recurrent neural network of the same or similar structure as the encoder. And then, inputting the first candidate answer and the second candidate answer into a pre-trained ranking model to obtain a ranking result. The ranking model may be used to rank candidate answers. In practice, the ranking model may use gradient-lifted trees (Gradient Boosting Decison Tree, GBDT) that are pre-fitted based on historical data. The ranking model may output probabilities of individual system behaviors and rank the probabilities of individual system behaviors from large to small. Finally, a target answer may be determined based on the ranking result, and the target answer may be output. For example, the answer ranked first is determined as the target answer.
In some optional implementations of this embodiment, in response to determining that the target system behavior is a query knowledge graph, the execution subject may retrieve knowledge point information associated with the target word from a preset knowledge graph, and output the knowledge point information.
As can be seen from fig. 4, the flow 400 of the method for processing information in this embodiment highlights the step of determining the target word and the step of determining the target system behavior based on the decision model, compared to the corresponding embodiment of fig. 2. Therefore, the scheme described in the embodiment improves the flexibility of deciding the target system behavior, and can also determine the target words in different modes according to different service types, so that the determined target words have more pertinence.
With further reference to fig. 5, as an implementation of the method shown in the above figures, the present application provides an embodiment of an apparatus for processing information, which corresponds to the method embodiment shown in fig. 2, and which is particularly applicable to various electronic devices.
As shown in fig. 5, an apparatus 500 for processing information according to the present embodiment includes: an obtaining unit 501 configured to obtain a text to be processed, and determine a service class corresponding to the text; a first determining unit 502 configured to determine a target word in the text based on the traffic class; a second determining unit 503 configured to acquire a history target word, determine a target system behavior based on the traffic class, the target word, and the history target word; an execution unit 504 configured to execute the above-described target system behavior.
In some optional implementations of the present embodiment, the acquiring unit 501 may include a receiving module and a converting module (not shown in the figure). Wherein the receiving module may be configured to receive voice information of the user. The conversion module may be configured to perform speech-to-text conversion on the speech information to generate text to be processed.
In some optional implementations of this embodiment, the acquiring unit 501 may be further configured to: and inputting the text into a pre-trained business classification model to determine business categories corresponding to the text, wherein the classification model is used for representing the corresponding relation between the text and the business categories.
In some optional implementations of this embodiment, the first determining unit 502 may include an extracting module and a first input module (not shown in the figure). The extracting module may be configured to extract feature information from the text in response to determining that the traffic class matches a preset traffic class, where the feature information includes a word vector sequence, and a context feature information sequence, the word vector sequence is composed of word vectors constituting words of the text, the word vector sequence is composed of word vectors of words obtained after word segmentation of the text, and the context feature information sequence is composed of feature information of contexts of words obtained after word segmentation of the text. The first input module may be configured to input the feature information extracted from the text into a pre-trained target word determining model, to obtain target words in the text, where the target word determining model is used to determine target words in the text.
In some optional implementations of this embodiment, the first determining unit 502 may include a first determining module and a matching module (not shown in the figure). Wherein the first determining module may be configured to determine a string length of the text in response to determining that the traffic class does not match a preset traffic class. The matching module may be configured to, in response to determining that the string length is less than a preset length, match the text with preset target words in a set of preset target words, and determine a word in the text that matches any preset target word as a target word of the text.
In some alternative implementations of the present embodiment, the apparatus may further include a storage unit (not shown in the figures). The storage unit may be configured to aggregate the previously stored historical dialog state information, the target word, the service class, and system behavior information indicating a last system behavior performed, and store the aggregated information as current dialog state information to update the historical dialog state information.
In some optional implementations of this embodiment, the second determining unit 503 may be further configured to use the service class, the target word, and the historical target word as input information, and input the input information into a pre-trained decision model to obtain a target system behavior, where the decision model is used to characterize a correspondence between the input information and the system behavior.
