CN112925961A - Intelligent question and answer method and device based on enterprise entity - Google Patents
Intelligent question and answer method and device based on enterprise entity Download PDFInfo
- Publication number
- CN112925961A CN112925961A CN201911245338.2A CN201911245338A CN112925961A CN 112925961 A CN112925961 A CN 112925961A CN 201911245338 A CN201911245338 A CN 201911245338A CN 112925961 A CN112925961 A CN 112925961A
- Authority
- CN
- China
- Prior art keywords
- entity
- enterprise
- question
- target
- attribute
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/903—Querying
- G06F16/9032—Query formulation
- G06F16/90332—Natural language query formulation or dialogue systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Abstract
The embodiment of the invention provides an intelligent question and answer method and device based on an enterprise entity, which are applied to a server, wherein the method comprises the following steps: identifying a target enterprise entity in a question input by a user; acquiring a final enterprise entity corresponding to the target enterprise entity based on the corresponding relation between the enterprise entity and an enterprise knowledge base; determining an attribute set of the final business entity, and taking the attribute set as a candidate answer set of the target business entity; and determining a matching value of the user input question and each attribute in the attribute set based on the MV-LSTM model, and taking the attribute corresponding to the maximum matching value as a final answer of the user input question. By applying the scheme provided by the embodiment of the invention, the accuracy of enterprise knowledge question answering can be improved.
Description
Technical Field
The invention relates to the technical field of knowledge question answering, in particular to an intelligent question answering method and device based on enterprise entities.
Background
The question answering system forms a knowledge base with a certain fixed structure by carrying out deep processing on data, and analyzes the requirements of users by the most advanced natural language processing technology, thereby quickly and accurately providing the users with required information.
The existing question-answering system is basically a knowledge question-answering system in the common general knowledge field, the question-answering method only solves the knowledge question-answering problem in a matching mode of question sentences and candidate answers, and the accuracy of the existing question-answering system on the knowledge question-answering in the enterprise field is low due to the lack of knowledge in the relevant fields of enterprise entities.
Disclosure of Invention
The embodiment of the invention aims to provide an intelligent question-answering method and system based on enterprise entities so as to improve the accuracy of enterprise knowledge question-answering. The specific technical scheme is as follows:
an intelligent question-answering method based on enterprise entities is applied to a server and comprises the following steps:
identifying a target enterprise entity in a question input by a user;
acquiring a final enterprise entity corresponding to the target enterprise entity based on the corresponding relation between the enterprise entity and an enterprise knowledge base;
determining an attribute set of the final business entity, and taking the attribute set as a candidate answer set of the target business entity;
and determining a matching value of the user input question and each attribute in the attribute set based on the MV-LSTM model, and taking the attribute corresponding to the maximum matching value as a final answer of the user input question.
Optionally, the obtaining a final business entity corresponding to the target business entity based on the corresponding relationship between the business entity and the business knowledge base includes:
performing entity disambiguation on the target enterprise entity, and acquiring a first candidate enterprise entity set corresponding to the target enterprise entity based on the corresponding relation between the enterprise entity and an enterprise knowledge base;
determining a popularity of each business entity in the first set of candidate business entities;
determining the similarity between each business entity in the first candidate business entity set and the target business entity, sequencing the similarities from large to small, and taking K business entities with large similarity results as a second candidate business entity set;
and linearly weighting the similarity and the popularity of each enterprise entity in the second candidate enterprise entity set, and determining the enterprise entity with the maximum value after linear weighting as the final enterprise entity corresponding to the target enterprise entity.
Optionally, before performing entity disambiguation processing on the target business entity, the method further includes:
judging whether an enterprise entity completely matched with the target enterprise entity exists in the enterprise knowledge base by adopting a complete matching method, and if so, taking the matched enterprise entity as a final enterprise entity corresponding to the target enterprise entity;
if not, judging whether an enterprise entity corresponding to the target enterprise entity exists in the enterprise knowledge base according to the corresponding relation between the short name or the brand name and the enterprise full name, and if so, taking the corresponding enterprise entity as a final enterprise entity corresponding to the target enterprise entity;
and if not, executing the step of carrying out entity disambiguation processing on the target enterprise entity.
Optionally, the determining, based on the MV-LSTM model, a matching value of the user input question and each attribute in the attribute set, and using an attribute corresponding to a maximum matching value as a final answer of the user input question includes:
carrying out Embedding vectorization processing on the question input by the user and the attribute set respectively;
respectively inputting the vectorized user input question and attribute set into a bidirectional LSTM network structure to obtain hidden layer output of each time t and splicing the bidirectional hidden layer outputs at the time to obtain the target
Calculating similarity values between any two time steps of question and attribute corresponding to hidden layer output by using Cosine function, bilinear function and Tensor function respectively, and forming three similarity matrixes M by the similarity values calculated by the three similarity functionsm×n。
For the three similarity matrixes respectivelyExtracting the similarity matrix M with K-Max Poolingm×nKey matching features of (1);
calculating scores of the extracted key matching features of the three similarity matrixes by using full-link connection and linear transformation, and determining the calculated scores as matching values of the user input question and each attribute in the attribute set;
and taking the attribute corresponding to the maximum matching value as the final answer of the question input by the user.
Optionally, the identifying a target business entity in the user-input question includes:
and identifying the target business entity in the question input by the user based on the BilSTM-CRF model.
Optionally, the identifying a target business entity in a question input by a user based on the BiLSTM-CRF model includes:
vectorizing each single character in the question input by the user and inputting the vectorized single character into a BilSTM-CRF model to obtain a label of each single character;
calculating a label sequence with the maximum probability by adopting a Viterbi algorithm;
and determining a target enterprise entity in the question input by the user according to the tag sequence.
Optionally, the attributes of the business entity include, but are not limited to, one or more of the following attributes:
company type, enterprise address, extent of operation, industry affiliated, registered capital, stockholder information, investments outside and a consensus actor.
The embodiment of the invention also provides an intelligent question-answering device based on the enterprise entity, which is applied to the server and comprises the following components:
the entity identification module is used for identifying a target enterprise entity in a question input by a user;
the entity determining module is used for obtaining a final enterprise entity corresponding to the target enterprise entity based on the corresponding relation between the enterprise entity and the enterprise knowledge base;
a set determining module, configured to determine a set of attributes of the final business entity, and use the set of attributes as a candidate answer set of the target business entity;
and the answer determining module is used for determining the matching value of the user input question and each attribute in the attribute set based on the MV-LSTM model, and taking the attribute corresponding to the maximum matching value as the final answer of the user input question.
