CN115017291B - Hotspot problem analysis method and device, computer equipment and storage medium - Google Patents
Hotspot problem analysis method and device, computer equipment and storage medium Download PDFInfo
- Publication number
- CN115017291B CN115017291B CN202210931690.7A CN202210931690A CN115017291B CN 115017291 B CN115017291 B CN 115017291B CN 202210931690 A CN202210931690 A CN 202210931690A CN 115017291 B CN115017291 B CN 115017291B
- Authority
- CN
- China
- Prior art keywords
- sentence
- sentences
- analyzed
- key
- determining
- 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
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/332—Query formulation
- G06F16/3329—Natural language query formulation or dialogue systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/3331—Query processing
- G06F16/334—Query execution
- G06F16/3344—Query execution using natural language analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/205—Parsing
- G06F40/211—Syntactic parsing, e.g. based on context-free grammar [CFG] or unification grammars
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/284—Lexical analysis, e.g. tokenisation or collocates
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- General Health & Medical Sciences (AREA)
- Human Computer Interaction (AREA)
- Machine Translation (AREA)
Abstract
The application relates to a hotspot problem analysis method, a hotspot problem analysis device, computer equipment, a storage medium and a computer program product. The method comprises the following steps: the method comprises the steps of obtaining a plurality of sentences to be analyzed from historical dialogue sentences, determining the sentences to be analyzed with the highest matching degree with a key sentence library from the plurality of sentences to be analyzed and the key sentence library, determining the key sentences corresponding to the sentences to be analyzed with the highest matching degree in the key sentence library as candidate sentences, and further determining hot sentences from the candidate sentences by using an attention mechanism algorithm so as to determine hot problems according to the hot sentences. By the method, the problem of hot spots of manual statistics is avoided, the time consumed by statistics is reduced, and the analysis efficiency is improved.
Description
Technical Field
The present application relates to the field of computer technologies, and in particular, to a method and an apparatus for analyzing a hotspot problem, a computer device, a storage medium, and a computer program product.
Background
With the development of artificial intelligence technology, more and more industries introduce customer service robots to automatically answer questions of users.
In the conventional technology, hot problems need to be counted manually based on historical question data, and then the hot problems and corresponding response contents need to be set for the customer service robot, so that when the customer service robot identifies the hot problems provided by the user, the corresponding response contents are provided for the user.
However, the conventional way of manually counting the hot spot problem is time-consuming and inefficient.
Disclosure of Invention
In view of the above, it is necessary to provide a hotspot problem analysis method, apparatus, computer device, computer readable storage medium and computer program product for solving the above technical problems.
In a first aspect, the present application provides a hotspot problem analysis method, including:
obtaining a plurality of sentences to be analyzed from historical dialogue sentences;
determining a sentence to be analyzed with the highest matching degree with the key sentence library in the sentences to be analyzed according to the sentences to be analyzed and the key sentence library, and determining a key sentence corresponding to the sentence to be analyzed with the highest matching degree in the key sentence library as a candidate sentence;
and determining a hot spot statement from the candidate statements by using an attention mechanism algorithm, and determining a hot spot problem according to the hot spot statement.
In one embodiment, determining, according to a plurality of sentences to be analyzed and a key sentence library, a sentence to be analyzed that has the highest matching degree with the key sentence library, includes:
determining a sentence vector of each sentence to be analyzed and a sentence vector of each key sentence in the key sentence library;
calculating the similarity between the sentence vector of each sentence to be analyzed and the sentence vector of each key sentence;
determining the number of matching sentences matched with the sentences to be analyzed; the matching sentences are key sentences in the key sentence library, wherein the similarity between the key sentences and the sentences to be analyzed is greater than a first similarity threshold;
and determining the sentence to be analyzed with the matching sentence number larger than the number threshold value in each sentence to be analyzed as the sentence to be analyzed with the highest matching degree.
In one embodiment, obtaining a plurality of statements to be analyzed in a history dialogue statement comprises:
determining keywords of historical dialogue sentences;
and obtaining the sentences including the keywords in the historical dialogue sentences as the sentences to be analyzed.
In one embodiment, the method further includes:
determining the appearance proportion of a plurality of preset keywords in the historical dialogue sentences according to the historical dialogue sentences;
determining a plurality of key sentences corresponding to preset keywords from historical dialogue sentences according to the appearance proportion of each preset keyword and the preset capacity of the key sentence library;
and constructing a key sentence library based on a plurality of key sentences corresponding to the preset key words.
In one embodiment, determining the hot-spot statement from the candidate statements using an attention mechanism algorithm includes:
determining the similarity between each candidate statement and each statement to be analyzed by using an attention mechanism algorithm;
and determining hot sentences in the candidate sentences according to the similarity.
In one embodiment, determining a hot statement in the candidate statements according to the similarity includes:
and determining the ratio of the number of the similarities of which the corresponding similarity is greater than the second similarity threshold to the total number of the similarities, and if the ratio of the number of the similarities is greater than the ratio threshold, determining that the candidate sentence of which the corresponding similarity is greater than the second similarity threshold is the hot sentence.
In a second aspect, the present application further provides a hotspot problem analysis device, including:
the sentence acquisition module is used for acquiring a plurality of sentences to be analyzed from the historical dialogue sentences;
the matching analysis module is used for determining a sentence to be analyzed with the highest matching degree with the key sentence library in the sentences to be analyzed according to the sentences to be analyzed and the key sentence library, and determining a key sentence corresponding to the sentence to be analyzed with the highest matching degree in the key sentence library as a candidate sentence;
and the hot spot determining module is used for determining a hot spot statement from the candidate statements by using an attention mechanism algorithm and determining a hot spot problem according to the hot spot statement.
In a third aspect, the application also provides a computer device. The computer device comprises a memory storing a computer program and a processor implementing the following steps when executing the computer program:
obtaining a plurality of sentences to be analyzed from historical dialogue sentences;
determining a sentence to be analyzed with the highest matching degree with the key sentence library in the sentences to be analyzed according to the sentences to be analyzed and the key sentence library, and determining a key sentence corresponding to the sentence to be analyzed with the highest matching degree in the key sentence library as a candidate sentence;
and determining a hot spot statement from the candidate statements by using an attention mechanism algorithm, and determining a hot spot problem according to the hot spot statement.