In some alternative implementations of the present embodiment, the executing unit 504 may include a retrieving module, a second determining module, and a first output module (not shown in the figure). Wherein the retrieving module may be configured to retrieve a candidate answer matching the target word from a preset knowledge graph in response to determining that the target system behavior is feedback voice information. The second determination module may be configured to determine a similarity of the converted text to the candidate answer. The first output module may be configured to output the candidate answer as the target answer in response to determining that the similarity is greater than a preset value.
In some optional implementations of this embodiment, the executing unit 504 may further include a second input module, a sorting module, and a second output module (not shown in the figure). The second input module may be configured to, in response to determining that the similarity is not greater than the preset value, use the candidate answer as a first candidate answer, and input the text into a pre-trained end-to-end model to obtain a second candidate answer, where the end-to-end model is used to characterize a correspondence between the text and the answer. The ranking module may be configured to input the first candidate answer and the second candidate answer to a pre-trained ranking model, where the ranking model is used to rank the candidate answers, to obtain a ranking result. The second output module may be configured to determine a target answer based on the ranking result, and output the target answer.
In some optional implementations of this embodiment, the executing unit 504 may be further configured to, in response to determining that the target system behavior is a query knowledge graph, retrieve knowledge-point information associated with the target word from a preset knowledge graph, and output the knowledge-point information.
The device provided by the above embodiment of the present application determines, through the obtaining unit 501, a business category corresponding to a text to be processed is obtained, then the first determining unit 502 determines a target word in the text based on the business category, then the second determining unit 503 determines a target system behavior based on the business category, the target word and the obtained history target word, and finally the executing unit 504 executes the target system behavior, so that no need of manually making a dialogue rule to perform matching, and the next system behavior can be automatically executed based on the business category and the target word, thereby improving the flexibility of making a decision on the target system behavior.
Referring now to FIG. 6, there is illustrated a schematic diagram of a computer system 600 suitable for use with a server embodying embodiments of the present application. The server illustrated in fig. 6 is merely an example, and should not be construed as limiting the functionality and scope of use of embodiments of the present application.
As shown in fig. 6, the computer system 600 includes a Central Processing Unit (CPU) 601, which can perform various appropriate actions and processes according to a program stored in a Read Only Memory (ROM) 602 or a program loaded from a storage section 608 into a Random Access Memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the system 600 are also stored. The CPU 601, ROM 602, and RAM 603 are connected to each other through a bus 604. An input/output (I/O) interface 605 is also connected to bus 604.
The following components are connected to the I/O interface 605: an input portion 606 including a keyboard, mouse, etc.; an output portion 607 including a Cathode Ray Tube (CRT), a Liquid Crystal Display (LCD), and the like, a speaker, and the like; a storage section 608 including a hard disk and the like; and a communication section 609 including a network interface card such as a LAN card, a modem, or the like. The communication section 609 performs communication processing via a network such as the internet. The drive 610 is also connected to the I/O interface 605 as needed. Removable media 611 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like is installed as needed on drive 610 so that a computer program read therefrom is installed as needed into storage section 608.
In particular, according to embodiments of the present disclosure, the processes described above with reference to 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 shown in the flowcharts. In such an embodiment, the computer program may be downloaded and installed from a network through the communication portion 609, and/or installed from the removable medium 611. The above-described functions defined in the method of the present application are performed when the computer program is executed by a Central Processing Unit (CPU) 601. The computer readable medium according to the present application may be a computer readable signal medium or a computer readable storage medium, or any combination of the two. The computer readable storage medium can be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or a combination of any 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 the context of this document, 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 the present application, however, the computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, with the computer-readable program code embodied therein. Such a propagated data signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination of the foregoing. 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: wireless, wire, fiber optic cable, RF, etc., or any suitable combination of the foregoing.