Optionally, the entity determining module includes:
the entity set obtaining sub-module is used for carrying out entity disambiguation on the target enterprise entity and obtaining a first candidate enterprise entity set corresponding to the target enterprise entity based on the corresponding relation between the enterprise entity and the enterprise knowledge base;
a popularity determination sub-module for determining a popularity of each business entity in the first set of candidate business entities;
the similarity determining submodule is used for determining the similarity between each enterprise entity in the first candidate enterprise entity set and the target enterprise entity, sequencing the similarities from large to small, and determining K enterprise entities with large similarity results as a second candidate enterprise entity set;
and the final entity determining submodule is used for linearly weighting the similarity and the popularity of each enterprise entity in the second candidate enterprise entity set, and determining the enterprise entity with the maximum value after the linear weighting as the final enterprise entity corresponding to the target enterprise entity.
Optionally, the apparatus further comprises:
a first judging module, configured to judge whether there is an enterprise entity that is completely matched with the target enterprise entity in the enterprise repository by using a complete matching method before entity disambiguation processing is performed on the target enterprise entity, and if yes, trigger a final entity module, where the final entity module is configured to use the matched enterprise entity as a final enterprise entity corresponding to the target enterprise entity; if not, trigger
A second judgment module;
a second judging module is triggered and used for judging whether an enterprise entity corresponding to the target enterprise entity exists in the enterprise knowledge base according to the corresponding relation between the short name or the brand name and the enterprise full name, and if so, the final entity module is triggered;
and if not, triggering the entity set obtaining submodule.
Optionally, the answer determining module includes:
the vector processing submodule is used for carrying out Embedding vectorization processing on the user input question and the attribute set respectively;
a vector input submodule for respectively inputting the vectorized user input question sentence and the attribute set into the bidirectional LSTM network structure to obtain hidden layer output of each time t and splicing the bidirectional hidden layer outputs of the time to obtain
A matrix forming submodule for calculating similarity values between any two time steps corresponding to hidden layer outputs of question and attribute by using Cosine function, bilinear function and Tensor function respectively, and forming three similar matrixes M by the similarity values calculated by the three similarity functionsm×n。
A feature extraction submodule for extracting the similarity matrix M by respectively using K-Max Pooling for the three similarity matricesm×nKey matching features of (1);
the score calculation submodule is used for calculating scores of the extracted key matching features of the three similarity matrixes by using full-link connection and linear transformation, and determining the calculated scores as the matching values of the user input question and each attribute in the attribute set;
and the answer determining submodule is used for taking the attribute corresponding to the maximum matching value as the final answer of the question input by the user.
Optionally, the entity identification module is specifically configured to:
and identifying the target business entity in the question input by the user based on the BilSTM-CRF model.
Optionally, the entity identification module includes:
the label obtaining submodule is used for vectorizing each single character in the question input by the user and inputting the vectorized single character into a BilSTM-CRF model to obtain a label of each single character;
the sequence obtaining submodule is used for calculating a label sequence with the maximum probability by adopting a Viterbi algorithm;
and the entity determining submodule is used for determining the target enterprise entity in the question input by the user according to the label sequence.
Optionally, the attributes of the business entity include, but are not limited to, one or more of the following attributes:
company type, enterprise address, extent of operation, industry affiliated, registered capital, stockholder information, investments outside and a consensus actor.
The embodiment of the invention also provides a server, which comprises a processor, a communication interface, a memory and a communication bus, wherein the processor and the communication interface are used for finishing mutual communication by the memory through the communication bus;
a memory for storing a computer program;
and the processor is used for realizing any one of the above intelligent question-answering method based on the enterprise entity when executing the program stored in the memory.
In yet another aspect of the present invention, there is also provided a storage medium having instructions stored therein, which when run on a computer, cause the computer to execute any one of the above-mentioned intelligent question-answering method based on business entities.
In yet another aspect of the present invention, the present invention further provides a computer program product containing instructions, which when run on a computer, causes the computer to execute any one of the above-mentioned intelligent question-answering method based on business entities.
The enterprise entity-based intelligent question-answering method and device provided by the embodiment of the invention adopt an enterprise knowledge base and deep learning technology-based intelligent question-answering method, the enterprise knowledge base provides a basis for question answer generation, the deep learning technology provides powerful technical support in the stages of question understanding and answer generation, and the method and device can play an important role in solving the question-answering scene surrounding the enterprise entity. The invention adopts an entity identification technology based on a BilSTM-CRF model and an entity disambiguation technology based on a PageRank graph model in the problem understanding process, and the short names and brand names of enterprise entities can be accurately linked to corresponding enterprise entities in an enterprise knowledge base. In the matching process, an MV-LSTM model is adopted, and the model can carry out deep semantic matching on question core questions and enterprise entity attributes in an enterprise knowledge base from multiple semantic dimensions, so that the accuracy of question answer matching is greatly improved.
Of course, it is not necessary for any product or method of practicing the invention to achieve all of the above-described advantages at the same time.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below.
Fig. 1 is a first flowchart of an intelligent question-answering method based on business entities according to an embodiment of the present invention;
FIG. 2 is a second flowchart of a business entity-based intelligent question answering method according to an embodiment of the present invention;
fig. 3 is a third flowchart of an intelligent question-answering method based on enterprise entities according to an embodiment of the present invention;
fig. 4 is a schematic structural diagram of a first structure of an intelligent question answering device based on a business entity according to an embodiment of the present invention;
fig. 5 is a schematic structural diagram of a second structure of an intelligent question answering device based on enterprise entities according to an embodiment of the present invention;
fig. 6 is a schematic structural diagram of a third structure of an intelligent question-answering device based on enterprise entities according to an embodiment of the present invention;
fig. 7 is a schematic structural diagram of a server according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that, for convenience of description, the identified business entity is named as a target business entity in the embodiment of the present invention. That is, the "target" in the target business entities mentioned in the embodiments of the present invention is only used for distinguishing from other business entities in terms of naming and has no limiting meaning.