In a fourth aspect, the present application further provides a computer-readable storage medium. The computer-readable storage medium having stored thereon a computer program which, when executed by a processor, performs the steps of:
obtaining a plurality of sentences to be analyzed from historical dialogue sentences;
determining a sentence to be analyzed with the highest matching degree with the key sentence library in the sentences to be analyzed according to the sentences to be analyzed and the key sentence library, and determining a key sentence corresponding to the sentence to be analyzed with the highest matching degree in the key sentence library as a candidate sentence;
and determining a hot spot statement from the candidate statements by using an attention mechanism algorithm, and determining a hot spot problem according to the hot spot statement.
In a fifth aspect, the present application further provides a computer program product. The computer program product comprising a computer program which when executed by a processor performs the steps of:
obtaining a plurality of sentences to be analyzed from historical dialogue sentences;
determining a sentence to be analyzed with the highest matching degree with the key sentence library in the sentences to be analyzed according to the sentences to be analyzed and the key sentence library, and determining a key sentence corresponding to the sentence to be analyzed with the highest matching degree in the key sentence library as a candidate sentence;
and determining a hot spot statement from the candidate statements by using an attention mechanism algorithm, and determining a hot spot problem according to the hot spot statement.
According to the method, the device, the computer equipment, the storage medium and the computer program product for analyzing the hot spot problem, a plurality of sentences to be analyzed are obtained from historical dialogue sentences, the sentences to be analyzed with the highest matching degree with the key sentence library in the sentences to be analyzed are determined according to the sentences to be analyzed and the key sentence library, the key sentences corresponding to the sentences to be analyzed with the highest matching degree in the key sentence library are determined as candidate sentences, and then the hot spot sentences are determined from the candidate sentences by using an attention mechanism algorithm so as to determine the hot spot problem according to the hot spot sentences. By the method, the problem of hot spots of manual statistics is avoided, the time consumed by statistics is reduced, and the analysis efficiency is improved.
Drawings
FIG. 1 is a schematic flow chart diagram of a hotspot problem analysis method in one embodiment;
FIG. 2 is a flow diagram illustrating the determination of a statement to be analyzed in one embodiment;
FIG. 3 is a diagram illustrating an exemplary process for determining a key sentence library;
FIG. 4 is a flow diagram illustrating the determination of a sentence to be analyzed with the highest degree of matching in one embodiment;
FIG. 5 is a flowchart illustrating the determination of hot spot statements in one embodiment;
FIG. 6 is a diagram illustrating the structure of a BiMPM model according to an embodiment;
FIG. 7 is a schematic flow chart illustrating 4 attention mechanism strategies in the BiMPM model according to an embodiment;
FIG. 8 is a block diagram of an embodiment of a hotspot problem analysis device;
FIG. 9 is a diagram illustrating an internal structure of a computer device according to an embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
In an embodiment, a hotspot problem analysis method is provided, and this embodiment is illustrated by applying the method to a terminal, and it can be understood that the method may also be applied to a server, and may also be applied to a system including a terminal and a server, and is implemented by interaction between the terminal and the server. The terminal can be but not limited to various personal computers, notebook computers, smart phones, tablet computers, internet of things devices and portable wearable devices, and the server can be realized by an independent server or a server cluster formed by a plurality of servers.
Note that, in the present embodiment, the hotspot problem (FAQ) is a problem that the consultation frequency and the popularity provided by the user reach a certain degree. The hot spot problem analysis method provided by the application is essentially a process for determining the hot spot problem.
As shown in fig. 1, in this embodiment, the method includes the following steps:
and S110, acquiring a plurality of sentences to be analyzed from the historical dialogue sentences.
The history dialogue sentences are sentences generated in the history dialogue process. Alternatively, the historical dialogue sentences may be consultation sentences received by the customer service system in the past period, or sentences obtained by manual customer service or communication between the customer service robot and the user through the customer service system in the past period.
Optionally, the sentence to be analyzed may be any sentence in the historical dialogue sentences, may also be a sentence in which the proportion/frequency of occurrence in the historical dialogue sentences reaches the corresponding proportion/frequency, and may also be a target sentence in the historical dialogue sentences. The target sentence is a sentence including a keyword. Optionally, the keyword may be a preset keyword, or a keyword determined based on a keyword analysis algorithm.
And S120, determining a sentence to be analyzed with the highest matching degree with the key sentence library in the sentences to be analyzed according to the sentences to be analyzed and the key sentence library, and determining the key sentence corresponding to the sentence to be analyzed with the highest matching degree in the key sentence library as a candidate sentence.
The key sentence library comprises a plurality of key sentences, and each key sentence comprises at least one industry keyword.
Optionally, the terminal determines a similarity between each sentence to be analyzed and each key sentence in the key sentence library, determines the sentence to be analyzed whose corresponding similarity satisfies a preset condition as the sentence to be analyzed with the highest matching degree with the key sentence library among the sentences to be analyzed, and further determines the key sentence corresponding to the sentence to be analyzed with the highest matching degree in the key sentence library as a candidate sentence. The preset condition may be that the similarity is greater than the similarity threshold and the obtained number of similarities greater than the similarity threshold is greater than the number threshold.
For example, the to-be-analyzed sentences include 100 (to-be-analyzed sentences 1 to 100), the keyword sentence library includes 1000 keyword sentences, and the terminal calculates the similarity between each to-be-analyzed sentence and each keyword sentence, so as to obtain 100 × 1000 similarities, that is, each to-be-analyzed sentence corresponds to 1000 similarities. If 400 similarity degrees of the 1000 similarity degrees corresponding to the sentence 10 to be analyzed are greater than the similarity threshold value, 350 similarity degrees of the 1000 similarity degrees corresponding to the sentence 20 to be analyzed are greater than the similarity threshold value, 351 similarity degrees of the 1000 similarity degrees corresponding to the sentence 21 to be analyzed are greater than the similarity threshold value, and the quantity threshold value is 300, the terminal can confirm that the sentence 10 to be analyzed, the sentence 20 to be analyzed and the sentence 21 to be analyzed are the sentences to be analyzed with the highest matching degree with the keyword library, and confirm that the 400, 350 and 351 similarity key sentences obtained respectively are candidate sentences.
Optionally, the similarity between each statement to be analyzed and each key statement may be determined according to a word vector of a keyword in the statement to be analyzed and a word vector of a keyword in the key statement, and may also be determined according to a sentence vector of the statement to be analyzed and a sentence vector of the key statement.
S130, determining hot spot sentences from the candidate sentences by using an attention mechanism algorithm, and determining hot spot problems according to the hot spot sentences.
In cognitive science, humans selectively focus on a portion of all information while ignoring other information due to bottlenecks in information processing, a mechanism commonly referred to as attentiveness mechanism.
Optionally, the terminal may determine similarity between the sentence to be analyzed and the candidate sentences by using a machine learning model based on an attention mechanism algorithm, determine a hot sentence from the candidate sentences, and then determine a hot problem by using the hot sentence.
Optionally, the terminal sets a corresponding concern question for each key sentence in the key sentence library, and after determining a hot spot sentence in the key sentence, the terminal may determine, according to a preset correspondence between the key sentence and the concern question, that the concern question corresponding to the hot spot sentence is the hot spot question.
In this embodiment, the terminal obtains a plurality of sentences to be analyzed from the historical dialogue sentences, determines, according to the plurality of sentences to be analyzed and the key sentence library, a sentence to be analyzed having the highest matching degree with the key sentence library from the plurality of sentences to be analyzed, determines a key sentence in the key sentence library corresponding to the sentence to be analyzed having the highest matching degree as a candidate sentence, and further determines a hot sentence from the candidate sentences by using an attention mechanism algorithm to determine a hot question according to the hot sentence. By the method, the problem of hot spots of manual statistics is avoided, the time consumed by statistics is reduced, and the analysis efficiency is improved.
In one embodiment, the sentence to be analyzed is a sentence including a keyword, as shown in fig. 2, the S110 includes:
s210, determining keywords of the historical dialogue sentences.
Wherein, the keywords of the historical dialogue sentences can be used for representing the subjects of the historical dialogue sentences. Alternatively, a plurality of keywords may be included in the history dialogue sentences.
Alternatively, the historical dialogue sentences may be speech data or text data. If the historical dialogue sentence is voice data, the terminal performs conversion processing of converting the voice of the historical dialogue sentence into text data in the same format, such as text data in an xml format. And if the historical dialogue sentences are text data, directly performing format identity processing on the historical dialogue sentences, and performing preprocessing such as analysis, data cleaning and the like on the historical dialogue sentences with the same format to obtain the processed historical dialogue sentences.
Optionally, the terminal may count the occurrence frequency of different words in the historical dialogue sentences, and perform common word elimination processing on the words with the occurrence frequency greater than the frequency threshold value to obtain the keywords. Keywords in the historical dialog sentences may also be determined using a keyword algorithm. For example, the historical dialogue sentences are input into a TF-IDF model (TF: term Frequency; IDF: inverse Document Frequency) based on a keyword algorithm, and keywords of the historical dialogue sentences are obtained.
S220, obtaining the sentences including the keywords in the historical dialogue sentences as the sentences to be analyzed.
Specifically, after determining the keywords of the historical dialogue sentences, the terminal traverses each historical dialogue sentence, and obtains the sentences including the keywords from the historical dialogue sentences as the sentences to be analyzed.
In this embodiment, the terminal first determines the keywords of the historical dialogue sentences, and then obtains the sentences including the keywords in the historical dialogue sentences as the sentences to be analyzed, so as to obtain the sentences to be analyzed including the keywords, thereby implementing data compression on the historical dialogue sentences, reducing subsequent data calculation amount, and further improving analysis efficiency.
In one embodiment, the method further includes a process of determining a key sentence library, as shown in fig. 3, and the method further includes:
s310, determining the appearance proportion of a plurality of preset keywords in the historical dialogue sentences according to the historical dialogue sentences.
The preset keywords are industry keywords, namely keywords related to industry. For example, for the insurance industry, the preset keywords may be insurance policy, premium, security, premium, compensation, etc., and may also be the name of the insurance policy product, etc.
Specifically, the terminal may traverse the historical dialogue sentences, determine the historical dialogue sentences including at least one preset keyword, and count the occurrence frequency of each preset keyword in the historical dialogue sentences (for simplification, the occurrence frequency of each preset keyword in a sentence is counted once), so as to obtain the occurrence ratio of each preset keyword in the historical dialogue sentences. For example, the preset keywords include a keyword a, a keyword B and a keyword C, the historical dialogue sentences include 10000 pieces, the frequency of occurrence of the keyword a in the historical dialogue sentences is 5000 times, the frequency of occurrence of the keyword B in the historical dialogue sentences is 3000 times, the frequency of occurrence of the keyword C in the historical dialogue sentences is 1000 times, and thus the occurrence ratio of the keyword a in the historical dialogue sentences is 5/9, the occurrence ratio of the keyword B in the historical dialogue sentences is 3/9, and the occurrence ratio of the keyword C in the historical dialogue sentences is 1/9.
And S320, determining a plurality of key sentences corresponding to the preset keywords from the historical dialogue sentences according to the occurrence proportion of each preset keyword and the preset capacity of the key sentence library.
S330, constructing a key sentence library based on a plurality of key sentences corresponding to the preset key words.
Specifically, after determining the occurrence ratio of each preset keyword in the historical dialogue sentences, the terminal determines the number of the historical dialogue sentences including the corresponding preset keyword according to the occurrence ratio of each preset keyword and the preset capacity of the key sentence library, acquires the corresponding number of key sentences corresponding to the preset keyword from the historical dialogue sentences, and further constructs the key sentence library from the corresponding number of key sentences corresponding to the preset keyword.
Continuing with the above example, the preset capacity of the keyword sentence library is 1000, and the number of the historical dialog sentences including the corresponding preset keywords is determined by combining the occurrence ratios of the determined keywords a, B and C: the number of the history dialog sentences including the keyword a is (5/9) × 1000 ≈ 556 pieces, the number of the history dialog sentences including the keyword B is (3/9) × 1000 ≈ 333 pieces, and the number of the history dialog sentences including the keyword C is (1/9) × 1000 ≈ 111 pieces. Correspondingly, the terminal extracts 556 historical dialogue sentences including the keywords a, 333 historical dialogue sentences including the keywords B and 111 historical dialogue sentences including the keywords C from the historical dialogue sentences, and forms the keyword sentence library by using the 556 historical dialogue sentences including the keywords a, the 333 historical dialogue sentences including the keywords B and the 111 historical dialogue sentences including the keywords C as the key sentences.
In this embodiment, the terminal determines the occurrence ratio of a plurality of preset keywords in the historical dialogue sentences according to the historical dialogue sentences, determines a plurality of key sentences corresponding to the preset keywords from the historical dialogue sentences according to the occurrence ratio of each preset keyword and the preset capacity of the key sentence library, and further constructs the key sentence library based on the plurality of key sentences corresponding to the preset keywords. The key sentence library determined by the method keeps the same proportion of the occurrence of the keywords with the historical dialogue sentences, so that the heat condition of each keyword can be accurately reflected, the accuracy of subsequently determining the hot sentences is improved, and the accuracy of determining the hot problems based on the hot sentences is further improved.
In one embodiment, the similarity between the sentence to be analyzed and the key sentence may be determined by using a sentence vector, as shown in fig. 4, the determining the sentence to be analyzed with the highest matching degree with the key sentence library in the plurality of sentences to be analyzed in S120 includes:
s410, determining a sentence vector of each sentence to be analyzed and a sentence vector of each key sentence in the key sentence library.
The sentence vector is a mapping vector of the sentence in the mathematical space, and is used for representing semantic features of the corresponding sentence. Alternatively, the sentence vector may be determined according to the word vector of the corresponding sentence, and may also be obtained based on a natural language processing technique.
Alternatively, the terminal may input each sentence to be analyzed into a bert model using a natural language processing technique, and output a sentence vector of each sentence to be analyzed by the bert model, and similarly, input each key sentence in the key sentence library into the bert model, and output a sentence vector of each key sentence in the key sentence library correspondingly.
S420, calculating the similarity between the sentence vector of each statement to be analyzed and the sentence vector of each key statement.
Specifically, the terminal may calculate a cosine similarity between the sentence vector of each sentence to be analyzed and the sentence vector of each key sentence.
And S430, determining the number of the matched sentences matched with the sentences to be analyzed.
The matching sentences are key sentences in the key sentence library, wherein the similarity between the key sentences and the sentences to be analyzed is greater than a first similarity threshold.
S440, determining the sentences to be analyzed with the number of the matched sentences larger than the number threshold value in each sentence to be analyzed as the sentences to be analyzed with the highest matching degree.
Specifically, the terminal may determine, according to the cosine similarity, a key sentence matching the sentence to be analyzed, that is, the matching sentence, in the key sentence library. And if the cosine similarity is greater than a first similarity threshold, determining that the sentence to be analyzed with the cosine similarity is matched with the key sentence. The terminal can obtain the matching sentences matched with the sentences to be analyzed in the key sentence library, further count the number of the matching sentences corresponding to each sentence to be analyzed, compare the number of the matching sentences corresponding to each sentence to be analyzed with the number threshold value, and determine the sentences to be analyzed with the number of the matching sentences larger than the number threshold value as the sentences to be analyzed with the highest matching degree. For a specific process, see the example in S120, which is not described herein again.
Optionally, the terminal may further input the sentence vector of the sentence to be analyzed into a lsh (Locality Sensitive Hashing) model, and query the initial hot question sentence generated by hierarchical sampling in a lsh model for a similar sentence, so as to obtain a final sentence to be analyzed.
In this embodiment, the terminal determines a sentence vector of each sentence to be analyzed and a sentence vector of each key sentence in the key sentence library to calculate a similarity between the sentence vector of each sentence to be analyzed and the sentence vector of each key sentence, determines the number of matching sentences having a similarity greater than a first similarity threshold with the sentence to be analyzed, and further determines the sentence to be analyzed having the highest matching degree, of which the number of matching sentences in each sentence to be analyzed is greater than a number threshold, as the sentence to be analyzed. The sentence vectors can accurately reflect the semantic features of the sentences, and the similarity between the two sentences can be accurately determined based on the similarity between the sentence vectors, so that the accuracy of the determined sentence to be analyzed with the highest matching degree is improved.
In one embodiment, to further improve the accuracy of the determined hotspot problem, as shown in fig. 5, the process of the hotspot statement in S130 includes:
and S510, determining the similarity between each candidate statement and each statement to be analyzed by using an attention mechanism algorithm.
Specifically, the terminal may input a sentence pair composed of each candidate sentence and each sentence to be analyzed into a machine learning model based on an attention mechanism algorithm, such as a bimp (binary Multi-Perspective Matching Network) model, and output the similarity of the sentence pair by the bimp model. For example, the candidate sentences include 200 sentences, the to-be-analyzed sentences include 100 sentences, and each candidate sentence and each to-be-analyzed sentence constitute a sentence pair and are input into the bimp model, so that the similarity corresponding to 200 × 100 sentence pairs, that is, 200 × 100 similarities is obtained.
The BiMPM model is an attention mechanism model, and is specifically introduced as follows:
preparing data:
and (4) adopting a third-party financial similar sentence data set and an open source financial word vector, and utilizing the pitcher to carry out data preprocessing to generate a word vector file and a dictionary. And converting the candidate sentence pair data set of the similar sentences into a pickle object data format, wherein the objects are id lists corresponding to words in the sentence pairs of the similar sentences obtained according to a dictionary.
Model structure:
as shown in fig. 6, the BiMPM model is divided into 5 layers: a word representation Layer (Embedding Layer), a coding Layer (Encoding Layer), a Matching Layer (Matching Layer), a fused Layer (Aggregated Layer), and a prediction Layer. The core of this is the 4 attention mechanism strategies shown in fig. 7: a) full match, b) max pooling match, c) attention match, d) max attention match.
And (3) realizing a model:
model structure is implemented by a pytorch. The core of the method is the design of a matching layer, and 4 attention mechanism strategies aim to fully fuse input sequences and achieve a better effect of judging semantic similarity. The core idea is that based on the output of a coding layer (a bidirectional lstm), the similarity weight (cosine similarity) of two sentences is calculated, then the weight is acted on the hidden vector of the corresponding word, and then the hidden vector is operated with the cosine matching matrix of each strategy to obtain the cosine matching value of each strategy as the output, and the output results of 4 strategies are fused.
Model training and verification:
the model input is a sentence-to-word id list generated in the preprocessing stage, and the output is 0 or 1, which represents whether the sentences are similar or not. Model training epoch used 50 rounds and the optimizer used Adam. And performing model verification once after every 200 batches of training, wherein the batch size is 32, and the model is stored when the verification precision is higher than the maximum precision. If the model does not converge continuously in the training process, the training of the current round is ended in advance, and the next round of training is started.
And (3) model evaluation:
training used open source third party financial and industry datasets for a total of 80 thousands. The ratio of similar to non-similar sentences is approximately 7:3, this is due to the model's consideration for similar sentence learning. The model precision can reach more than 91%.
And (3) outputting a model:
and outputting similar sentence pairs with the similarity larger than a given threshold, wherein each hot question corresponds to 10 similar questions. When ten million pieces of input data are input, 20-50 hot spot problems can be guaranteed to be output. This depends on the number of rough question candidates and the similarity threshold.
S520, determining hot spot sentences in the candidate sentences according to the similarity.
Specifically, based on the determined similarity of all sentence pairs, the terminal further determines the number of similarities of which the corresponding similarities are greater than a second similarity threshold, calculates the number ratio of the number of similarities to the total number of similarities, and determines that the candidate sentence of which the corresponding similarity is greater than the second similarity threshold is the hot sentence if the number ratio is greater than the ratio threshold. For example, continuing the above example, the terminal obtains 200 × 100 similarities through the above calculation, obtains, through statistics, 10001 similarities among the 200 × 100 similarities are greater than the second similarity threshold, further determines that a ratio 10001/20000 of the number of similarities 10001 whose corresponding similarities are greater than the second similarity threshold to the total number of similarities 20000 is greater than a ratio threshold 50%, and the terminal may determine that the 10001 candidate sentences whose corresponding similarities are greater than the second similarity threshold are hot sentences and output.
Optionally, if the number ratio is not greater than the proportional threshold, it is determined that none of the candidate sentences reaches the hot scale, and an output result that the hot sentence cannot be determined may be output.
In this embodiment, the terminal determines the similarity between each candidate statement and each statement to be analyzed by using an attention mechanism algorithm, determines a hot statement in the candidate statement according to the similarity, specifically determines a quantity ratio of the quantity of the similarities, the corresponding similarities of which are greater than a second similarity threshold, to a total quantity of the similarities, and determines that the candidate statement, the corresponding similarities of which are greater than the second similarity threshold, is the hot statement when the quantity ratio is greater than a ratio threshold. Therefore, hot spot sentences which are highly related to the industry and high in frequency are determined, and the accuracy of the determined hot spot problem is further improved.
It should be understood that, although the steps in the flowcharts related to the embodiments as described above are sequentially displayed as indicated by arrows, the steps are not necessarily performed sequentially as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least a part of the steps in the flowcharts related to the embodiments described above may include multiple steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, and the execution order of the steps or stages is not necessarily sequential, but may be rotated or alternated with other steps or at least a part of the steps or stages in other steps.
Based on the same inventive concept, the embodiment of the present application further provides a hotspot problem analysis device for implementing the hotspot problem analysis method. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme described in the above method, so specific limitations in one or more embodiments of the hotspot problem analysis device provided below can be referred to the limitations of the hotspot problem analysis method in the above description, and details are not repeated herein.
In one embodiment, as shown in fig. 8, there is provided a hotspot problem analysis device, including: statement acquisition module 801, matching analysis module 802, and hotspot determination module 803, wherein:
the statement acquisition module 801 is configured to acquire a plurality of statements to be analyzed from a historical dialogue statement;
the matching analysis module 802 is configured to determine, according to the multiple sentences to be analyzed and the key sentence library, a sentence to be analyzed that has the highest matching degree with the key sentence library among the multiple sentences to be analyzed, and determine a key sentence in the key sentence library that corresponds to the sentence to be analyzed that has the highest matching degree as a candidate sentence;
the hot spot determining module 803 is configured to determine a hot spot statement from the candidate statements by using an attention mechanism algorithm, and determine a hot spot problem according to the hot spot statement.
In one embodiment, the matching analysis module 802 is specifically configured to:
determining a sentence vector of each sentence to be analyzed and a sentence vector of each key sentence in the key sentence library; calculating the similarity between the sentence vector of each sentence to be analyzed and the sentence vector of each key sentence; determining the number of matching sentences matched with the sentences to be analyzed; the matching sentences are key sentences in the key sentence library, wherein the similarity between the key sentences and the sentences to be analyzed is greater than a first similarity threshold; and determining the sentence to be analyzed with the matching sentence number larger than the number threshold value in each sentence to be analyzed as the sentence to be analyzed with the highest matching degree.
In one embodiment, the statement obtaining module 801 is specifically configured to:
determining keywords of historical dialogue sentences; and obtaining the sentences including the keywords in the historical dialogue sentences as the sentences to be analyzed.
In one embodiment, the apparatus further comprises a key determination module; the key determination module is specifically configured to:
determining the appearance proportion of a plurality of preset keywords in the historical dialogue sentences according to the historical dialogue sentences; determining a plurality of key sentences corresponding to preset keywords from historical dialogue sentences according to the appearance proportion of each preset keyword and the preset capacity of the key sentence library; and constructing a key sentence library based on a plurality of key sentences corresponding to the preset key words.
In one embodiment, the hotspot determination module 803 is specifically configured to:
determining the similarity between each candidate statement and each statement to be analyzed by using an attention mechanism algorithm; and determining hot-spot sentences in the candidate sentences according to the similarity.
In one embodiment, the hotspot determination module 803 is specifically configured to:
and determining the ratio of the number of the similarities of which the corresponding similarity is greater than the second similarity threshold to the total number of the similarities, and if the ratio of the number of the similarities is greater than the ratio threshold, determining that the candidate sentence of which the corresponding similarity is greater than the second similarity threshold is the hot sentence.
The modules in the hotspot problem analysis device can be wholly or partially realized by software, hardware and a combination thereof. The modules can be embedded in a hardware form or independent of a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.
In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as shown in fig. 9. The computer device includes a processor, a memory, a communication interface, a display screen, and an input device 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 comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The communication interface of the computer device is used for carrying out wired or wireless communication with an external terminal, and the wireless communication can be realized through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. The computer program is executed by a processor to implement a hotspot problem analysis method. The display screen of the computer equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, a key, a track ball or a touch pad arranged on the shell of the computer equipment, an external keyboard, a touch pad or a mouse and the like.
Those skilled in the art will appreciate that the architecture shown in fig. 9 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those 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 a computer program stored therein, the processor implementing the following steps when executing the computer program:
obtaining a plurality of sentences to be analyzed from historical dialogue sentences; determining a sentence to be analyzed with the highest matching degree with the key sentence library in the sentences to be analyzed according to the sentences to be analyzed and the key sentence library, and determining a key sentence corresponding to the sentence to be analyzed with the highest matching degree in the key sentence library as a candidate sentence; and determining a hot spot statement from the candidate statements by using an attention mechanism algorithm, and determining a hot spot problem according to the hot spot statement.
In one embodiment, the processor when executing the computer program further performs the steps of:
determining a sentence vector of each sentence to be analyzed and a sentence vector of each key sentence in the key sentence library; calculating the similarity between the sentence vector of each sentence to be analyzed and the sentence vector of each key sentence; determining the number of matching sentences matched with the sentences to be analyzed; the matching sentences are key sentences in the key sentence library, wherein the similarity between the key sentences and the sentences to be analyzed is greater than a first similarity threshold; and determining the sentence to be analyzed with the matching sentence number larger than the number threshold value in each sentence to be analyzed as the sentence to be analyzed with the highest matching degree.
In one embodiment, the processor, when executing the computer program, further performs the steps of:
determining keywords of historical dialogue sentences; and obtaining the sentences including the keywords in the historical dialogue sentences as the sentences to be analyzed.
In one embodiment, the processor, when executing the computer program, further performs the steps of:
determining the appearance proportion of a plurality of preset keywords in the historical dialogue sentences according to the historical dialogue sentences; determining a plurality of key sentences corresponding to preset keywords from historical dialogue sentences according to the appearance proportion of each preset keyword and the preset capacity of the key sentence library; and constructing a key sentence library based on a plurality of key sentences corresponding to the preset key words.
In one embodiment, the processor, when executing the computer program, further performs the steps of:
determining the similarity between each candidate statement and each statement to be analyzed by using an attention mechanism algorithm; and determining hot sentences in the candidate sentences according to the similarity.
In one embodiment, the processor, when executing the computer program, further performs the steps of:
and determining the ratio of the number of the similarities of which the corresponding similarity is greater than the second similarity threshold to the total number of the similarities, and if the ratio of the number of the similarities is greater than the ratio threshold, determining that the candidate sentence of which the corresponding similarity is greater than the second similarity threshold is the hot sentence.
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 a plurality of sentences to be analyzed from historical dialogue sentences; determining a sentence to be analyzed with the highest matching degree with the key sentence library in the sentences to be analyzed according to the sentences to be analyzed and the key sentence library, and determining a key sentence corresponding to the sentence to be analyzed with the highest matching degree in the key sentence library as a candidate sentence; and determining a hot spot statement from the candidate statements by using an attention mechanism algorithm, and determining a hot spot problem according to the hot spot statement.
In one embodiment, the computer program when executed by the processor further performs the steps of:
determining a sentence vector of each sentence to be analyzed and a sentence vector of each key sentence in the key sentence library; calculating the similarity between the sentence vector of each sentence to be analyzed and the sentence vector of each key sentence; determining the number of matching sentences matched with the sentences to be analyzed; the matching sentences are key sentences in the key sentence library, wherein the similarity between the key sentences and the sentences to be analyzed is greater than a first similarity threshold; and determining the sentences to be analyzed with the matching sentence number larger than the number threshold value in each sentence to be analyzed as the sentences to be analyzed with the highest matching degree.
In one embodiment, the computer program when executed by the processor further performs the steps of:
determining keywords of historical dialogue sentences; and obtaining the sentences including the keywords in the historical dialogue sentences as the sentences to be analyzed.
In one embodiment, the computer program when executed by the processor further performs the steps of:
determining the appearance proportion of a plurality of preset keywords in the historical dialogue sentences according to the historical dialogue sentences; determining a plurality of key sentences corresponding to preset keywords from historical dialogue sentences according to the appearance proportion of each preset keyword and the preset capacity of the key sentence library; and constructing a key sentence library based on a plurality of key sentences corresponding to the preset key words.
In one embodiment, the computer program when executed by the processor further performs the steps of:
determining the similarity between each candidate statement and each statement to be analyzed by using an attention mechanism algorithm; and determining hot sentences in the candidate sentences according to the similarity.
In one embodiment, the computer program when executed by the processor further performs the steps of:
and determining the ratio of the number of the similarities of which the corresponding similarity is greater than the second similarity threshold to the total number of the similarities, and if the ratio of the number of the similarities is greater than the ratio threshold, determining that the candidate sentence of which the corresponding similarity is greater than the second similarity threshold is the hot sentence.
In one embodiment, a computer program product is provided, comprising a computer program which, when executed by a processor, performs the steps of:
obtaining a plurality of sentences to be analyzed from historical dialogue sentences; determining a sentence to be analyzed with the highest matching degree with the key sentence library in the sentences to be analyzed according to the sentences to be analyzed and the key sentence library, and determining a key sentence corresponding to the sentence to be analyzed with the highest matching degree in the key sentence library as a candidate sentence; and determining a hot spot statement from the candidate statements by using an attention mechanism algorithm, and determining a hot spot problem according to the hot spot statement.
In one embodiment, the computer program when executed by the processor further performs the steps of:
determining a sentence vector of each sentence to be analyzed and a sentence vector of each key sentence in the key sentence library; calculating the similarity between the sentence vector of each sentence to be analyzed and the sentence vector of each key sentence; determining the number of matching sentences matched with the sentences to be analyzed; the matching sentences are key sentences in the key sentence library, wherein the similarity between the key sentences and the sentences to be analyzed is greater than a first similarity threshold; and determining the sentence to be analyzed with the matching sentence number larger than the number threshold value in each sentence to be analyzed as the sentence to be analyzed with the highest matching degree.
In one embodiment, the computer program when executed by the processor further performs the steps of:
determining keywords of historical dialogue sentences; and obtaining the sentences including the keywords in the historical dialogue sentences as the sentences to be analyzed.
In one embodiment, the computer program when executed by the processor further performs the steps of:
determining the appearance proportion of a plurality of preset keywords in the historical dialogue sentences according to the historical dialogue sentences; determining a plurality of key sentences corresponding to preset keywords from historical dialogue sentences according to the occurrence proportion of each preset keyword and the preset capacity of the key sentence library; and constructing a key sentence library based on a plurality of key sentences corresponding to the preset key words.
In one embodiment, the computer program when executed by the processor further performs the steps of:
determining the similarity between each candidate statement and each statement to be analyzed by using an attention mechanism algorithm; and determining hot sentences in the candidate sentences according to the similarity.
In one embodiment, the computer program when executed by the processor further performs the steps of:
and determining the ratio of the number of the similarities of which the corresponding similarity is greater than the second similarity threshold to the total number of the similarities, and if the ratio of the number of the similarities is greater than the ratio threshold, determining that the candidate sentence of which the corresponding similarity is greater than the second similarity threshold is the hot sentence.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, database, or other medium used in the 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, high-density embedded nonvolatile Memory, resistive Random Access Memory (ReRAM), magnetic Random Access Memory (MRAM), ferroelectric Random Access Memory (FRAM), phase Change Memory (PCM), graphene Memory, and the like. Volatile Memory can include Random Access Memory (RAM), external cache Memory, and the like. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM), among others. The databases referred to in various embodiments provided herein may include at least one of relational and non-relational databases. The non-relational database may include, but is not limited to, a block chain based distributed database, and the like. The processors referred to in the embodiments provided herein may be general purpose processors, central processing units, graphics processors, digital signal processors, programmable logic devices, quantum computing based data processing logic devices, etc., without limitation.
All possible combinations of the technical features in the above embodiments may not be described for the sake of brevity, but should be considered as being within the scope of the present disclosure as long as there is no contradiction between the combinations of the technical features.
The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is specific and detailed, but not construed as limiting the scope of the present application. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present application shall be subject to the appended claims.
Claims (10)
1. A method for hotspot problem analysis, the method comprising:
obtaining a plurality of sentences to be analyzed from historical dialogue sentences;
determining a sentence to be analyzed with the highest matching degree with the key sentence library in the sentences to be analyzed according to the sentences to be analyzed and the key sentence library, and determining a key sentence corresponding to the sentence to be analyzed with the highest matching degree in the key sentence library as a candidate sentence; the key sentences in the key sentence library are sentences comprising preset key words, and the occurrence proportion of the key words in the key sentence library is the same as that of the key words in the historical dialogue sentences; the sentence to be analyzed with the highest matching degree is the sentence to be analyzed, wherein the similarity between the plurality of sentences to be analyzed and each key sentence in the key sentence library is greater than a similarity threshold value, and the number of the key sentences the similarity of which is greater than the similarity threshold value is greater than a number threshold value;
and determining a hot spot statement from the candidate statements by using an attention mechanism algorithm, and determining a hot spot problem according to the hot spot statement.
2. The method according to claim 1, wherein said determining the sentence to be analyzed having the highest matching degree with the keyword sentence library from the sentences to be analyzed and the keyword sentence library comprises:
determining a sentence vector of each sentence to be analyzed and a sentence vector of each key sentence in the key sentence library;
calculating the similarity between the sentence vector of each sentence to be analyzed and the sentence vector of each key sentence;
determining the number of matching sentences matched with the sentences to be analyzed; the matching statement is a key statement in the key statement library, wherein the similarity between the matching statement and the statement to be analyzed is greater than a first similarity threshold;
and determining the sentence to be analyzed with the matching sentence number larger than the number threshold value in each sentence to be analyzed as the sentence to be analyzed with the highest matching degree.
3. The method of claim 1, wherein the obtaining a plurality of sentences to be analyzed from the historical conversational sentences comprises:
determining keywords of the historical dialogue sentences;
and obtaining the sentence of the historical dialogue sentence, which comprises the keyword, as the sentence to be analyzed.
4. The method of any one of claims 1~3, further comprising:
determining the appearance proportion of a plurality of preset keywords in the historical dialogue sentences according to the historical dialogue sentences;
determining a plurality of key sentences corresponding to the preset keywords from the historical dialogue sentences according to the occurrence proportion of each preset keyword and the preset capacity of the key sentence library;
and constructing the key sentence library based on a plurality of key sentences corresponding to the preset key words.
5. The method of claim 1, wherein the determining a hot spot statement from the candidate statements using an attention mechanism algorithm comprises:
determining the similarity between each candidate statement and each statement to be analyzed by using the attention mechanism algorithm;
and determining the hot sentence in the candidate sentences according to the similarity.
6. The method of claim 5, wherein the determining the hot sentence in the candidate sentences according to the similarity comprises:
and determining the ratio of the number of the similarities of which the corresponding similarities are greater than the second similarity threshold to the total number of the similarities, and if the ratio of the number of the similarities is greater than a proportional threshold, determining that the candidate sentence of which the corresponding similarity is greater than the second similarity threshold is the hot sentence.
7. An apparatus for hotspot problem analysis, the apparatus comprising:
the sentence acquisition module is used for acquiring a plurality of sentences to be analyzed from the historical dialogue sentences;
the matching analysis module is used for determining a sentence to be analyzed with the highest matching degree with the key sentence library in the sentences to be analyzed according to the sentences to be analyzed and the key sentence library, and determining a key sentence corresponding to the sentence to be analyzed with the highest matching degree in the key sentence library as a candidate sentence; the key sentences in the key sentence library are sentences comprising preset key words, and the occurrence proportion of the key words in the key sentence library is the same as that of the key words in the historical dialogue sentences; the sentence to be analyzed with the highest matching degree is the sentence to be analyzed, wherein the similarity between the plurality of sentences to be analyzed and each key sentence in the key sentence library is greater than a similarity threshold value, and the number of the key sentences the similarity of which is greater than the similarity threshold value is greater than a number threshold value;
and the hot spot determining module is used for determining a hot spot statement from the candidate statements by using an attention mechanism algorithm and determining a hot spot problem according to the hot spot statement.
8. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor, when executing the computer program, implements the steps of the method of any of claims 1 to 6.
9. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 6.
10. A computer program product comprising a computer program, characterized in that the computer program realizes the steps of the method of any one of claims 1 to 6 when executed by a processor.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210931690.7A CN115017291B (en) | 2022-08-04 | 2022-08-04 | Hotspot problem analysis method and device, computer equipment and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210931690.7A CN115017291B (en) | 2022-08-04 | 2022-08-04 | Hotspot problem analysis method and device, computer equipment and storage medium |
Publications (2)
Publication Number | Publication Date |
---|---|
CN115017291A CN115017291A (en) | 2022-09-06 |
CN115017291B true CN115017291B (en) | 2023-01-10 |
Family
ID=83065852
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210931690.7A Active CN115017291B (en) | 2022-08-04 | 2022-08-04 | Hotspot problem analysis method and device, computer equipment and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115017291B (en) |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103500175A (en) * | 2013-08-13 | 2014-01-08 | 中国人民解放军国防科学技术大学 | Method for microblog hot event online detection based on emotion analysis |
CN112100617A (en) * | 2020-09-15 | 2020-12-18 | 全球能源互联网研究院有限公司 | Abnormal SQL detection method and device |
Family Cites Families (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110852095B (en) * | 2018-08-02 | 2023-09-19 | ***股份有限公司 | Statement hot spot extraction method and system |
CN112650829B (en) * | 2019-10-11 | 2024-06-25 | 阿里巴巴集团控股有限公司 | Customer service processing method and device |
CN111061831A (en) * | 2019-10-29 | 2020-04-24 | 深圳绿米联创科技有限公司 | Method and device for switching machine customer service to manual customer service and electronic equipment |
CN111143530B (en) * | 2019-12-24 | 2024-04-05 | 平安健康保险股份有限公司 | Intelligent reply method and device |
-
2022
- 2022-08-04 CN CN202210931690.7A patent/CN115017291B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103500175A (en) * | 2013-08-13 | 2014-01-08 | 中国人民解放军国防科学技术大学 | Method for microblog hot event online detection based on emotion analysis |
CN112100617A (en) * | 2020-09-15 | 2020-12-18 | 全球能源互联网研究院有限公司 | Abnormal SQL detection method and device |
Also Published As
Publication number | Publication date |
---|---|
CN115017291A (en) | 2022-09-06 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109376222B (en) | Question-answer matching degree calculation method, question-answer automatic matching method and device | |
US11704500B2 (en) | Techniques to add smart device information to machine learning for increased context | |
CN110858226A (en) | Conversation management method and device | |
CN113239176B (en) | Semantic matching model training method, device, equipment and storage medium | |
US20220114644A1 (en) | Recommendation system with sparse feature encoding | |
CN114841161A (en) | Event element extraction method, device, equipment, storage medium and program product | |
CN113343091A (en) | Industrial and enterprise oriented science and technology service recommendation calculation method, medium and program | |
CN116308551A (en) | Content recommendation method and system based on digital financial AI platform | |
Wang et al. | Application of an emotional classification model in e-commerce text based on an improved transformer model | |
CN112749557A (en) | Text processing model construction method and text processing method | |
CN112069404A (en) | Commodity information display method, device, equipment and storage medium | |
CN115017291B (en) | Hotspot problem analysis method and device, computer equipment and storage medium | |
JP7213890B2 (en) | Accelerated large-scale similarity computation | |
CN116401522A (en) | Financial service dynamic recommendation method and device | |
CN111858899B (en) | Statement processing method, device, system and medium | |
CN114429384A (en) | Intelligent product recommendation method and system based on e-commerce platform | |
CN114490996B (en) | Intention recognition method and device, computer equipment and storage medium | |
CN116151331B (en) | Training method of commodity marketing text generation model and commodity marketing text generation method | |
CN115374273A (en) | Session text analysis method, device, computer equipment and storage medium | |
CN116860972A (en) | Interactive information classification method, device, apparatus, storage medium and program product | |
CN117436501A (en) | Training method and device for product recommendation model, computer equipment and storage medium | |
CN117010334A (en) | Text information generation method, device, computer equipment and storage medium | |
CN114637926A (en) | Content recommendation method and device, electronic equipment and storage medium | |
CN117593090A (en) | Graph structure prediction model training method and related device based on multi-task learning | |
CN116796729A (en) | Text recommendation method, device, equipment and storage medium based on feature enhancement |
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 |