The flowcharts 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 application. 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 involved in the embodiments of the present application may be implemented in software or in hardware. The described units may also be provided in a processor, for example, described as: a processor includes an acquisition unit, a first determination unit, a second determination unit, and an execution unit. The names of these units do not constitute a limitation on the unit itself in some cases, and the acquisition unit may also be described as "a unit that acquires text to be processed", for example.
As another aspect, the present application also provides a computer-readable medium that may be contained in the apparatus described in the above embodiments; or may be present alone without being fitted into the device. The computer readable medium carries one or more programs which, when executed by the apparatus, cause the apparatus to: acquiring a text to be processed, and determining a business category corresponding to the text; determining a target word in the text based on the business category; acquiring historical target words, and determining target system behaviors based on the business category, the target words and the historical target words; and executing the target system behavior.
The above description is only illustrative of the preferred embodiments of the present application and of the principles of the technology employed. It will be appreciated by persons skilled in the art that the scope of the application referred to in the present application is not limited to the specific combinations of the technical features described above, but also covers other technical features formed by any combination of the technical features described above or their equivalents without departing from the inventive concept described above. Such as the above-mentioned features and the technical features disclosed in the present application (but not limited to) having similar functions are replaced with each other.
Claims (10)
1. A method for processing information, comprising:
acquiring a text to be processed, and determining a business category corresponding to the text;
determining a target word in the text based on the business category;
acquiring historical target words, and determining target system behaviors based on the business category, the target words and the historical target words;
Executing the target system behavior;
Wherein the determining, based on the business category, a target word in the text includes: extracting feature information from the text in response to determining that the service class matches a preset service class, wherein the feature information comprises a word vector sequence, a word vector sequence and a context feature information sequence, the word vector sequence is composed of word vectors of words forming the text, the word vector sequence is composed of word vectors of words obtained after word segmentation of the text, and the context feature information sequence is composed of feature information of contexts of the words obtained after word segmentation of the text; inputting the characteristic information extracted from the text into a pre-trained target word determining model to obtain target words in the text, wherein the target word determining model is used for determining target words in the text, and each business category corresponds to one pre-trained target word determining model;
Determining the character string length of the text in response to determining that the service class does not match a preset service class; in response to determining that the character string length is smaller than a preset length, matching the text with preset target words in a preset target word set, and determining words in the text, which are matched with any preset target word, as target words of the text;
wherein the obtaining the historical target word, determining the target system behavior based on the business category, the target word and the historical target word, comprises:
The business category, the target word and the historical target word are used as input information, the input information is input into a pre-trained decision model to obtain target system behaviors, and the decision model is used for representing the corresponding relation between the input information and the system behaviors;
Wherein the target word includes one or more of the following: product words, brand words, attribute words.
2. The method for processing information according to claim 1, wherein the acquiring text to be processed includes:
Receiving voice information of a user;
and performing voice text conversion on the voice information to generate a text to be processed.
3. The method for processing information of claim 1, wherein the determining a traffic class corresponding to the text comprises:
And inputting the text into a pre-trained business classification model, and determining a business category corresponding to the text, wherein the classification model is used for representing the corresponding relation between the text and the business category.
4. The method for processing information according to claim 1, wherein the method further comprises:
and summarizing the pre-stored historical dialogue state information, the target words, the business categories and the system behavior information for indicating the last system behavior executed, and storing the summarized information as current dialogue state information so as to update the historical dialogue state information.
5. The method for processing information of claim 1, wherein the performing the target system behavior comprises:
in response to determining that the target system behavior is feedback voice information, searching candidate answers matched with the target words from a preset knowledge graph;
Determining the similarity of the converted text and the candidate answer;
and outputting the candidate answer as a target answer in response to determining that the similarity is greater than a preset value.
6. The method for processing information of claim 5, wherein the performing the target system behavior further comprises:
In response to determining that the similarity is not greater than the preset value, taking the candidate answer as a first candidate answer, and inputting the text into a pre-trained end-to-end model to obtain a second candidate answer, wherein the end-to-end model is used for representing the corresponding relation between the text and the answer;
Inputting the first candidate answers and the second candidate answers into a pre-trained ranking model to obtain ranking results, wherein the ranking model is used for ranking the candidate answers;
and determining a target answer based on the sorting result, and outputting the target answer.
