CN111859986B - Semantic matching method, device, equipment and medium based on multi-task twin network - Google Patents
Semantic matching method, device, equipment and medium based on multi-task twin network Download PDFInfo
- Publication number
- CN111859986B CN111859986B CN202010732024.1A CN202010732024A CN111859986B CN 111859986 B CN111859986 B CN 111859986B CN 202010732024 A CN202010732024 A CN 202010732024A CN 111859986 B CN111859986 B CN 111859986B
- Authority
- CN
- China
- Prior art keywords
- word
- text
- vector
- processed
- standard
- 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
Images
Classifications
-
- 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
- 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
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/205—Parsing
-
- 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
Abstract
The application relates to the field of artificial intelligence, in particular to a semantic matching method, device, equipment and medium based on a multi-task twin network. The method comprises the following steps: acquiring a text to be processed and a standard text; inputting the text to be processed into a first feature extractor to obtain a word quantity, a word vector and a word combination vector corresponding to the text to be processed, and inputting the standard text into a second feature extractor to obtain a word vector, a word vector and a word combination vector corresponding to the standard text; the first feature extractor and the second feature extractor are mutually twinned networks; inputting the obtained word vectors, word vectors and word combination vectors into a pre-trained target semantic matching model to obtain target similarity of the text to be processed and the standard text; and outputting standard texts corresponding to the texts to be processed according to the target similarity. In addition, the application relates to blockchain technology, and standard text and target semantic matching models can be stored in the blockchain. By adopting the method, the accuracy can be improved.
Description
Technical Field
The application relates to the technical field of artificial intelligence, in particular to a semantic matching method, device, equipment and medium based on a multi-task twin network.
Background
With the development of artificial intelligence technology, a natural language processing technology appears, wherein semantic matching is involved, namely whether two sentences have the same meaning is judged, a semantic matching task is completed through semantic matching, namely, data analysis processing is carried out on natural language expressed by a user, and the deep meaning actually intended to be expressed by the user is identified, so that the content intended to be inquired by the user can be accurately identified, the needs of the user can be known, the quick response of enterprises can be helped, the problem of the user can be timely solved, and the satisfaction degree of the user can be improved.
Traditional semantic matching adopts a one-hot coding mode or a mode of pre-training word vectors. But one-hot codes have only a few dimensions of 1 in the space of the bag of words size, and the high-dimensional sparse space is difficult to represent semantic information of sentences. The word vector is pre-trained by mapping words into a low-dimensional dense space, the semantic representation mode of the words is more mature, but semantic characterization trained in a language model task is difficult to adapt to the requirement of the semantic matching task, so that the result of the semantic matching task is inaccurate.
Disclosure of Invention
Based on the foregoing, it is necessary to provide a semantic matching method, device, equipment and medium based on a multi-task twin network, which can improve accuracy.
A semantic matching method based on a multitasking twin network, the method comprising:
acquiring a text to be processed and a standard text;
inputting the text to be processed into a first feature extractor to obtain a word quantity, a word vector and a word combination vector corresponding to the text to be processed, and inputting the standard text into a second feature extractor to obtain a word vector, a word vector and a word combination vector corresponding to the standard text; the first feature extractor and the second feature extractor are twinned networks;
inputting word vectors, word vectors and word combination vectors corresponding to the text to be processed and the standard text into a pre-trained target semantic matching model so as to calculate target similarity of the text to be processed and the standard text through the target semantic matching model;
and outputting standard texts corresponding to the texts to be processed according to the target similarity.
In one embodiment, the obtaining the text to be processed and the standard text includes:
Acquiring a text to be processed and an initial standard text;
extracting a preset number of characters of the text to be processed according to a preset rule, and identifying the text according to the extracted characters to determine extraction parameters;
and selecting standard texts from the initial standard texts according to the extraction parameters.
In one embodiment, the extraction parameter is a keyword, and the selecting standard text from the initial standard text according to the extraction parameter includes:
matching the keywords with standard keywords corresponding to a plurality of service types;
acquiring service types with the number of successfully matched keywords being greater than a preset number as the service types corresponding to the text to be processed;
and selecting standard texts from the initial standard texts according to the service types.
In one embodiment, the inputting the text to be processed into the first feature extractor to obtain the word quantity, the word vector and the word combination vector corresponding to the text to be processed, and inputting the standard text into the second feature extractor to obtain the word vector, the word vector and the word combination vector corresponding to the standard text include:
inputting the text to be processed into a first feature extractor, and performing word segmentation processing and word segmentation processing on the text to be processed through the first feature extractor to obtain a subsequence and a word sequence corresponding to the text to be processed;
Generating word quantity, word vector and word combination vector corresponding to the text to be processed according to the subsequence and word sequence corresponding to the text to be processed;
inputting the standard text into a second feature extractor, and performing word segmentation processing and word segmentation processing on the standard text through the second feature extractor to obtain a subsequence and a word sequence corresponding to the standard text;
and generating the word quantity, the word vector and the word combination vector corresponding to the standard text according to the subsequence and the word sequence corresponding to the standard text.
In one embodiment, the generating, according to the subsequence and the word sequence corresponding to the text to be processed, the word quantity, the word vector and the word combination vector corresponding to the text to be processed includes:
inputting the subsequence and the word sequence corresponding to the text to be processed into a preset neural network in parallel to obtain a word vector, a word vector and a word combination vector corresponding to the text to be processed;
the generating the word quantity, the word vector and the word combination vector corresponding to the standard text according to the subsequence and the word sequence corresponding to the standard text comprises the following steps:
and inputting the subsequence and the word sequence corresponding to the standard text into a preset neural network in parallel to obtain a word vector, a word vector and a word combination vector corresponding to the standard text.
In one embodiment, the obtaining the text to be processed and the standard text includes:
receiving a service processing request sent by a terminal;
extracting an initial text in the service processing request;
and preprocessing the initial text according to the sentence type of the initial text to obtain a text to be processed.
In one embodiment, the method further comprises:
acquiring a training text, a standard text and a corresponding relation between the pre-marked training text and the standard text;
inputting the training text into a first feature extractor to obtain a word quantity, a word vector and a word combination vector corresponding to the training text, and inputting the standard text into a second feature extractor to obtain a word vector, a word vector and a word combination vector corresponding to the standard text, wherein the first feature extractor and the second feature extractor are mutually twinned networks;
inputting the word vectors, the word vectors and the word combination vectors corresponding to the training texts and the standard texts into an initial semantic matching model, respectively calculating the similarity corresponding to the word vectors, the word vectors and the word combination vectors, and calculating the respective corresponding loss functions according to the calculated similarity and the pre-labeled corresponding relation;
Obtaining a target loss function according to the loss functions of the word vector, the word vector and the word combination vector, and carrying out parameter adjustment on the initial semantic matching model according to the target loss function through gradient back propagation to obtain a target semantic matching model.
A semantic matching apparatus based on a multitasking twin network, the apparatus comprising:
the text acquisition module is used for acquiring a text to be processed and a standard text;
the vector generation module is used for inputting the text to be processed into the first feature extractor to obtain a word quantity, a word vector and a word combination vector corresponding to the text to be processed, and inputting the standard text into the second feature extractor to obtain a word vector, a word vector and a word combination vector corresponding to the standard text; the first feature extractor and the second feature extractor are twinned networks;
the model processing module is used for inputting word vectors, word vectors and word combination vectors corresponding to the text to be processed and the standard text into a pre-trained target semantic matching model so as to calculate target similarity of the text to be processed and the standard text through the target semantic matching model;
And the output module is used for outputting standard texts corresponding to the texts to be processed according to the target similarity.
A computer device comprising a memory storing a computer program and a processor implementing the steps of the method described above when the processor executes the computer program.
A computer readable storage medium having stored thereon a computer program which, when executed by a processor, performs the steps of the method described above.
According to the semantic matching method, device, equipment and medium based on the multi-task twin network, the word vector and the word combination vector are input into the semantic matching model, so that a large number of errors caused by word segmentation can be avoided in a plurality of modes, meanwhile, the information of the word granularity level is possessed, better tolerance is provided for wrongly written words or synonyms in single words, and accuracy is improved.
Drawings
FIG. 1 is an application environment diagram of a semantic matching method based on a multi-tasking twin network in one embodiment;
FIG. 2 is a flow diagram of a semantic matching method based on a multi-tasking twin network in one embodiment;
FIG. 3 is a schematic diagram of a twin network in one embodiment;
FIG. 4 is a flow diagram of a method of training a target semantic matching model in one embodiment;
FIG. 5 is a block diagram of a semantic matching device based on a multi-tasking twin network in one embodiment;
fig. 6 is an internal structural diagram of a computer device in one embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application will be further described in detail with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the present application.
The semantic matching method based on the multi-task twin network can be applied to an application environment shown in fig. 1. Wherein the terminal 102 communicates with the server 104 via a network. The terminal 102 sends the text to be processed to the server 104, so that the server 104 obtains the text to be processed and the standard text, the server 104 inputs the text to be processed into the first feature extractor to obtain the word quantity, the word vector and the word combination vector corresponding to the text to be processed, and inputs the standard text into the second feature extractor to obtain the word vector, the word vector and the word combination vector corresponding to the standard text. In addition, the first feature extractor and the second feature extractor are mutually twinned networks, so that the same processing of the text to be processed and the standard text is guaranteed. The server 104 inputs word vectors, word vectors and word combination vectors corresponding to the text to be processed and the standard text into a pre-trained target semantic matching model, so that target similarity between the text to be processed and the standard text is calculated through the target semantic matching model, and the standard text matched with the text to be processed is obtained according to the similarity. The terminal 102 may be, but not limited to, various personal computers, notebook computers, smartphones, tablet computers, and portable wearable devices, and the server 104 may be implemented by a stand-alone server or a server cluster composed of a plurality of servers.
In one embodiment, as shown in fig. 2, a semantic matching method based on a multi-task twin network is provided, and the method is applied to the server in fig. 1 for illustration, and includes the following steps:
s202: and acquiring a text to be processed and a standard text.
Specifically, the text to be processed is a text whose meaning needs to be determined, the standard text is a standard text whose meaning has been clarified, and the standard text also corresponds to a corresponding processing manner, for example, in the field of intelligent question-answering, the text to be processed may refer to a question asked by a user, the standard text is a standard expression corresponding to the question asked by the user and stored in a database in advance, and a standard answer corresponding to the standard text is stored in the database, so that the server may obtain the standard answer according to the standard text corresponding to the text to be processed finally and feed back the standard answer to the user.
S204: inputting the text to be processed into a first feature extractor to obtain a word quantity, a word vector and a word combination vector corresponding to the text to be processed, and inputting the standard text into a second feature extractor to obtain a word vector, a word vector and a word combination vector corresponding to the standard text; the first feature extractor and the second feature extractor are twinning networks.
Specifically, the first feature extractor and the second feature extractor are mutually twinned, namely the first feature extractor and the second feature extractor share weights, so that the text to be processed and the standard text are ensured to be processed identically, and the obtained word vector, word vector and word combination vector are comparable. Specifically, the twin network may be shown in fig. 3, where input1 and input2 are the text to be processed and the standard text, vector1 and Vector2 are the resulting vectors, and sim is the similarity.
Specifically, the text to be processed and the standard text may be processed separately to obtain the corresponding word vector, word vector and word combination vector, for example, the text to be processed may be processed in parallel, so that one thread of the server processes the text to be processed to obtain the corresponding word vector, word vector and word combination vector, and the other thread processes the standard text to obtain the corresponding word vector, word vector and word combination vector.
S206: and inputting the word vectors, the word vectors and the word combination vectors corresponding to the text to be processed and the standard text into a pre-trained target semantic matching model so as to calculate the target similarity of the text to be processed and the standard text through the target semantic matching model.
Specifically, the target semantic matching model predicts through a word vector, a word vector and a word combination vector, and improves accuracy through comprehensive processing of different granularities, for example, the target semantic matching model is to calculate similarity of the word vector, similarity of the word vector and similarity of the word combination vector of a text to be processed and a standard text, and then calculate the similarity of the word vector, similarity of the word vector and similarity of the word combination vector according to parameters obtained by training the target semantic matching model to obtain target similarity, for example, target similarity=a×similarity of the word vector+b×similarity of the word vector+c×word combination vector, where a, b and c are parameters obtained by training the target semantic matching model.
S208: and outputting standard texts corresponding to the texts to be processed according to the target similarity.
Specifically, the server may select one standard text with the greatest similarity as the text corresponding to the text to be processed, and after the server outputs the standard text, the server may perform other processing according to the standard text, for example, may obtain an answer to a question corresponding to the standard text, or obtain an applet, shopping, querying, or other links, or a processing flow (skip, or other, etc.) corresponding to the standard text.
According to the semantic matching method based on the multi-task twin network, the word vectors and the word combination vectors are input into the semantic matching model, so that a large number of errors caused by word segmentation can be avoided in a plurality of modes, meanwhile, the information of the word granularity level is possessed, and the method has better tolerance to wrongly written words or synonyms in single words and the like, and improves accuracy.
In one embodiment, obtaining the text to be processed and the standard text includes: acquiring a text to be processed and an initial standard text; extracting a preset number of characters of a text to be processed according to a preset rule, and identifying the text according to the extracted characters to determine extraction parameters; and selecting standard texts from the initial standard texts according to the extraction parameters.
Specifically, in order to improve the processing efficiency, the determining of the standard text may be determined according to the text to be processed, for example, a preset number of characters of the text to be processed, for example, a preset number, or a preset number of characters corresponding to a certain preset position, may be read in advance. The server determines the classification corresponding to the text to be processed according to the extracted preset number of characters, for example, determines extraction parameters, which may be a business region (e.g. determined by language) or a business type (e.g. determined by keywords) corresponding to the text to be processed. In this way, the server selects the standard text from the initial standard text according to the determined extraction parameters, and the input data amount of the subsequent second feature extractor is reduced, so that the processing efficiency is improved.
In practical application, the server may first read a pre-preset number of characters, for example, 50 characters, of the text to be processed, then perform text classification and recognition according to the pre-preset number of characters, for example, the language or keyword adopted by the text, so as to extract the standard text according to the adopted language or keyword, obtain the standard text corresponding to the language or keyword, and then input the obtained standard text into the second feature extractor, so that the amount of the standard text can be reduced, and the processing efficiency is improved.
It should be emphasized that, in order to further guarantee the privacy and security of the initial standard text and the target semantic segmentation model, the initial standard text and the target semantic segmentation model may also be stored in a node of a blockchain.
In one embodiment, the extracting parameters are keywords, and selecting the standard text from the initial standard text according to the extracting parameters includes: matching the keywords with standard keywords corresponding to a plurality of service types; acquiring service types with the number of successfully matched keywords being greater than the preset number as service types corresponding to the text to be processed; and selecting standard texts from the initial standard texts according to the service types.
Specifically, the server may first determine the standard text according to the keywords in the text to be processed, so that the number of standard texts may be reduced, and then match the text to be processed with the standard text determined by the keywords. The method comprises the steps of classifying according to keywords, wherein the extraction of the keywords can be related to the service, namely, a plurality of service types and keywords corresponding to the service types are preset, so that the server extracts the keywords in the text to be processed, and compared with the keywords corresponding to the service types, when the keywords are matched with the plurality of service types as the service types corresponding to the text to be processed, the standard text can be selected according to the determined service, and the processing capacity of the standard text is reduced.
In the above embodiment, the standard text is screened according to the text to be processed, so that the processing amount of the subsequent target semantic matching model can be reduced, and the processing efficiency is improved.
In one embodiment, inputting a text to be processed into a first feature extractor to obtain a word quantity, a word vector and a word combination vector corresponding to the text to be processed, and inputting a standard text into a second feature extractor to obtain a word vector, a word vector and a word combination vector corresponding to the standard text, wherein the method comprises the following steps: inputting the text to be processed into a first feature extractor, and performing word segmentation and word segmentation on the text to be processed through the first feature extractor to obtain a subsequence and a word sequence corresponding to the text to be processed; generating word quantity, word vector and word combination vector corresponding to the text to be processed according to the subsequence and word sequence corresponding to the text to be processed; inputting the standard text into a second feature extractor, and performing word segmentation and word segmentation on the standard text through the second feature extractor to obtain a subsequence and a word sequence corresponding to the standard text; and generating the word quantity, the word vector and the word combination vector corresponding to the standard text according to the subsequence and the word sequence corresponding to the standard text.
Specifically, the word segmentation and word segmentation of the text to be processed and the standard text are performed through the same word segmentation logic and word segmentation logic, and word vectors, word vectors and word combination vectors are extracted through a feature extractor Network sharing weights. And the word segmentation logic are used for carrying out word segmentation and word segmentation processing on the text to be processed and the standard text in advance, so that the obtained word segmentation and word segmentation are comparably, and the word quantity, word vector and word combination vector obtained according to the word segmentation and word segmentation are comparably.
In one embodiment, generating a word quantity, a word vector and a word combination vector corresponding to the text to be processed according to the subsequence and the word sequence corresponding to the text to be processed includes: inputting the subsequence and the word sequence corresponding to the text to be processed into a preset neural network in parallel to obtain a word vector, a word vector and a word combination vector corresponding to the text to be processed; generating word quantity, word vector and word combination vector corresponding to the standard text according to the subsequence and word sequence corresponding to the standard text, including: and inputting the subsequence and the word sequence corresponding to the standard text into a preset neural network in parallel to obtain a word vector, a word vector and a word combination vector corresponding to the standard text.
Specifically, the step of calculating the word vector, the word vector and the word combination vector is performed through a neural network, for example, the word vector and the word combination vector are input into a feature extraction neural network to be calculated. Optionally, the word sequence and the word combination sequence can be input into the feature extraction network in parallel to calculate to obtain a word vector, a word vector and a word combination vector, so that the processing efficiency can be improved.
In one embodiment, obtaining the text to be processed and the standard text includes: receiving a service processing request sent by a terminal; extracting an initial text in a service processing request; and preprocessing the initial text according to the sentence type of the initial text to obtain a text to be processed.
Specifically, the text to be processed may be obtained according to a service processing request sent by the terminal, so that the server extracts an initial text, determines a sentence type of the initial text, and pre-processes the initial text according to the sentence type of the initial text to obtain the text to be processed. For example, when the sentence type is a question sentence, the interfering words in the initial text, such as the question words, are filtered, and then the text to be processed is obtained. When the sentence type is a statement sentence, the sentence can be directly used as a text to be processed or the like.
In the above embodiment, the initial text is preprocessed before processing, which reduces the data size and improves the processing efficiency.
In one embodiment, referring to fig. 4, fig. 4 is a flowchart of a training method of a target semantic matching model in one embodiment, the training method of the target semantic matching model includes:
s402: and acquiring the corresponding relation among the training text, the standard text and the pre-marked training text.
Specifically, the training samples may be extracted from the system server according to the log, the standard text may refer to the above, and the corresponding relationship between the training text and the standard text may be labeled in advance.
S404: the training text is input into a first feature extractor to obtain word quantity, word vector and word combination vector corresponding to the training text, the standard text is input into a second feature extractor to obtain word vector, word vector and word combination vector corresponding to the standard text, and the first feature extractor and the second feature extractor are mutually twinned networks.
Specifically, the processing of training text and standard text in this step may be referred to above, and will not be described here.
S406: inputting the word vectors, the word vectors and the word combination vectors corresponding to the training texts and the standard texts into an initial semantic matching model, respectively calculating the similarity corresponding to the word vectors, the word vectors and the word combination vectors, and calculating the respective corresponding loss functions according to the calculated similarity and the pre-labeled corresponding relation.
S408: obtaining a target loss function according to the loss functions of the word vector, the word vector and the word combination vector, and carrying out parameter adjustment on the initial semantic matching model according to the target loss function through gradient back propagation to obtain a target semantic matching model.
In the model training of the technical scheme, a plurality of expression forms of a sentence in practice are input, namely word level, word level and word combination level, the traditional mode only considers the expression form of the word level, errors in word segmentation can be introduced, and the trained model can be greatly influenced by the word segmentation accuracy. Therefore, granularity of word level and word combination level is introduced, the influence of word segmentation links can be effectively removed by expansion of the word level, meanwhile, the overall result is more robust due to training of the word combination level, and the final target loss is in the following form:
L=a_1×l_char+a_2×l_word+a_3×l_ (word_char), where a1+a2+a3=1, l_ (word_char) is a word combination vector, l_word is a word vector, and l_char is a word vector.
For ease of understanding, a complete embodiment is presented herein, specifically comprising:
firstly, the text to be processed is applied for hypertension.
Secondly, converting the text to be processed into word granularity: w= [ hypertension application ], let w= [ w1, w2, w3, …, wn ].
Then, w= [ w1, w2, w3, …, wn ] is input into the model, and the word list is searched to obtain a vector t= [ v1, v2, v3, …, vn ] corresponding to each w, and then the whole is w= [ t1, t2, t3, …, tn ], wherein each t is a vector with m dimensions. Thus, w= [ t1, t2, t3, …, tn ] is summed and averaged to obtain the corresponding semantic vector s=1/n (t1+t2+t3+ … +tn).
The corresponding semantic vectors are obtained by referring to the above mode for other granularities, such as word granularity/word+word granularity, and the word list in the model is of a preset structure, namely each word corresponds to a unique m-dimensional vector.
Further assume that there is standard text: diabetes mellitus is warranted. Then the two sentences are respectively input into the model to obtain two semantic vectors s1 and s2, and the similarity between the two vectors, namely the semantic similarity, can be calculated.
It should be noted that, the similarity may be calculated separately for different granularities, and according to the difference between the similarity and the label, for example, the similarity is 0.5, but the label is 1, then the difference is (1-0.5) =0.5, which is the back propagation loss when the model is optimized, and the three granularity losses are added and averaged.
It should be understood that, although the steps in the flowcharts of fig. 2-4 are shown in order as indicated by the arrows, these steps are not necessarily performed in order as indicated by the arrows. The steps are not strictly limited to the order of execution unless explicitly recited herein, and the steps may be executed in other orders. Moreover, at least some of the steps in fig. 2-4 may include multiple steps or stages that are not necessarily performed at the same time, but may be performed at different times, nor does the order in which the steps or stages are performed necessarily performed in sequence, but may be performed alternately or alternately with at least a portion of the steps or stages in other steps or other steps.
In one embodiment, as shown in fig. 5, there is provided a semantic matching apparatus based on a multitasking twin network, including: a text acquisition module 100, a vector generation module 200, a model processing module 300, and an output module 400, wherein:
A text acquisition module 100, configured to acquire a text to be processed and a standard text;
the vector generation module 200 is configured to input a text to be processed into the first feature extractor to obtain a word quantity, a word vector and a word combination vector corresponding to the text to be processed, and input a standard text into the second feature extractor to obtain a word vector, a word vector and a word combination vector corresponding to the standard text; the first feature extractor and the second feature extractor are mutually twinned networks;
the model processing module 300 is configured to input a word vector, a word vector and a word combination vector corresponding to the text to be processed and the standard text into a target semantic matching model that is trained in advance, so as to calculate target similarity between the text to be processed and the standard text through the target semantic matching model;
and the output module 400 is used for outputting standard text corresponding to the text to be processed according to the target similarity.
In one embodiment, the text obtaining module 100 may include:
the text acquisition unit is used for acquiring a text to be processed and an initial standard text;
the extraction parameter determining unit is used for extracting a preset number of characters of the text to be processed according to a preset rule, and identifying and determining extraction parameters according to the extracted characters;
And the selecting unit is used for selecting the standard text from the initial standard text according to the extraction parameters.
In one embodiment, the extracting parameter is a keyword, and the selecting unit includes:
the matching subunit is used for matching the keywords with standard keywords corresponding to a plurality of service types;
the business type obtaining subunit is used for obtaining business types with the number of successfully matched keywords being greater than the preset number as the business types corresponding to the text to be processed;
and the selecting subunit is used for selecting standard texts from the initial standard texts according to the service types.
In one embodiment, the vector generation module 200 may include:
the first sequence generating unit is used for inputting the text to be processed into the first feature extractor so as to perform word segmentation processing and word segmentation processing on the text to be processed through the first feature extractor to obtain a subsequence and a word sequence corresponding to the text to be processed;
the first vector generation unit is used for generating word quantity, word vector and word combination vector corresponding to the text to be processed according to the subsequence and word sequence corresponding to the text to be processed;
the second sequence generating unit is used for inputting the standard text into the second feature extractor so as to perform word segmentation processing and word segmentation processing on the standard text through the second feature extractor and obtain a subsequence and a word sequence corresponding to the standard text;
And the second vector generation unit is used for generating the word quantity, the word vector and the word combination vector corresponding to the standard text according to the subsequence and the word sequence corresponding to the standard text.
In one embodiment, the first vector generating unit is further configured to input the subsequence and the word sequence corresponding to the text to be processed into a preset neural network in parallel to obtain a word vector, a word vector and a word combination vector corresponding to the text to be processed;
the second vector generation unit is further configured to input the subsequence and the word sequence corresponding to the standard text into a preset neural network in parallel to obtain a word vector, a word vector and a word combination vector corresponding to the standard text.
In one embodiment, the text obtaining module 100 may include:
the receiving unit is used for receiving the service processing request sent by the terminal;
the extraction unit is used for extracting the initial text in the service processing request;
and the preprocessing unit is used for preprocessing the initial text according to the sentence type of the initial text to obtain a text to be processed.
In one embodiment, the semantic matching device based on the multi-task twin network may further include:
the training data acquisition module is used for acquiring the training text, the standard text and the corresponding relation between the pre-marked training text and the standard text;
The training vector acquisition module is used for inputting training texts into the first feature extractor to obtain word quantities, word vectors and word combination vectors corresponding to the training texts, inputting standard texts into the second feature extractor to obtain word vectors, word vectors and word combination vectors corresponding to the standard texts, wherein the first feature extractor and the second feature extractor are mutually twinned networks;
the loss function generation module is used for inputting word vectors, word vectors and word combination vectors corresponding to the training texts and the standard texts into the initial semantic matching model, respectively calculating the similarity corresponding to the word vectors, the word vectors and the word combination vectors, and calculating the loss functions corresponding to the word vectors, the word vectors and the word combination vectors according to the calculated similarity and the corresponding relation marked in advance;
the training module is used for obtaining a target loss function according to the loss functions of the word vector, the word vector and the word combination vector, and carrying out parameter adjustment on the initial semantic matching model according to the target loss function through gradient back propagation to obtain a target semantic matching model.
For specific limitations on the semantic matching means based on the multi-tasking twin network, reference may be made to the above limitation on the semantic matching method based on the multi-tasking twin network, and will not be repeated here. The above-mentioned respective modules in the semantic matching apparatus based on the multitasking twin network may be implemented in whole or in part by software, hardware, and combinations thereof. The above modules may be embedded in hardware or may be independent of a processor in the computer device, or may be stored in software in a memory in the computer device, so that the processor may call and execute operations corresponding to the above modules.
In one embodiment, a computer device is provided, which may be a server, the internal structure of which may be as shown in fig. 6. The computer device includes a processor, a memory, and a network interface connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database of the computer device is used to store standard text and a target semantic matching model. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program, when executed by a processor, implements a semantic matching method based on a multi-tasking twin network.
It will be appreciated by those skilled in the art that the structure shown in fig. 6 is merely a block diagram of some of the structures associated with the present application and is not limiting of the computer device to which the present application may be applied, and that a particular computer device may include more or fewer components than shown, or may combine certain components, or have a different arrangement of components.
In one embodiment, a computer device is provided comprising a memory and a processor, the memory having stored therein a computer program, the processor when executing the computer program performing the steps of: acquiring a text to be processed and a standard text; inputting the text to be processed into a first feature extractor to obtain a word quantity, a word vector and a word combination vector corresponding to the text to be processed, and inputting the standard text into a second feature extractor to obtain a word vector, a word vector and a word combination vector corresponding to the standard text; the first feature extractor and the second feature extractor are mutually twinned networks; inputting word vectors, word vectors and word combination vectors corresponding to the text to be processed and the standard text into a pre-trained target semantic matching model so as to calculate target similarity of the text to be processed and the standard text through the target semantic matching model; and outputting standard texts corresponding to the texts to be processed according to the target similarity.
In one embodiment, the obtaining the text to be processed and the standard text implemented when the processor executes the computer program includes: acquiring a text to be processed and an initial standard text; extracting a preset number of characters of a text to be processed according to a preset rule, and identifying the text according to the extracted characters to determine extraction parameters; and selecting standard texts from the initial standard texts according to the extraction parameters.
In one embodiment, the extraction parameters implemented when the processor executes the computer program are keywords, and selecting the standard text from the initial standard text according to the extraction parameters includes: matching the keywords with standard keywords corresponding to a plurality of service types; acquiring service types with the number of successfully matched keywords being greater than the preset number as service types corresponding to the text to be processed; and selecting standard texts from the initial standard texts according to the service types.
In one embodiment, inputting the text to be processed into the first feature extractor to obtain the word quantity, the word vector and the word combination vector corresponding to the text to be processed, and inputting the standard text into the second feature extractor to obtain the word vector, the word vector and the word combination vector corresponding to the standard text, wherein the method comprises the following steps: inputting the text to be processed into a first feature extractor, and performing word segmentation and word segmentation on the text to be processed through the first feature extractor to obtain a subsequence and a word sequence corresponding to the text to be processed; generating word quantity, word vector and word combination vector corresponding to the text to be processed according to the subsequence and word sequence corresponding to the text to be processed; inputting the standard text into a second feature extractor, and performing word segmentation and word segmentation on the standard text through the second feature extractor to obtain a subsequence and a word sequence corresponding to the standard text; and generating the word quantity, the word vector and the word combination vector corresponding to the standard text according to the subsequence and the word sequence corresponding to the standard text.
In one embodiment, the generating, by the processor executing the computer program, the word quantity, the word vector, and the word combination vector corresponding to the text to be processed according to the subsequence and the word sequence corresponding to the text to be processed includes: inputting the subsequence and the word sequence corresponding to the text to be processed into a preset neural network in parallel to obtain a word vector, a word vector and a word combination vector corresponding to the text to be processed; generating, by the processor, an amount of words, a word vector, and a word combination vector corresponding to the standard text from the subsequence and the word sequence corresponding to the standard text, when the computer program is executed, including: and inputting the subsequence and the word sequence corresponding to the standard text into a preset neural network in parallel to obtain a word vector, a word vector and a word combination vector corresponding to the standard text.
In one embodiment, the obtaining the text to be processed and the standard text implemented when the processor executes the computer program includes: receiving a service processing request sent by a terminal; extracting an initial text in a service processing request; and preprocessing the initial text according to the sentence type of the initial text to obtain a text to be processed.
In one embodiment, the processor when executing the computer program further performs the steps of: acquiring a corresponding relation among a training text, a standard text and a pre-marked training text; inputting the training text into a first feature extractor to obtain a word quantity, a word vector and a word combination vector corresponding to the training text, and inputting the standard text into a second feature extractor to obtain a word vector, a word vector and a word combination vector corresponding to the standard text, wherein the first feature extractor and the second feature extractor are mutually twinned networks; inputting the word vectors, the word vectors and the word combination vectors corresponding to the training texts and the standard texts into an initial semantic matching model, respectively calculating the similarity corresponding to the word vectors, the word vectors and the word combination vectors, and calculating the respective corresponding loss functions according to the calculated similarity and the pre-labeled corresponding relation; obtaining a target loss function according to the loss functions of the word vector, the word vector and the word combination vector, and carrying out parameter adjustment on the initial semantic matching model according to the target loss function through gradient back propagation to obtain a target semantic matching model.
In one embodiment, a computer readable storage medium is provided having a computer program stored thereon, which when executed by a processor, performs the steps of obtaining text to be processed and standard text; inputting the text to be processed into a first feature extractor to obtain a word quantity, a word vector and a word combination vector corresponding to the text to be processed, and inputting the standard text into a second feature extractor to obtain a word vector, a word vector and a word combination vector corresponding to the standard text; the first feature extractor and the second feature extractor are mutually twinned networks; inputting word vectors, word vectors and word combination vectors corresponding to the text to be processed and the standard text into a pre-trained target semantic matching model so as to calculate target similarity of the text to be processed and the standard text through the target semantic matching model; and outputting standard texts corresponding to the texts to be processed according to the target similarity.
In one embodiment, the obtaining of the text to be processed and the standard text, which is implemented when the computer program is executed by the processor, includes: acquiring a text to be processed and an initial standard text; extracting a preset number of characters of a text to be processed according to a preset rule, and identifying the text according to the extracted characters to determine extraction parameters; and selecting standard texts from the initial standard texts according to the extraction parameters.
In one embodiment, the extraction parameters implemented when the computer program is executed by the processor are keywords, and selecting the standard text from the initial standard text according to the extraction parameters comprises: matching the keywords with standard keywords corresponding to a plurality of service types; acquiring service types with the number of successfully matched keywords being greater than the preset number as service types corresponding to the text to be processed; and selecting standard texts from the initial standard texts according to the service types.
In one embodiment, the method for inputting the text to be processed into the first feature extractor to obtain the word quantity, the word vector and the word combination vector corresponding to the text to be processed, and inputting the standard text into the second feature extractor to obtain the word vector, the word vector and the word combination vector corresponding to the standard text, wherein the method comprises the steps of: inputting the text to be processed into a first feature extractor, and performing word segmentation and word segmentation on the text to be processed through the first feature extractor to obtain a subsequence and a word sequence corresponding to the text to be processed; generating word quantity, word vector and word combination vector corresponding to the text to be processed according to the subsequence and word sequence corresponding to the text to be processed; inputting the standard text into a second feature extractor, and performing word segmentation and word segmentation on the standard text through the second feature extractor to obtain a subsequence and a word sequence corresponding to the standard text; and generating the word quantity, the word vector and the word combination vector corresponding to the standard text according to the subsequence and the word sequence corresponding to the standard text.
In one embodiment, a computer program, when executed by a processor, generates a word quantity, a word vector, and a word combination vector corresponding to a text to be processed from a subsequence and a word sequence corresponding to the text to be processed, comprising: inputting the subsequence and the word sequence corresponding to the text to be processed into a preset neural network in parallel to obtain a word vector, a word vector and a word combination vector corresponding to the text to be processed; the computer program, when executed by a processor, generates word amounts, word vectors and word combination vectors corresponding to standard texts according to subsequences and word sequences corresponding to the standard texts, comprising: and inputting the subsequence and the word sequence corresponding to the standard text into a preset neural network in parallel to obtain a word vector, a word vector and a word combination vector corresponding to the standard text.
In one embodiment, the obtaining of the text to be processed and the standard text, which is implemented when the computer program is executed by the processor, includes: receiving a service processing request sent by a terminal; extracting an initial text in a service processing request; and preprocessing the initial text according to the sentence type of the initial text to obtain a text to be processed.
In one embodiment, the computer program when executed by the processor further performs the steps of: acquiring a corresponding relation among a training text, a standard text and a pre-marked training text; inputting the training text into a first feature extractor to obtain a word quantity, a word vector and a word combination vector corresponding to the training text, and inputting the standard text into a second feature extractor to obtain a word vector, a word vector and a word combination vector corresponding to the standard text, wherein the first feature extractor and the second feature extractor are mutually twinned networks; inputting the word vectors, the word vectors and the word combination vectors corresponding to the training texts and the standard texts into an initial semantic matching model, respectively calculating the similarity corresponding to the word vectors, the word vectors and the word combination vectors, and calculating the respective corresponding loss functions according to the calculated similarity and the pre-labeled corresponding relation; obtaining a target loss function according to the loss functions of the word vector, the word vector and the word combination vector, and carrying out parameter adjustment on the initial semantic matching model according to the target loss function through gradient back propagation to obtain a target semantic matching model.
Those skilled in the art will appreciate that implementing all or part of the above described methods may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed, may comprise the steps of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in embodiments provided herein may include at least one of non-volatile and volatile memory. The nonvolatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical Memory, or the like. Volatile memory can include random access memory (Random Access Memory, RAM) or external cache memory. By way of illustration, and not limitation, RAM can be in the form of a variety of forms, such as static random access memory (Static Random Access Memory, SRAM) or dynamic random access memory (Dynamic Random Access Memory, DRAM), and the like.
The blockchain referred to in the application is a novel application mode of computer technologies such as distributed data storage, point-to-point transmission, consensus mechanism, encryption algorithm and the like. The Blockchain (Blockchain), which is essentially a decentralised database, is a string of data blocks that are generated by cryptographic means in association, each data block containing a batch of information of network transactions for verifying the validity of the information (anti-counterfeiting) and generating the next block. The blockchain may include a blockchain underlying platform, a platform product services layer, an application services layer, and the like.
The technical features of the above embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.
The above examples merely represent a few embodiments of the present application, which are described in more detail and are not to be construed as limiting the scope of the invention. It should be noted that it would be apparent to those skilled in the art that various modifications and improvements could be made without departing from the spirit of the present application, which would be within the scope of the present application. Accordingly, the scope of protection of the present application is to be determined by the claims appended hereto.
Claims (10)
1. A semantic matching method based on a multitasking twin network, the method comprising:
acquiring a text to be processed and a standard text;
inputting the text to be processed into a first feature extractor to obtain a word vector, a word vector and a word combination vector corresponding to the text to be processed, and inputting the standard text into a second feature extractor to obtain a word vector, a word vector and a word combination vector corresponding to the standard text; the first feature extractor and the second feature extractor are twinned networks;
Inputting word vectors, word vectors and word combination vectors corresponding to the text to be processed and the standard text into a pre-trained target semantic matching model so as to calculate target similarity of the text to be processed and the standard text through the target semantic matching model;
outputting standard texts corresponding to the texts to be processed according to the target similarity;
obtaining a standard answer or a link or a processing flow corresponding to the standard text;
the obtaining the text to be processed and the standard text comprises the following steps:
acquiring a text to be processed and an initial standard text;
extracting a preset number of characters of the text to be processed, and carrying out text recognition according to the extracted characters to determine extraction parameters, wherein the extraction parameters comprise service areas determined by languages or service types determined by keywords;
selecting standard texts from the initial standard texts according to the extraction parameters;
the step of inputting the text to be processed into a first feature extractor to obtain a word vector, a word vector and a word combination vector corresponding to the text to be processed, and the step of inputting the standard text into a second feature extractor to obtain a word vector, a word vector and a word combination vector corresponding to the standard text, comprises the following steps:
Inputting the text to be processed into a first feature extractor, and performing word segmentation and word segmentation on the text to be processed through the first feature extractor to obtain a word sequence and a word sequence corresponding to the text to be processed;
generating a word vector, a word vector and a word combination vector corresponding to the text to be processed according to the word sequence and the word sequence corresponding to the text to be processed;
inputting the standard text into a second feature extractor, and performing word segmentation processing and word segmentation processing on the standard text through the second feature extractor to obtain a word sequence and a word sequence corresponding to the standard text;
generating a word vector, a word vector and a word combination vector corresponding to the standard text according to the word sequence and the word sequence corresponding to the standard text;
generating a word vector, a word vector and a word combination vector corresponding to the text to be processed according to the word sequence and the word sequence corresponding to the text to be processed, wherein the word vector, the word vector and the word combination vector comprise:
inputting a word sequence, a word sequence and a word combination sequence corresponding to the text to be processed into a preset neural network in parallel to obtain a word vector, a word vector and a word combination vector corresponding to the text to be processed;
The generating the word vector, the word vector and the word combination vector corresponding to the standard text according to the word sequence and the word sequence corresponding to the standard text comprises the following steps:
and inputting the word sequence, the word sequence and the word combination sequence corresponding to the standard text into a preset neural network in parallel to obtain a word vector, a word vector and a word combination vector corresponding to the standard text.
2. The method of claim 1, wherein said selecting standard text from said initial standard text based on said extraction parameters comprises:
matching the keywords with standard keywords corresponding to a plurality of service types;
acquiring service types with the number of successfully matched keywords being greater than a preset number as the service types corresponding to the text to be processed;
and selecting standard texts from the initial standard texts according to the service types.
3. The method according to claim 1 or 2, wherein the obtaining the text to be processed and the standard text comprises:
receiving a service processing request sent by a terminal;
extracting an initial text in the service processing request;
and preprocessing the initial text according to the sentence type of the initial text to obtain a text to be processed.
4. The method according to claim 1 or 2, characterized in that the method further comprises:
acquiring a training text, a standard text and a corresponding relation between the pre-marked training text and the standard text;
inputting the training text into a first feature extractor to obtain a word vector, a word vector and a word combination vector corresponding to the training text, and inputting the standard text into a second feature extractor to obtain a word vector, a word vector and a word combination vector corresponding to the standard text, wherein the first feature extractor and the second feature extractor are mutually twinned networks;
inputting the word vectors, the word vectors and the word combination vectors corresponding to the training texts and the standard texts into an initial semantic matching model, respectively calculating the similarity corresponding to the word vectors, the word vectors and the word combination vectors, and calculating the respective corresponding loss functions according to the calculated similarity and the pre-labeled corresponding relation;
obtaining a target loss function according to the loss functions of the word vector, the word vector and the word combination vector, and carrying out parameter adjustment on the initial semantic matching model according to the target loss function through gradient back propagation to obtain a target semantic matching model.
5. The method of claim 1 or 2, wherein the initial standard text and the target semantic matching model are stored in nodes of a blockchain.
6. A semantic matching apparatus based on a multitasking twin network, the apparatus comprising:
the text acquisition module is used for acquiring a text to be processed and a standard text;
the vector generation module is used for inputting the text to be processed into the first feature extractor to obtain a word vector, a word vector and a word combination vector corresponding to the text to be processed, and inputting the standard text into the second feature extractor to obtain a word vector, a word vector and a word combination vector corresponding to the standard text; the first feature extractor and the second feature extractor are twinned networks;
the model processing module is used for inputting word vectors, word vectors and word combination vectors corresponding to the text to be processed and the standard text into a pre-trained target semantic matching model so as to calculate target similarity of the text to be processed and the standard text through the target semantic matching model;
the output module is used for outputting standard texts corresponding to the texts to be processed according to the target similarity; obtaining a standard answer or a link or a processing flow corresponding to the standard text;
The text acquisition module comprises:
the text acquisition unit is used for acquiring a text to be processed and an initial standard text;
the extraction parameter determining unit is used for extracting the characters of the pre-preset number of the text to be processed, and carrying out text recognition according to the extracted characters to determine extraction parameters, wherein the extraction parameters comprise service areas determined by languages or service types determined by keywords;
the selecting unit is used for selecting standard texts from the initial standard texts according to the extraction parameters;
the vector generation module includes:
the first sequence generating unit is used for inputting the text to be processed into a first feature extractor, so that word segmentation processing and word segmentation processing are carried out on the text to be processed through the first feature extractor, and a word sequence corresponding to the text to be processed are obtained;
the first vector generation unit is used for generating a word vector, a word vector and a word combination vector corresponding to the text to be processed according to the word sequence and the word sequence corresponding to the text to be processed;
the second sequence generating unit is used for inputting the standard text into a second feature extractor so as to perform word segmentation processing and word segmentation processing on the standard text through the second feature extractor and obtain a word sequence and a word sequence corresponding to the standard text;
The second vector generation unit is used for generating a word vector, a word vector and a word combination vector corresponding to the standard text according to the word sequence and the word sequence corresponding to the standard text;
the first vector generation unit is further used for inputting a word sequence, a word sequence and a word combination sequence corresponding to the text to be processed into a preset neural network in parallel to obtain a word vector, a word vector and a word combination vector corresponding to the text to be processed;
the second vector generation unit is further configured to input a word sequence, a word sequence and a word combination sequence corresponding to the standard text into a preset neural network in parallel to obtain a word vector, a word vector and a word combination vector corresponding to the standard text.
7. The apparatus of claim 6, wherein the selection unit comprises:
the matching subunit is used for matching the keywords with standard keywords corresponding to a plurality of service types;
the service type acquisition subunit is used for acquiring service types with the number of successfully matched keywords being greater than the preset number as the service types corresponding to the text to be processed;
and the selecting subunit is used for selecting standard texts from the initial standard texts according to the service types.
8. The apparatus of claim 6 or 7, wherein the text acquisition module comprises:
the receiving unit is used for receiving the service processing request sent by the terminal;
an extracting unit, configured to extract an initial text in the service processing request;
and the preprocessing unit is used for preprocessing the initial text according to the sentence type of the initial text to obtain a text to be processed.
9. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method of any one of claims 1 to 5 when the computer program is executed.
10. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010732024.1A CN111859986B (en) | 2020-07-27 | 2020-07-27 | Semantic matching method, device, equipment and medium based on multi-task twin network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010732024.1A CN111859986B (en) | 2020-07-27 | 2020-07-27 | Semantic matching method, device, equipment and medium based on multi-task twin network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111859986A CN111859986A (en) | 2020-10-30 |
| CN111859986B true CN111859986B (en) | 2023-06-20 |
Family
ID=72947672
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010732024.1A Active CN111859986B (en) | 2020-07-27 | 2020-07-27 | Semantic matching method, device, equipment and medium based on multi-task twin network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111859986B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112328734B (en) * | 2020-11-05 | 2024-02-13 | 中国平安人寿保险股份有限公司 | Method, device and computer equipment for generating text data |
| CN112711652B (en) * | 2021-01-15 | 2022-10-14 | 清华大学 | Term standardization method and device |
| CN112800248B (en) * | 2021-01-19 | 2022-08-05 | 天河超级计算淮海分中心 | Similar case retrieval method, similar case retrieval device, computer equipment and storage medium |
| CN112800777B (en) * | 2021-04-14 | 2021-07-30 | 北京育学园健康管理中心有限公司 | Semantic determination method |
| CN114614989A (en) * | 2021-11-18 | 2022-06-10 | 亚信科技（中国）有限公司 | Feasibility verification method and device of network service based on digital twin technology |
| CN115497633B (en) * | 2022-10-19 | 2024-01-30 | 联仁健康医疗大数据科技股份有限公司 | Data processing method, device, equipment and storage medium |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110413988A (en) * | 2019-06-17 | 2019-11-05 | 平安科技（深圳）有限公司 | Method, apparatus, server and the storage medium of text information matching measurement |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108595416A (en) * | 2018-03-27 | 2018-09-28 | 义语智能科技（上海）有限公司 | Character string processing method and equipment |
| CN109871446B (en) * | 2019-01-31 | 2023-06-06 | 平安科技（深圳）有限公司 | Refusing method in intention recognition, electronic device and storage medium |
| CN110569500A (en) * | 2019-07-23 | 2019-12-13 | 平安国际智慧城市科技股份有限公司 | Text semantic recognition method and device, computer equipment and storage medium |
| CN111427995B (en) * | 2020-02-26 | 2023-05-26 | 平安科技（深圳）有限公司 | Semantic matching method, device and storage medium based on internal countermeasure mechanism |
-
2020
- 2020-07-27 CN CN202010732024.1A patent/CN111859986B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110413988A (en) * | 2019-06-17 | 2019-11-05 | 平安科技（深圳）有限公司 | Method, apparatus, server and the storage medium of text information matching measurement |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111859986A (en) | 2020-10-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111859960B (en) | Semantic matching method, device, computer equipment and medium based on knowledge distillation | |
| CN111859986B (en) | Semantic matching method, device, equipment and medium based on multi-task twin network | |
| CN110598206B (en) | Text semantic recognition method and device, computer equipment and storage medium | |
| CN108563782B (en) | Commodity information format processing method and device, computer equipment and storage medium | |
| CN110968695A (en) | Intelligent labeling method, device and platform based on active learning of weak supervision technology | |
| CN110334186B (en) | Data query method and device, computer equipment and computer readable storage medium | |
| CN107291840B (en) | User attribute prediction model construction method and device | |
| US11373117B1 (en) | Artificial intelligence service for scalable classification using features of unlabeled data and class descriptors | |
| CN112231224A (en) | Business system testing method, device, equipment and medium based on artificial intelligence | |
| CN113298197B (en) | Data clustering method, device, equipment and readable storage medium | |
| CN110968725B (en) | Image content description information generation method, electronic device and storage medium | |
| CN111506732A (en) | Text multi-level label classification method | |
| CN111783873A (en) | Incremental naive Bayes model-based user portrait method and device | |
| CN112632258A (en) | Text data processing method and device, computer equipment and storage medium | |
| CN115600109A (en) | Sample set optimization method and device, equipment, medium and product thereof | |
| CN107341152B (en) | Parameter input method and device | |
| CN114691525A (en) | Test case selection method and device | |
| JP2019082860A (en) | Generation program, generation method and generation device | |
| CN112732423B (en) | Process migration method, device, equipment and medium | |
| CN114021541A (en) | Presentation generation method, device, equipment and storage medium | |
| CN113627514A (en) | Data processing method and device of knowledge graph, electronic equipment and storage medium | |
| CN111967253A (en) | Entity disambiguation method and device, computer equipment and storage medium | |
| CN116383367B (en) | Data processing method, device, equipment and medium for cold start stage of dialogue system | |
| CN116681054A (en) | Text similarity recognition method, device, computer equipment and storage medium | |
| CN114880578A (en) | User behavior data processing method, device, equipment and storage medium |
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 |