In addition, the intelligent question answering method based on the enterprise entity provided by the embodiment of the invention is applied to a server, and particularly, the functional software for realizing the intelligent question answering method based on the enterprise entity provided by the embodiment of the invention can be special intelligent question answering software or a plug-in the existing software, which is reasonable.
Example one
Referring to fig. 1, fig. 1 is a first flowchart of a method for intelligent question answering based on business entities according to an embodiment of the present invention, where the method is applied to a server, as shown in fig. 1, and the method includes the following steps:
s110, identifying a target enterprise entity in a question input by a user;
the process is to obtain the target business entity in the question input by the user for linking with the business entity in the business knowledge base. The business entities in the question may be the full name, short name and brand name of the company, such as "Chinese Bank stocks Limited", "Baidu", "hungry" and "company full name, short name and brand name, respectively, which are herein collectively referred to as business entities.
Optionally, the present embodiment identifies a target business entity in a question input by a user based on the BiLSTM-CRF model.
The BilSTM _ CRF model is composed of a bidirectional LSTM (Long Short-Term Memory) model and a CRF (Conditional Random Fields) model. The LSTM model overcomes the defects of the original cyclic neural network, becomes the most popular neural network model at present, and is successfully applied to many fields such as speech recognition, picture description, natural language processing and the like.
In the training stage of the BilSTM _ CRF model, the training data set adopts the labeled corpus of 1998 edition of the people's daily report, and simultaneously adopts a data enhancement strategy to expand the corpus, namely, enterprise entities in an enterprise knowledge base randomly replace some enterprise entities in the people's daily report corpus according to a certain proportion, and experiments show that the strategy can improve the identification accuracy of the enterprise entities. And in the model training stage, a BIES and O labeling system is adopted to label the corpus, namely Begin, Inner, End, Single and Other, the loss function is a cross entropy loss function, and a Momentum (Momentum) optimization algorithm is adopted to learn model parameters.
During data training, firstly, vectorizing each word in each training sentence in an Embedding mode, inputting the vectorized words into a bidirectional LSTM model to obtain hidden layer output of each time t, and splicing the bidirectional hidden layer outputs of all the times to obtain the hidden layer output of each time t
Finally using the full connection layer
R mapped to n x k dimensionsn×kA vector P is obtained in space, where k is the number of labels of all named entities. Inputting the vector P into the CRF layer, recording the final score function
Wherein A is a label from yiTo yi+1Then the Score function is globally normalized using the SoftMax function.
Specifically, identifying a target business entity in a question input by a user based on a BilSTM-CRF model comprises the following steps:
vectorizing each single character in the question input by the user and inputting the vectorized single character into a BilSTM-CRF model to obtain a label of each single character;
calculating a label sequence with the maximum probability by adopting a Viterbi algorithm;
and determining a target enterprise entity in the question input by the user according to the tag sequence.
S120, obtaining a final enterprise entity corresponding to the target enterprise entity based on the corresponding relation between the enterprise entity and the enterprise knowledge base;
in order to further improve the accuracy of the enterprise entities linked to the enterprise knowledge base, a company abbreviation dictionary of common well-known enterprise entities is manually collected to supplement the abbreviation knowledge of the enterprise knowledge base, such as common internet companies, financial technology companies, and the like.
In the question sentence, there may occur a series of problems that the company is called "Baidu", "Teng" and so on for short, the name of the company brand "is hungry", "Mei Tuo" and so on, the knowledge base enterprise entity corresponding to "Baidu" is "Baidu online network technology (Beijing) Limited company", "Teng Feng" is "Shenzhen Teng Xun computer systems Limited company", "hungry" is "Shanghai Lazas information technology Limited company", the knowledge base enterprise entity corresponding to "Mei Tuo" is "Beijing san Kung technology Limited company", and so on. In order to solve the problem that the company name, brand name, and name-past name are assigned to the same enterprise entity, the embodiment disambiguates the entity based on the PageRank graph model, the BM25 algorithm, and the alias dictionary, and links the entity to the same enterprise entity in the enterprise knowledge base, thereby obtaining the final enterprise entity corresponding to the target enterprise entity.
S130, determining an attribute set of the final enterprise entity, and taking the attribute set as a candidate answer set of the target enterprise entity;
after a final enterprise entity corresponding to a target enterprise entity in a question input by a user is obtained, all attribute sets within one degree or two degrees of the corresponding target enterprise entity are retrieved from an enterprise knowledge base, wherein the attributes of the enterprise entity include but are not limited to one or more of the following attributes:
company type, enterprise address, extent of operation, industry affiliated, registered capital, stockholder information, investments outside and a consensus actor.
And S140, determining a matching value of the user input question and each attribute in the attribute set based on the MV-LSTM model, and taking the attribute corresponding to the maximum matching value as a final answer of the user input question.
After the final attribute set of the enterprise entity is retrieved from the enterprise knowledge base, it is determined which attribute is consistent with the question input by the user, and a text matching technology is adopted to calculate the matching value between the question input by the user and the attribute, and then the attribute with the highest matching value is used as the final answer of the question input by the user.
Specifically, an MV-LSTM model is adopted in the matching calculation process, the MV-LSTM model is a Multi-semantic model based on a bidirectional LSTM network, two sentences are processed by adopting the bidirectional LSTM, then matching values are calculated pairwise on output of an LSTM hidden layer so as to confirm attributes of target enterprise entities in questions input by a user, and the process can be considered as a Multi-View (MV) process, and the meaning of each word under different contexts can be considered.
Specifically, the training process of the MV-LSTM model comprises the following steps: training is carried out by adopting training data which is manually sorted and labeled, the training data covers several problems of enterprise addresses, industry classification, operation range, registered capital, high management, legal persons, external investment and the like, each training sample comprises a question and an attribute, and negative samples of the training data are obtained by random sampling. Using Hinge Loss function (Hinge Loss):
wherein
The score is matched for the positive sample,
a negative sample match score.
In the model prediction stage, after word embedding is carried out on a question and candidate attributes, firstly, hidden layer representation of each word is obtained, hidden layer representation of the question and the candidate attributes at any two moments is calculated through a Cosine function, a bilinear function and a Tensor function to obtain three similarity matrixes, then K-Max Pooling (Pooling) and MLP (fully connected neural network) layers are input, finally, a final matching score is obtained through one-time linear conversion, and the candidate attributes with the maximum matching score are taken as final answers of the question.
After the final enterprise entity and the most matched attribute of the question in the enterprise knowledge base are determined, the most matched entity attribute is directly returned to the user to serve as the answer of the question.
Specifically, based on the MV-LSTM model, determining a matching value between a user input question and each attribute in the attribute set, and using an attribute corresponding to a maximum matching value as a final answer of the user input question, including:
carrying out Embedding vectorization processing on the question input by the user and the attribute set respectively;
respectively inputting the vectorized user input question and attribute set into a bidirectional LSTM network structure to obtain hidden layer output of each time t and splicing the bidirectional hidden layer outputs at the time to obtain the target
Calculating similarity values between any two time steps of question and attribute corresponding to hidden layer output by using Cosine function, bilinear function and Tensor function respectively, and forming three similarity matrixes M by the similarity values calculated by the three similarity functionsm×n。
Extracting the similarity matrix M by respectively using K-Max Pooling on the three similarity matricesm×nKey matching features of (1);
calculating scores of the extracted key matching features of the three similarity matrixes by using full-link connection and linear transformation, and determining the calculated scores as matching values of the user input question and each attribute in the attribute set;
and taking the attribute corresponding to the maximum matching value as the final answer of the question input by the user.
The intelligent question-answering method based on the enterprise entity provided by the embodiment adopts an intelligent question-answering method based on an enterprise knowledge base and a deep learning technology, the enterprise knowledge base provides a basis for generating the answer to the question, the deep learning technology provides powerful technical support in the stage of understanding the question and generating the answer, and the method can play an important role in solving the question-answering scene around the enterprise entity. The invention adopts an entity identification technology based on the BilSTM-CRF model in the problem understanding process, and can accurately identify the short names and brand names of the enterprise entities. In the matching process, an MV-LSTM model is adopted, and the model can carry out deep semantic matching on question core questions and enterprise entity attributes in an enterprise knowledge base from multiple semantic dimensions, so that the accuracy of question answer matching is greatly improved.
Example two
Referring to fig. 2, fig. 2 is a second flowchart of an intelligent question-answering method based on an enterprise entity according to an embodiment of the present invention, where the method is applied to a server, and on the basis of the method according to the embodiment, in this embodiment, obtaining a final enterprise entity corresponding to the target enterprise entity based on a corresponding relationship between the enterprise entity and an enterprise knowledge base specifically includes:
s2201, performing entity disambiguation on the target enterprise entity, and acquiring a first candidate enterprise entity set corresponding to the target enterprise entity based on the corresponding relation between the enterprise entity and an enterprise knowledge base;
for example, assuming that the target business entity is hundred degrees, the corresponding first candidate business entity set may be { Baidu cloud, Baidu online networking technology, Baidu translation, Baidu music }.
S2202, determining the popularity of each business entity in the first candidate set of business entities;
specifically, based on the PageRank graph model, the popularity of each enterprise entity in the candidate enterprise entity set is calculated, specifically, the proportion of investment relations among the enterprise entities in the enterprise knowledge base is used as the weight of an investment edge in the graph, the popularity of each enterprise entity is initialized to be a certain equal default value, such as 1 or 0.25, the popularity of each enterprise entity is obtained after the PageRank algorithm is stabilized through multiple iterations, and the calculated popularity can be stored in the enterprise knowledge base in advance.
S2203, determining the similarity between each enterprise entity in the first candidate enterprise entity set and the target enterprise entity, sequencing the similarities from large to small, and taking K enterprise entities with large similarity results as a second candidate enterprise entity set;
specifically, the BM25 algorithm is used to calculate the literal similarity between the target business entity and each business entity in the first candidate business entity set, and after ranking the literal similarities, the K business entities with the highest similarity result are taken as the second candidate business entity set of the target business entity.
S2204, linearly weighting the similarity and the popularity of each business entity in the second candidate business entity set, and determining the business entity with the largest value after the linear weighting as the final business entity corresponding to the target business entity.
And then, taking the pre-calculated values obtained by linear weighting calculation of the popularity and the literal similarity of each entity as final similarity values of the candidate enterprise entities and the target enterprise entity in the second candidate enterprise entity set, and taking the enterprise entity corresponding to the highest numerical value in the final similarity values obtained by calculation of the K enterprise entities as the final enterprise entity of the target enterprise entity.
It should be noted that, in this embodiment, the sequence of calculating the similarity and the popularity is not limited, that is, the process of performing entity disambiguation processing and entity linking through the PageRank graph model and the BM25 algorithm is not limited in sequence, that is, in this embodiment, the step of obtaining the final enterprise entity corresponding to the target enterprise entity based on the corresponding relationship between the enterprise entity and the enterprise knowledge base may further specifically include:
the method comprises the steps of obtaining a candidate enterprise entity set corresponding to a target enterprise entity based on the corresponding relation between the enterprise entity and an enterprise knowledge base, utilizing a BM25 algorithm to conduct literal matching on the target enterprise entity in enterprise entity full names and question sentences in the candidate enterprise entity set, obtaining the similarity between the enterprise entity and the target enterprise entity, selecting K candidate enterprise entities according to the relation from large to small of the similarity, then calculating K popularity degrees of the candidate enterprise entities based on a PageRank graph model, using the proportion of investment relations between the enterprise entities in the enterprise knowledge base as the weight of investment edges in the graph by the PageRank graph model, and obtaining the importance ranking value of each enterprise entity in the enterprise knowledge base as the popularity degree of each enterprise entity after the PageRank graph model converges. And calculating a new numerical value by adopting a linear weighting mode for the similarity calculated by the BM25 algorithm and the popularity based on the PageRank graph model, and finally taking the enterprise entity corresponding to the highest numerical value as the final enterprise entity corresponding to the target enterprise entity in the question input by the user.
The intelligent question-answering method based on the enterprise entities provided by the embodiment adopts an entity disambiguation technology of the PageRank graph model, so that the short names and brand names of the enterprise entities can be accurately linked to the corresponding enterprise entities in the enterprise knowledge base, and the accuracy rate of the enterprise knowledge question-answering is further improved.
EXAMPLE III
Referring to fig. 3, fig. 3 is a third flowchart of an intelligent question-answering method based on an enterprise entity according to an embodiment of the present invention, where the method is applied to a server, and on the basis of the method according to the second embodiment, before performing entity disambiguation processing on the target enterprise entity, the method according to the present embodiment further includes:
S30A, judging whether a complete matching enterprise entity completely matched with the target enterprise entity exists in the enterprise knowledge base by adopting a complete matching method, if so, executing S30C, taking the matched enterprise entity as a final enterprise entity corresponding to the target enterprise entity, and if not, executing S30B;
specifically, the target business entity and the entity in the business knowledge base are linked first, specifically, a complete matching method is adopted. Because the enterprise company full name has uniqueness, if the enterprise entity in the corresponding enterprise knowledge base can be searched, the target enterprise entity can be mapped to the only enterprise entity in the searched enterprise knowledge base; the process is based on matching the enterprise entity full names in the enterprise knowledge base, so the step can also be called entity full name matching. In order to accelerate the retrieval matching speed, enterprise entities in the enterprise knowledge base can be stored in the Redis base and the Elastic Search base in advance, so that the retrieval matching speed can be greatly improved in engineering practice. If the enterprise knowledge base has the enterprise entity completely matched with the target enterprise entity, executing S30C, taking the corresponding enterprise entity as the final enterprise entity corresponding to the target enterprise entity, and if the enterprise knowledge base does not completely match with the full name of the enterprise, executing S30B.
S30B, judging whether an enterprise entity corresponding to the target enterprise entity exists in the enterprise knowledge base according to the corresponding relation between the short name or the brand name and the enterprise full name, if so, executing S30C, and taking the corresponding enterprise entity as a final enterprise entity corresponding to the target enterprise entity; if not, executing S2201, and carrying out entity disambiguation processing on the target enterprise entity.
In order to improve the accuracy of the link between the target enterprise entity and the enterprise knowledge base, the enterprise knowledge base also comprises a matching dictionary of the abbreviation and the brand name, wherein the matching dictionary is mainly collected in a manual mode and comprises two columns of the company abbreviation or the brand name and the company full name, the key value pair of the company abbreviation or the brand name to the entity full name of the enterprise knowledge base, and the key value pair of the company abbreviation or the brand name to the entity full name of the enterprise knowledge base mainly comprises the entity full name mapping from companies such as common internet companies, financial technology companies and the like to the enterprise knowledge base. In this embodiment, the matching dictionary may be stored in the Redis library and the Elastic Search library to accelerate the Search speed. And then judging whether the enterprise knowledge base has an enterprise entity corresponding to the target enterprise entity according to the corresponding relation between the short name or the brand name and the enterprise full name, if so, executing S30C, and if not, executing S2201.
According to the enterprise entity-based intelligent question-answering method provided by the embodiment, the enterprise entity corresponding to the target enterprise entity is searched by adopting a complete matching method, the enterprise entity full name is stored in advance, the searching and matching speed is improved, the enterprise knowledge question-answering speed is further improved, the matching dictionary of the short name and the brand name is stored in the enterprise knowledge base in advance, the accuracy of the link between the target enterprise entity and the enterprise knowledge base is improved, and the accuracy of the enterprise knowledge question-answering is further improved.
Example four
Corresponding to the first method embodiment, an embodiment of the present invention further provides an intelligent question answering device based on an enterprise entity, which is applied to a server, and as shown in fig. 3, the device includes:
an entity identification module 410, configured to identify a target business entity in a question input by a user;
an entity determining module 420, configured to obtain a final enterprise entity corresponding to the target enterprise entity based on a corresponding relationship between the enterprise entity and an enterprise knowledge base;
a set determining module 430, configured to determine a set of attributes of the final business entity, and use the set of attributes as a candidate answer set of the target business entity;
and the answer determining module 440 is configured to determine, based on the MV-LSTM model, a matching value between the user input question and each attribute in the attribute set, and use an attribute corresponding to the maximum matching value as a final answer of the user input question.
Optionally, the entity identifying module 410 is specifically configured to:
and identifying the target business entity in the question input by the user based on the BilSTM-CRF model.
Optionally, the entity identification module 410 includes:
the label obtaining submodule is used for vectorizing each single character in the question input by the user and inputting the vectorized single character into a BilSTM-CRF model to obtain a label of each single character;
the sequence obtaining submodule is used for calculating a label sequence with the maximum probability by adopting a Viterbi algorithm;
and the entity determining submodule is used for determining the target enterprise entity in the question input by the user according to the label sequence.
Optionally, the answer determining module 440 includes:
the vector processing submodule is used for carrying out Embedding vectorization processing on the user input question and the attribute set respectively;
a vector input submodule for respectively inputting the vectorized user input question sentence and the attribute set into the bidirectional LSTM network structure to obtain hidden layer output of each time t and splicing the bidirectional hidden layer outputs of the time to obtain
A matrix forming submodule for calculating similarity values between any two time steps corresponding to hidden layer outputs of question and attribute by using Cosine function, bilinear function and Tensor function respectively, and forming three similar matrixes M by the similarity values calculated by the three similarity functionsm×n。
A feature extraction submodule for extracting the similarity matrix M by respectively using K-Max Pooling for the three similarity matricesm×nKey matching features of (1);
the score calculation submodule is used for calculating scores of the extracted key matching features of the three similarity matrixes by using full-link connection and linear transformation, and determining the calculated scores as the matching values of the user input question and each attribute in the attribute set;
and the answer determining submodule is used for taking the attribute corresponding to the maximum matching value as the final answer of the question input by the user.
Optionally, the attributes of the business entity include, but are not limited to, one or more of the following attributes:
company type, enterprise address, extent of operation, industry affiliated, registered capital, stockholder information, investments outside and a consensus actor.
The intelligent question-answering device based on the enterprise entity provided by the embodiment adopts an intelligent question-answering method based on an enterprise knowledge base and a deep learning technology, the enterprise knowledge base provides a basis for generating the answer to the question, the deep learning technology provides powerful technical support in the stage of understanding the question and generating the answer, and the intelligent question-answering device can play an important role in solving the question-answering scene around the enterprise entity. The invention adopts an entity identification technology based on the BilSTM-CRF model in the problem understanding process, and can accurately identify the short names and brand names of the enterprise entities. In the matching process, an MV-LSTM model is adopted, and the model can carry out deep semantic matching on question core questions and enterprise entity attributes in an enterprise knowledge base from multiple semantic dimensions, so that the accuracy of question answer matching is greatly improved.
EXAMPLE five
Corresponding to the second embodiment of the method, an embodiment of the present invention further provides an intelligent question answering apparatus based on an enterprise entity, which is applied to a server, and as shown in fig. 5, in the apparatus, an entity determining module 420 specifically includes:
an entity set obtaining sub-module 5201, configured to perform entity disambiguation on the target enterprise entity, and obtain a first candidate enterprise entity set corresponding to the target enterprise entity based on a correspondence between the enterprise entity and an enterprise knowledge base;
a popularity determination sub-module 5202 configured to determine a popularity of each business entity in the first set of candidate business entities;
a similarity determining sub-module 5203, configured to determine a similarity between each business entity in the first candidate business entity set and the target business entity, sort the similarities from large to small, and determine K business entities with large similarity results as a second candidate business entity set;
the final entity determining sub-module 5204 is configured to linearly weight the similarity and popularity of each business entity in the second candidate business entity set, and determine that the business entity with the largest value after the linear weighting is the final business entity corresponding to the target business entity.
The intelligent question-answering device based on the enterprise entities provided by the embodiment adopts an entity disambiguation technology of the PageRank graph model, so that the short names and brand names of the enterprise entities can be accurately linked to the corresponding enterprise entities in the enterprise knowledge base, and the accuracy rate of the enterprise knowledge question-answering is further improved.
EXAMPLE six
Corresponding to the third method embodiment, an embodiment of the present invention further provides an intelligent question answering device based on an enterprise entity, which is applied to a server, and as shown in fig. 6, the device further includes:
a first determining module 60A, configured to determine whether there is an enterprise entity that is completely matched with the target enterprise entity in the enterprise repository by using a complete matching method before performing entity disambiguation on the target enterprise entity, if so, trigger a final entity module 60C, and trigger a final entity module to execute 60C, where the matched enterprise entity is used as a final enterprise entity corresponding to the target enterprise entity; if not, triggering a second judgment module 60B;
and a second triggering judgment module 60B, configured to judge whether there is an enterprise entity corresponding to the target enterprise entity in the enterprise knowledge base according to a correspondence between the short name or the brand name and the enterprise full name, if yes, trigger the final entity module to execute 60C, and if not, trigger the entity set obtaining sub-module 5201.
The enterprise entity-based intelligent question-answering device provided by the embodiment retrieves the enterprise entity corresponding to the target enterprise entity by adopting a complete matching method, stores the enterprise entity full name in advance, improves retrieval matching speed, further improves enterprise knowledge question-answering speed, and stores the matching dictionary of the name of.
An embodiment of the present invention further provides an electronic device, as shown in fig. 7, including a processor 701, a communication interface 702, a memory 703 and a communication bus 704, where the processor 701, the communication interface 702, and the memory 703 complete mutual communication through the communication bus 704,
a memory 703 for storing a computer program;
the processor 701 is configured to implement the following steps when executing the program stored in the memory 703:
an intelligent question-answering method based on enterprise entities is applied to a server and comprises the following steps:
identifying a target enterprise entity in a question input by a user;
acquiring a final enterprise entity corresponding to the target enterprise entity based on the corresponding relation between the enterprise entity and an enterprise knowledge base;
determining an attribute set of the final business entity, and taking the attribute set as a candidate answer set of the target business entity;
and determining a matching value of the user input question and each attribute in the attribute set based on the MV-LSTM model, and taking the attribute corresponding to the maximum matching value as a final answer of the user input question.
Optionally, the obtaining a final business entity corresponding to the target business entity based on the corresponding relationship between the business entity and the business knowledge base includes:
performing entity disambiguation on the target enterprise entity, and acquiring a first candidate enterprise entity set corresponding to the target enterprise entity based on the corresponding relation between the enterprise entity and an enterprise knowledge base;
determining a popularity of each business entity in the first set of candidate business entities;
determining the similarity between each business entity in the first candidate business entity set and the target business entity, sequencing the similarities from large to small, and taking K business entities with large similarity results as a second candidate business entity set;
and linearly weighting the similarity and the popularity of each enterprise entity in the second candidate enterprise entity set, and determining the enterprise entity with the maximum value after linear weighting as the final enterprise entity corresponding to the target enterprise entity.
Optionally, before performing entity disambiguation processing on the target business entity, the method further includes:
judging whether an enterprise entity completely matched with the target enterprise entity exists in the enterprise knowledge base by adopting a complete matching method, and if so, taking the matched enterprise entity as a final enterprise entity corresponding to the target enterprise entity;
if not, judging whether an enterprise entity corresponding to the target enterprise entity exists in the enterprise knowledge base according to the corresponding relation between the short name or the brand name and the enterprise full name, and if so, taking the corresponding enterprise entity as a final enterprise entity corresponding to the target enterprise entity;
and if not, executing the step of carrying out entity disambiguation processing on the target enterprise entity.
Optionally, the determining, based on the MV-LSTM model, a matching value of the user input question and each attribute in the attribute set, and using an attribute corresponding to a maximum matching value as a final answer of the user input question includes:
carrying out Embedding vectorization processing on the question input by the user and the attribute set respectively;
respectively inputting the vectorized user input question and attribute set into a bidirectional LSTM network structure to obtain hidden layer output of each time t and splicing the bidirectional hidden layer outputs at the time to obtain the target
Calculating similarity values between any two time steps of question and attribute corresponding to hidden layer output by using Cosine function, bilinear function and Tensor function respectively, and forming three similarity matrixes M by the similarity values calculated by the three similarity functionsm×n。
Extracting the similarity matrix M by respectively using K-Max Pooling on the three similarity matricesm×nKey matching features of (1);
calculating scores of the extracted key matching features of the three similarity matrixes by using full-link connection and linear transformation, and determining the calculated scores as matching values of the user input question and each attribute in the attribute set;
and taking the attribute corresponding to the maximum matching value as the final answer of the question input by the user.
Optionally, the identifying a target business entity in the user-input question includes:
and identifying the target business entity in the question input by the user based on the BilSTM-CRF model.
Optionally, the identifying a target business entity in a question input by a user based on the BiLSTM-CRF model includes:
vectorizing each single character in the question input by the user and inputting the vectorized single character into a BilSTM-CRF model to obtain a label of each single character;
calculating a label sequence with the maximum probability by adopting a Viterbi algorithm;
and determining a target enterprise entity in the question input by the user according to the tag sequence.
Optionally, the attributes of the business entity include, but are not limited to, one or more of the following attributes:
company type, enterprise address, extent of operation, industry affiliated, registered capital, stockholder information, investments outside and a consensus actor.
The communication bus mentioned in the electronic device may be a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, or the like. The communication bus may be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, only one thick line is shown, but this does not mean that there is only one bus or one type of bus.
The communication interface is used for communication between the electronic equipment and other equipment.
The Memory may include a Random Access Memory (RAM) or a Non-Volatile Memory (NVM), such as at least one disk Memory. Optionally, the memory may also be at least one memory device located remotely from the processor.
The Processor may be a general-purpose Processor, including a Central Processing Unit (CPU), a Network Processor (NP), and the like; but also Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field Programmable Gate Arrays (FPGAs) or other Programmable logic devices, discrete Gate or transistor logic devices, discrete hardware components.
In another embodiment of the present invention, a storage medium is further provided, where the storage medium stores instructions that, when executed on a computer, cause the computer to perform any one of the above-mentioned method for intelligent question-answering based on business entities.
In yet another embodiment of the present invention, there is provided a computer program product containing instructions which, when run on a computer, cause the computer to perform the business entity-based intelligent question answering method described in any one of the above embodiments.
In the above embodiments, the implementation may be wholly or partially realized by software, hardware, firmware, or any combination thereof. When implemented in software, may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When loaded and executed on a computer, cause the processes or functions described in accordance with the embodiments of the invention to occur, in whole or in part. The computer may be a general purpose computer, a special purpose computer, a network of computers, or other programmable device. The computer instructions may be stored in a computer readable storage medium or transmitted from one computer readable storage medium to another, for example, from one website site, computer, server, or data center to another website site, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device, such as a server, a data center, etc., that incorporates one or more of the available media. The usable medium may be a magnetic medium (e.g., floppy Disk, hard Disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., Solid State Disk (SSD)), among others.
It is noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
All the embodiments in the present specification are described in a related manner, and the same and similar parts among the embodiments may be referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the system embodiment, since it is substantially similar to the method embodiment, the description is simple, and for the relevant points, reference may be made to the partial description of the method embodiment.
The above description is only for the preferred embodiment of the present invention, and is not intended to limit the scope of the present invention. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention shall fall within the protection scope of the present invention.
Claims (10)
1. An intelligent question-answering method based on enterprise entities is applied to a server and is characterized by comprising the following steps:
identifying a target enterprise entity in a question input by a user;
acquiring a final enterprise entity corresponding to the target enterprise entity based on the corresponding relation between the enterprise entity and an enterprise knowledge base;
determining an attribute set of the final business entity, and taking the attribute set as a candidate answer set of the target business entity;
and determining a matching value of the user input question and each attribute in the attribute set based on the MV-LSTM model, and taking the attribute corresponding to the maximum matching value as a final answer of the user input question.
2. The method of claim 1, wherein obtaining a final business entity corresponding to the target business entity based on the correspondence between the business entities and a business knowledge base comprises:
performing entity disambiguation on the target enterprise entity, and acquiring a first candidate enterprise entity set corresponding to the target enterprise entity based on the corresponding relation between the enterprise entity and an enterprise knowledge base;
determining a popularity of each business entity in the first set of candidate business entities;
determining the similarity between each business entity in the first candidate business entity set and the target business entity, sequencing the similarities from large to small, and taking K business entities with large similarity results as a second candidate business entity set;
and linearly weighting the similarity and the popularity of each enterprise entity in the second candidate enterprise entity set, and determining the enterprise entity with the maximum value after linear weighting as the final enterprise entity corresponding to the target enterprise entity.
3. The method of claim 2, wherein prior to performing entity disambiguation processing on the target business entity, the method further comprises:
judging whether an enterprise entity completely matched with the target enterprise entity exists in the enterprise knowledge base by adopting a complete matching method, and if so, taking the matched enterprise entity as a final enterprise entity corresponding to the target enterprise entity;
if not, judging whether an enterprise entity corresponding to the target enterprise entity exists in the enterprise knowledge base according to the corresponding relation between the short name or the brand name and the enterprise full name, and if so, taking the corresponding enterprise entity as a final enterprise entity corresponding to the target enterprise entity;
and if not, executing the step of carrying out entity disambiguation processing on the target enterprise entity.
4. The method according to any one of claims 1-3, wherein the determining, based on the MV-LSTM model, a matching value of the user-input question with each attribute in the attribute set, and using an attribute corresponding to a maximum matching value as a final answer of the user-input question, comprises:
carrying out Embedding vectorization processing on the question input by the user and the attribute set respectively;
respectively inputting the vectorized user input question and attribute set into a bidirectional LSTM network structure to obtain hidden layer output of each time t and splicing the bidirectional hidden layer outputs at the time to obtain the target
Calculating similarity values between any two time steps of question and attribute corresponding to hidden layer output by using Cosine function, bilinear function and Tensor function respectively, and forming three similarity matrixes M by the similarity values calculated by the three similarity functionsm×n。
Extracting the similarity matrix M by respectively using K-Max Pooling on the three similarity matricesm×nKey matching features of (1);
calculating scores of the extracted key matching features of the three similarity matrixes by using full-link connection and linear transformation, and determining the calculated scores as matching values of the user input question and each attribute in the attribute set;
and taking the attribute corresponding to the maximum matching value as the final answer of the question input by the user.
5. The method of claim 4, wherein identifying the target business entity in the user-entered question comprises:
and identifying the target business entity in the question input by the user based on the BilSTM-CRF model.
6. The method of claim 5, wherein identifying the target business entity in the user-entered question based on the BilSTM-CRF model comprises:
vectorizing each single character in the question input by the user and inputting the vectorized single character into a BilSTM-CRF model to obtain a label of each single character;
calculating a label sequence with the maximum probability by adopting a Viterbi algorithm;
and determining a target enterprise entity in the question input by the user according to the tag sequence.
7. The method of claim 6, wherein the attributes of the business entity include, but are not limited to, one or more of the following:
company type, enterprise address, extent of operation, industry affiliated, registered capital, stockholder information, investments outside and a consensus actor.
8. An intelligent question-answering device based on enterprise entities, which is applied to a server, and is characterized in that the device comprises:
the entity identification module is used for identifying a target enterprise entity in a question input by a user;
the entity determining module is used for obtaining a final enterprise entity corresponding to the target enterprise entity based on the corresponding relation between the enterprise entity and the enterprise knowledge base;
a set determining module, configured to determine a set of attributes of the final business entity, and use the set of attributes as a candidate answer set of the target business entity;
and the answer determining module is used for determining the matching value of the user input question and each attribute in the attribute set based on the MV-LSTM model, and taking the attribute corresponding to the maximum matching value as the final answer of the user input question.
9. A server is characterized by comprising a processor, a communication interface, a memory and a communication bus, wherein the processor and the communication interface are used for realizing the communication between the processor and the memory through the communication bus;
a memory for storing a computer program;
a processor for implementing the method steps of any of claims 1 to 7 when executing a program stored in the memory.
10. A storage medium, characterized in that a computer program is stored in the storage medium, which computer program, when being executed by a processor, carries out the method steps of any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911245338.2A CN112925961A (en) | 2019-12-06 | 2019-12-06 | Intelligent question and answer method and device based on enterprise entity |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911245338.2A CN112925961A (en) | 2019-12-06 | 2019-12-06 | Intelligent question and answer method and device based on enterprise entity |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112925961A true CN112925961A (en) | 2021-06-08 |
Family
ID=76162008
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911245338.2A Pending CN112925961A (en) | 2019-12-06 | 2019-12-06 | Intelligent question and answer method and device based on enterprise entity |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112925961A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115375190A (en) * | 2022-10-18 | 2022-11-22 | 北京北方科诚信息技术股份有限公司 | Enterprise operation management system and method |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105045826A (en) * | 2015-06-29 | 2015-11-11 | 华东师范大学 | Entity linkage algorithm based on graph model |
| CN107797989A (en) * | 2017-10-16 | 2018-03-13 | 平安科技（深圳）有限公司 | Enterprise name recognition methods, electronic equipment and computer-readable recording medium |
| CN109408627A (en) * | 2018-11-15 | 2019-03-01 | 众安信息技术服务有限公司 | A kind of answering method and system merging convolutional neural networks and Recognition with Recurrent Neural Network |
| CN109508458A (en) * | 2018-10-31 | 2019-03-22 | 北京国双科技有限公司 | The recognition methods of legal entity and device |
| CN109635285A (en) * | 2018-11-26 | 2019-04-16 | 平安科技（深圳）有限公司 | Enterprise's full name and abbreviation matching method, apparatus, computer equipment and storage medium |
| CN110134964A (en) * | 2019-05-20 | 2019-08-16 | 中国科学技术大学 | A kind of text matching technique based on stratification convolutional neural networks and attention mechanism |
-
2019
- 2019-12-06 CN CN201911245338.2A patent/CN112925961A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105045826A (en) * | 2015-06-29 | 2015-11-11 | 华东师范大学 | Entity linkage algorithm based on graph model |
| CN107797989A (en) * | 2017-10-16 | 2018-03-13 | 平安科技（深圳）有限公司 | Enterprise name recognition methods, electronic equipment and computer-readable recording medium |
| CN109508458A (en) * | 2018-10-31 | 2019-03-22 | 北京国双科技有限公司 | The recognition methods of legal entity and device |
| CN109408627A (en) * | 2018-11-15 | 2019-03-01 | 众安信息技术服务有限公司 | A kind of answering method and system merging convolutional neural networks and Recognition with Recurrent Neural Network |
| CN109635285A (en) * | 2018-11-26 | 2019-04-16 | 平安科技（深圳）有限公司 | Enterprise's full name and abbreviation matching method, apparatus, computer equipment and storage medium |
| CN110134964A (en) * | 2019-05-20 | 2019-08-16 | 中国科学技术大学 | A kind of text matching technique based on stratification convolutional neural networks and attention mechanism |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115375190A (en) * | 2022-10-18 | 2022-11-22 | 北京北方科诚信息技术股份有限公司 | Enterprise operation management system and method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112528672B (en) | Aspect-level emotion analysis method and device based on graph convolution neural network | |
| CN108875051B (en) | Automatic knowledge graph construction method and system for massive unstructured texts | |
| CN111274394B (en) | Method, device and equipment for extracting entity relationship and storage medium | |
| CN110110062B (en) | Machine intelligent question and answer method and device and electronic equipment | |
| US8452772B1 (en) | Methods, systems, and articles of manufacture for addressing popular topics in a socials sphere | |
| CN110704576B (en) | Text-based entity relationship extraction method and device | |
| CN110674317B (en) | Entity linking method and device based on graph neural network | |
| CN111159485B (en) | Tail entity linking method, device, server and storage medium | |
| CN108536807B (en) | Information processing method and device | |
| CN112100354A (en) | Man-machine conversation method, device, equipment and storage medium | |
| CN111191002A (en) | Neural code searching method and device based on hierarchical embedding | |
| CN112287069B (en) | Information retrieval method and device based on voice semantics and computer equipment | |
| WO2023207096A1 (en) | Entity linking method and apparatus, device, and nonvolatile readable storage medium | |
| CN113177412A (en) | Named entity identification method and system based on bert, electronic equipment and storage medium | |
| CN113326702B (en) | Semantic recognition method, semantic recognition device, electronic equipment and storage medium | |
| CN111611807A (en) | Keyword extraction method and device based on neural network and electronic equipment | |
| CN111414746A (en) | Matching statement determination method, device, equipment and storage medium | |
| CN111241410A (en) | Industry news recommendation method and terminal | |
| CN114661872A (en) | Beginner-oriented API self-adaptive recommendation method and system | |
| CN112581327B (en) | Knowledge graph-based law recommendation method and device and electronic equipment | |
| CN113705207A (en) | Grammar error recognition method and device | |
| CN112949293A (en) | Similar text generation method, similar text generation device and intelligent equipment | |
| CN112925961A (en) | Intelligent question and answer method and device based on enterprise entity | |
| CN111723583A (en) | Statement processing method, device, equipment and storage medium based on intention role | |
| CN116662518A (en) | Question answering method, question answering device, electronic equipment and readable 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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210608 |
|
| RJ01 | Rejection of invention patent application after publication |