7. The method for processing information of claim 1, wherein the performing the target system behavior comprises:
And in response to determining that the target system behavior is a query knowledge graph, retrieving knowledge point information associated with the target word from a preset knowledge graph, and outputting the knowledge point information.
8. An apparatus for processing information, comprising:
An acquisition unit configured to acquire a text to be processed, and determine a service class corresponding to the text;
a first determining unit configured to determine a target word in the text based on the traffic class;
a second determining unit configured to acquire a history target word, determine a target system behavior based on the traffic class, the target word, and the history target word;
an execution unit configured to execute the target system behavior;
Wherein the first determination unit is further configured to: extracting feature information from the text in response to determining that the service class matches a preset service class, wherein the feature information comprises a word vector sequence, a word vector sequence and a context feature information sequence, the word vector sequence is composed of word vectors of words forming the text, the word vector sequence is composed of word vectors of words obtained after word segmentation of the text, and the context feature information sequence is composed of feature information of contexts of the words obtained after word segmentation of the text; inputting the characteristic information extracted from the text into a pre-trained target word determining model to obtain target words in the text, wherein the target word determining model is used for determining target words in the text, and each business category corresponds to one pre-trained target word determining model;
Determining the character string length of the text in response to determining that the service class does not match a preset service class; in response to determining that the character string length is smaller than a preset length, matching the text with preset target words in a preset target word set, and determining words in the text, which are matched with any preset target word, as target words of the text;
wherein the second determining unit is specifically configured to: the business category, the target word and the historical target word are used as input information, the input information is input into a pre-trained decision model to obtain target system behaviors, and the decision model is used for representing the corresponding relation between the input information and the system behaviors;
Wherein the target word includes one or more of the following: product words, brand words, attribute words.
9. A server, comprising:
one or more processors;
a storage device having one or more programs stored thereon,
When executed by the one or more processors, causes the one or more processors to implement the method of any of claims 1-7.
10. A computer readable medium having stored thereon a computer program, wherein the program when executed by a processor implements the method of any of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810630153.2A CN110019742B (en) | 2018-06-19 | 2018-06-19 | Method and device for processing information |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810630153.2A CN110019742B (en) | 2018-06-19 | 2018-06-19 | Method and device for processing information |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110019742A CN110019742A (en) | 2019-07-16 |
| CN110019742B true CN110019742B (en) | 2024-05-17 |
Family
ID=67188315
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810630153.2A Active CN110019742B (en) | 2018-06-19 | 2018-06-19 | Method and device for processing information |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110019742B (en) |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111783810B (en) * | 2019-09-24 | 2023-12-08 | 北京沃东天骏信息技术有限公司 | Method and device for determining attribute information of user |
| CN110727775B (en) * | 2019-10-11 | 2021-03-30 | 京东数字科技控股有限公司 | Method and apparatus for processing information |
| CN110909725B (en) * | 2019-10-18 | 2023-09-19 | 平安科技（深圳）有限公司 | Method, device, equipment and storage medium for recognizing text |
| CN110874438B (en) * | 2019-11-19 | 2023-04-07 | 中国银行股份有限公司 | Function navigation method and device |
| CN112908319B (en) * | 2019-12-04 | 2022-10-25 | 海信视像科技股份有限公司 | Method and equipment for processing information interaction |
| CN111199430A (en) * | 2020-01-08 | 2020-05-26 | 中信银行股份有限公司 | Service solution generation method and device and electronic equipment |
| CN113033170B (en) * | 2021-04-23 | 2023-08-04 | 中国平安人寿保险股份有限公司 | Form standardization processing method, device, equipment and storage medium |
| CN114861680B (en) * | 2022-05-27 | 2023-07-25 | 马上消费金融股份有限公司 | Dialogue processing method and device |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105893465A (en) * | 2016-03-28 | 2016-08-24 | 北京京东尚科信息技术有限公司 | Automatic question answering method and device |
| CN107908803A (en) * | 2017-12-26 | 2018-04-13 | 上海智臻智能网络科技股份有限公司 | The response method and device, storage medium, terminal of question and answer interaction |
| CN107958091A (en) * | 2017-12-28 | 2018-04-24 | 北京贝塔智投科技有限公司 | A kind of NLP artificial intelligence approaches and interactive system based on financial vertical knowledge mapping |
| CN108022586A (en) * | 2017-11-30 | 2018-05-11 | 百度在线网络技术（北京）有限公司 | Method and apparatus for controlling the page |
| CN108121824A (en) * | 2018-01-12 | 2018-06-05 | 北京融快线科技有限公司 | A kind of chat robots and system towards financial service |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9754037B2 (en) * | 2014-08-27 | 2017-09-05 | Facebook, Inc. | Blending by query classification on online social networks |
-
2018
- 2018-06-19 CN CN201810630153.2A patent/CN110019742B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105893465A (en) * | 2016-03-28 | 2016-08-24 | 北京京东尚科信息技术有限公司 | Automatic question answering method and device |
| CN108022586A (en) * | 2017-11-30 | 2018-05-11 | 百度在线网络技术（北京）有限公司 | Method and apparatus for controlling the page |
| CN107908803A (en) * | 2017-12-26 | 2018-04-13 | 上海智臻智能网络科技股份有限公司 | The response method and device, storage medium, terminal of question and answer interaction |
| CN107958091A (en) * | 2017-12-28 | 2018-04-24 | 北京贝塔智投科技有限公司 | A kind of NLP artificial intelligence approaches and interactive system based on financial vertical knowledge mapping |
| CN108121824A (en) * | 2018-01-12 | 2018-06-05 | 北京融快线科技有限公司 | A kind of chat robots and system towards financial service |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110019742A (en) | 2019-07-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110019742B (en) | Method and device for processing information | |
| CN111428010B (en) | Man-machine intelligent question-answering method and device | |
| CN108388674B (en) | Method and device for pushing information | |
| KR102288249B1 (en) | Information processing method, terminal, and computer storage medium | |
| CN110046254B (en) | Method and apparatus for generating a model | |
| CN114840671A (en) | Dialogue generation method, model training method, device, equipment and medium | |
| CN112530408A (en) | Method, apparatus, electronic device, and medium for recognizing speech | |
| CN109582825B (en) | Method and apparatus for generating information | |
| CN108268450B (en) | Method and apparatus for generating information | |
| CN111639162A (en) | Information interaction method and device, electronic equipment and storage medium | |
| CN112926308B (en) | Method, device, equipment, storage medium and program product for matching text | |
| CN111930792A (en) | Data resource labeling method and device, storage medium and electronic equipment | |
| CN109190123B (en) | Method and apparatus for outputting information | |
| CN112163081A (en) | Label determination method, device, medium and electronic equipment | |
| CN113239204A (en) | Text classification method and device, electronic equipment and computer-readable storage medium | |
| CN114218945A (en) | Entity identification method, device, server and storage medium | |
| CN116303537A (en) | Data query method and device, electronic equipment and storage medium | |
| CN113051380A (en) | Information generation method and device, electronic equipment and storage medium | |
| CN110738056B (en) | Method and device for generating information | |
| CN111538817A (en) | Man-machine interaction method and device | |
| CN113642324A (en) | Text abstract generation method and device, electronic equipment and storage medium | |
| CN112818096A (en) | Dialog generating method and device | |
| CN114676227B (en) | Sample generation method, model training method and retrieval method | |
| CN112579733A (en) | Rule matching method, rule matching device, storage medium and electronic equipment | |
| CN107656627B (en) | Information input method and device |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |