WO2020082609A1 - 利用深度学习模型的证券研报分析方法及装置 - Google Patents
利用深度学习模型的证券研报分析方法及装置 Download PDFInfo
- Publication number
- WO2020082609A1 WO2020082609A1 PCT/CN2019/070287 CN2019070287W WO2020082609A1 WO 2020082609 A1 WO2020082609 A1 WO 2020082609A1 CN 2019070287 W CN2019070287 W CN 2019070287W WO 2020082609 A1 WO2020082609 A1 WO 2020082609A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- clause
- research report
- analyzed
- opinion
- securities research
- 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.)
- Ceased
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
- G06F40/35—Discourse or dialogue representation
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; 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 OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/284—Lexical analysis, e.g. tokenisation or collocates
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; 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
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; 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
Definitions
- This application relates to the field of computer technology, and in particular to a method and device for analyzing securities research reports using a deep learning model.
- Securities research report also referred to as securities research report, refers to the research report made by relevant researchers (such as researchers in a securities company, etc.) to analyze the value of securities and related products, or factors that affect their market prices. .
- the purpose of this application is to provide a securities research report analysis method, device, computer equipment, and readable storage medium using a deep learning model, and to a certain extent, overcome the above-mentioned problems in the prior art and improve the securities research report Analysis efficiency and accuracy.
- a method for analyzing securities research reports using a deep learning model including the following steps:
- S02 Perform clause processing on the securities research report to be analyzed to obtain each clause in the securities research report to be analyzed;
- S05 use the pre-trained third deep learning model to score the opinion clauses of each industry in the securities research report to be analyzed, to obtain a third score for each industry opinion category in the securities research report to be analyzed, The third score is used to determine the sentiment trend of opinion clauses in various industries in the securities research report to be analyzed;
- S06 Determine the overall sentiment trend of the securities research report to be analyzed based on the second score and the third score of the opinion clauses of various industries in the securities research report to be analyzed.
- this application also provides a securities research report analysis device using a deep learning model, including:
- the receiving module is used to receive the input securities research report to be analyzed
- Sentence module which is used to perform sentence processing on the securities research report to be analyzed to obtain each clause in the securities research report to be analyzed;
- the first scoring module is used to score each clause in the securities research report to be analyzed using a pre-trained first deep learning model to obtain a first score for each clause in the securities research report to be analyzed, and Determine opinion clauses from each clause in the securities research report to be analyzed based on the first score;
- the second scoring module is used to score each opinion clause in the securities research report to be analyzed using a pre-trained second deep learning model to obtain the first sentence for each opinion category clause in the securities research report to be analyzed Second score, and based on the second score, determine the industry opinion clause from each opinion clause in the securities research report to be analyzed;
- the third scoring module is used to score the perspective clauses of various industries in the securities research report to be analyzed using a pre-trained third deep learning model to obtain the perspective clauses of each industry in the securities research report to be analyzed
- the third score of the third score is used to determine the sentiment trend of the opinion clauses of various industries in the securities research report to be analyzed;
- the sentiment trend determination module is used to determine the overall sentiment trend of the securities research report to be analyzed according to the second score and the third score of the opinion clauses of various industries in the securities research report to be analyzed.
- the present application also provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor.
- a deep learning model is used.
- S02 Perform clause processing on the securities research report to be analyzed to obtain each clause in the securities research report to be analyzed;
- S05 use the pre-trained third deep learning model to score the opinion clauses of each industry in the securities research report to be analyzed, to obtain a third score for each industry opinion category in the securities research report to be analyzed, The third score is used to determine the sentiment trend of opinion clauses in various industries in the securities research report to be analyzed;
- S06 Determine the overall sentiment trend of the securities research report to be analyzed based on the second score and the third score of the opinion clauses of various industries in the securities research report to be analyzed.
- the present application also provides a computer-readable storage medium on which a computer program is stored, and when the program is executed by a processor, the following steps of a securities research report analysis method using a deep learning model are implemented:
- S02 Perform clause processing on the securities research report to be analyzed to obtain each clause in the securities research report to be analyzed;
- S05 use the pre-trained third deep learning model to score the opinion clauses of each industry in the securities research report to be analyzed, to obtain a third score for each industry opinion category in the securities research report to be analyzed, The third score is used to determine the sentiment trend of opinion clauses in various industries in the securities research report to be analyzed;
- the securities research report analysis method, device, computer equipment, and readable storage medium using the deep learning model provided in this application can first perform sentence processing on the securities research report to be analyzed, and then use each pre-trained first sentence
- the deep learning model scores to get the first score of each clause, and judges whether the clause is an opinion clause according to the first score, and then uses the pre-trained second deep learning model to judge the judged opinion clause Score to get the second score of each opinion clause, and determine whether each opinion clause is an industry opinion clause according to the second score, and then use the pre-trained third for the judged industry opinion clause
- the deep learning model scores to get the third score of opinion clauses in various industries, and judges the emotional trend of each industry opinion clause according to the third score, and finally according to the second score and third score of each industry opinion clause Determine the overall sentiment trend of the securities research report to be analyzed.
- FIG. 1 is a schematic flowchart of an optional method for analyzing a securities research report using a deep learning model according to an embodiment of the present application
- FIG. 2 is a schematic diagram of an optional program module of a securities research report analysis device using a deep learning model according to an embodiment of the present application
- FIG. 3 is a schematic diagram of another optional program module of a securities research report analysis device using a deep learning model according to an embodiment of the present application;
- FIG. 4 is a schematic diagram of another optional program module of a securities research report analysis device using a deep learning model according to an embodiment of the present application;
- FIG. 5 is a schematic diagram of an optional hardware architecture of a computer device according to an embodiment of the present application.
- FIG. 1 is a schematic diagram of an optional process of a securities research report analysis method using a deep learning model in this application. As shown in FIG. 1, the method may include the following steps:
- a securities research report can be used as a unit to extract its core opinions.
- the format of the securities research report can be judged first. Specifically, it can be determined whether the format of the securities research report is a text type, such as ".txt" format, ".doc” format, etc. If the judgment result is no, that is, the format of the securities research report is not a text type, the format of the securities research report can be converted to a text type, for example, the format of the securities research report is a portable file format (PDF format), then You can use the existing "PDFParser" and other tools to convert the PDF format of the securities research report into a txt format of the securities research report. In this way, the format of the received securities research report can be unified to read the content of the securities research report to be analyzed more conveniently and efficiently.
- PDF format portable file format
- the securities research report to be analyzed can be segmented according to a preset type of symbol, such as comma ",”, period “.”, Period “.”, Dash “-”, square bracket “ [], [] “, Semicolon”; "and other symbols, to deal with the analysis of the securities research report clauses, in order to get the analysis of the securities research report clauses.
- the jieba word segmentation module may be further used to perform word segmentation processing on each clause to obtain each word segment in the securities research report to be analyzed.
- the value corresponding to each word segment in the securities research report to be analyzed can be determined based on a preset dictionary that includes the corresponding conversion relationship between the word segment and the value (in this embodiment: Integer value). Then, according to the determination results of each participle and its corresponding value, each clause composed of multiple participles in the securities research report to be analyzed is converted into a clause in a numeric vector format. In this way, in subsequent steps, the clauses in the numerical vector format can be input into the deep learning model, so that the deep learning model can score each clause and can judge the type of the clause according to the score.
- the training process may include the following steps:
- Step 110 Determine the first data set.
- pre-select a preset number of securities research reports such as 300. For example, you can select a certain percentage of opinions that are more obvious, especially those with more obvious industry views and clearer emotional trends. You can also select a certain percentage of securities research reports.
- the clauses are used to label the first type of tags.
- the first type of tags may include opinion classes and non-opinion classes. That is to say, after being labeled by the first type label, the clauses in the first data set may include opinion clauses and non-opinion clauses.
- Step 120 Perform segmentation processing on each clause labeled with the first type label to obtain each segmentation labeled with the first type label.
- the existing jieba word segmentation module may be used to perform word segmentation processing on each clause labeled with the first type label in the first data set, so as to obtain each word segment labeled with the first type label.
- Step 130 Convert each participle marked with the first type label to a corresponding first value according to a preset rule, and store each participle marked with the first type label and its corresponding first value in a preset including participle and The dictionary of the corresponding conversion relationship of numerical values.
- each participle marked with a label of the first type can be converted into a numeric value one by one according to preset rules (such as random conversion, or the conversion rule of words to numeric values can be preset according to actual needs, etc.) (in this embodiment , May be referred to as a first value), where the first value may be, for example, an integer value, and may store each participle labeled with the label of the first type and the first value in a preset correspondence including the participle and the value The conversion relationship in the dictionary.
- preset rules such as random conversion, or the conversion rule of words to numeric values can be preset according to actual needs, etc.
- each participle marked with the first type label and the first value can be stored in the dictionary in the form of a file or a data table, and the content can be shown in Table 1.
- each clause labeled with the label of the first type is converted into a clause in a numeric vector format to form a first numeric vector clause set.
- each clause composed of a plurality of word segments labeled with the first type label can be converted into a clause in a numeric vector format, and a numeric vector segment can be formed
- the sentence set in this embodiment, may be referred to as a first numerical vector clause set.
- Step 150 Select a first preset number of clauses from the first numerical vector clause set as the first training data.
- the training set used for training and the test set used for testing can be determined first.
- a preset number can be selected from the above first numerical vector clause set (in this
- the first preset number of clauses for example, 90%
- the remaining (for example, 10%) clauses may be used as the first test data.
- Step 160 Train the first training data through the deep learning model to obtain the first deep learning model.
- the deep learning model may be a long-short-term memory network (LSTM, Long Short-Term Memory) machine learning model.
- LSTM Long Short-Term Memory
- some key parameters may be set, such as embedding size (embedding size), Hidden layer size (hidden layer size), batch training size (batch size), full sample number of cycles (num epochs), forgetting rate (dropout), activation function (activation, using sigmoid for variable types), loss function (loss, Use binary-crossentropy) and other parameters for different types of problems.
- the first training data can be input into the LSTM machine learning model for training to obtain the first deep learning model after the training is completed.
- the first test data can also be input into the first deep learning model for testing to obtain the accuracy of the first deep learning model.
- our first The accuracy of the deep learning model on the first test data can reach about 85%.
- the accuracy rate is lower than the predetermined standard (such as 70%)
- a good first-depth model can score clauses more accurately and determine sentence types.
- each clause in the securities research report to be analyzed can be scored using the above-mentioned first deep learning model to obtain a first score for each clause in the securities research report to be analyzed, and based on the first score Analyze the opinion clauses in each clause in the Securities Research Report.
- the first score may be a score considered by the model, and the score may be regarded as a probability (between 0 and 1). For example, if the model thinks that the clause that is more biased towards the opinion category, the score is closer to 1 (such as 0.75), the model believes that the clauses that are more biased towards non-opinion categories, the score is closer to 0 (such as 0.22).
- the first score can be compared with the preset first score threshold, and the opinion clause can be judged according to the comparison result, for example, the sentence with the first score greater than the first score threshold can be judged
- the first score threshold can be set to 0.5, that is to say, clauses with a first score not greater than 0.5 can be determined as non-opinion clauses, and no longer participate in subsequent steps; the first score is greater than Clauses of 0.5 can be determined as opinion clauses, and you can continue to participate in subsequent steps.
- the first deep learning model pre-trained is used to analyze the types of clauses (sentence clauses or non-opinion clauses) in an intelligent scoring manner. On the one hand, it can greatly save manpower and improve the analysis efficiency; On the other hand, the accuracy of the analysis results can be improved.
- the training process may include the following steps:
- Step 210 Determine the second data set.
- pre-select a preset number (such as 300) of securities research reports For example, a certain percentage of opinions can be selected, especially those of the industry with obvious opinions and clear emotional trends. A certain percentage of opinions can also be selected. Fuzzy research, especially the securities research report with vague industry views and unclear emotional trends. Then, the summary part of the securities research report can be sentenced to obtain multiple clauses, and then each clause can be manually
- the first type of labeling is performed, and the first type of label may include an opinion class and a non-opinion class. That is to say, after being labeled by the first type label, the clauses in the second data set may include opinion clauses and non-opinion clauses.
- each opinion class clause with a second type label, which may include an industry opinion class and a non-industry opinion class. That is to say, after being labeled by the second type of label, the clauses in the second data set may include opinion clauses and non-opinion clauses, and the opinion clauses also include industry opinion clauses and non-industry opinions Category clause.
- Step 220 Perform segmentation processing on each clause labeled with the second type label to obtain each segmentation labeled with the second type label.
- the clauses labeled with the second type label in the second data set are subjected to word segmentation processing, for example, the current
- the jieba word segmentation module performs word segmentation processing on the above clauses to obtain each word segment marked with a second type label.
- Step 230 Convert each participle marked with the second type label to a corresponding second value according to a preset rule, and store each participle marked with the second type label and its corresponding second value in the preset including participle and The dictionary of the corresponding conversion relationship of numerical values.
- each participle marked with a second type label can be converted into a numeric value one by one according to preset rules (such as random conversion, or a word-to-value conversion rule can be preset according to actual needs) (in this embodiment , May be referred to as a second value), where the first value may be, for example, an integer value, and each participle labeled with a label of the second type and the second value may be stored in a preset correspondence including the participle and the value The conversion relationship in the dictionary.
- preset rules such as random conversion, or a word-to-value conversion rule can be preset according to actual needs
- each participle marked with the second type label and the second value can be stored in the dictionary in the form of a file or a data table, and the content can be as shown in Table 1 above.
- each clause labeled with the label of the second type is converted into a clause in a numeric vector format to form a second numeric vector clause set.
- each clause composed of a plurality of word segments labeled with a second type label can be converted into a clause in a numeric vector format, and a numeric vector segment can be formed
- the sentence set in this embodiment, may be referred to as a second numerical vector clause set.
- Step 250 Select a second preset number of clauses from the second numerical vector clause set as the second training data.
- the training set for training and the test set for testing can be determined first.
- a preset number can be selected from the above second numerical vector clause set (in this In an embodiment, it may be called a second preset number, such as 90% of the clauses, as second training data, and the remaining (eg, 10%) clauses may be used as the second test data.
- Step 260 Train the second training data through the deep learning model to obtain the second deep learning model.
- the deep learning model may be a long-short-term memory network (LSTM, Long Short-Term Memory) machine learning model.
- LSTM Long Short-Term Memory
- some key parameters may be set, such as embedding size (embedding size), Hidden layer size (hidden layer size), batch training size (batch size), full sample number of cycles (num epochs), forgetting rate (dropout), activation function (activation, using sigmoid for variable types), loss function (loss, Use binary-crossentropy) and other parameters for different types of problems.
- the second training data can be input to the LSTM machine learning model for training to obtain the second deep learning model after the training is completed.
- the second test data may also be input into the second deep learning model for testing to obtain the accuracy rate of the second deep learning model.
- the accuracy of our second deep learning on the second test data can reach about 85%.
- the accuracy rate is lower than the predetermined standard (such as 70%)
- a good second-depth model can score clauses more accurately and determine sentence types.
- the opinion clauses in the securities research report to be analyzed determined in S03 can be scored using the above-mentioned second deep learning model to obtain a second score for each opinion category clause in the securities research report to be analyzed , And can determine the industry opinion class clause from the opinion class clauses in the securities research report to be analyzed based on the second score.
- the second score may be a score considered by the model, and the score may be regarded as a probability (between 0 and 1). For example, if the model considers a clause that is more biased towards industry opinions, the score is closer to At 1 (such as 0.85), the model believes that the clauses that are more biased towards non-industry opinion categories, the score is closer to 0 (such as 0.3).
- the second score can be compared with the preset second score threshold, and the opinion clause can be judged according to the comparison result, for example, the sentence with the second score greater than the second score threshold can be judged Clauses for industry opinion categories, clauses with a second score not greater than the second score threshold are determined as non-industry opinion clauses.
- the second score threshold can be set to 0.5, that is to say, clauses with a second score not greater than 0.5 can be determined as non-industry opinion clauses, and no longer participate in subsequent steps; second score Clauses greater than 0.5 can be determined as industry opinion clauses and can continue to participate in subsequent steps.
- S05 use the pre-trained third deep learning model to score the opinion clauses of each industry in the securities research report to be analyzed, to obtain a third score for each industry opinion category in the securities research report to be analyzed, The third score is used to determine the sentiment trend of each industry opinion clause in the securities research report to be analyzed.
- the training process may include the following steps:
- Step 310 Determine the third data set.
- a preset number such as 300
- Each clause is used to label the first type of label, and the first type of label may include opinion and non-opinion categories. That is to say, after being labeled by the first type label, the clauses in the first data set may include opinion clauses and non-opinion clauses.
- each opinion class clause with a second type label, which may include an industry opinion class and a non-industry opinion class. That is to say, after being labeled by the second type of label, the clauses in the second data set may include opinion clauses and non-opinion clauses, and the opinion clauses also include industry opinion clauses and non-industry opinion categories Clause.
- each industry opinion category clause with a third type label, which may include a bullish category and a bearish category. That is to say, after being labeled by the third type of label, the clauses in the third data set may include opinion clauses and non-opinion clauses, and the opinion clauses also include industry opinion clauses and non-industry opinion categories Clause, and the industry opinion clause also includes bullish industry opinion clause and bearish industry opinion clause.
- Step 320 Perform segmentation processing on each clause labeled with the third type label to obtain each segmentation labeled with the third type label.
- the third data set is marked with each clause of the third type (that is, the bullish industry opinion class clause and the bearish industry opinion class clause in the third data set) for word segmentation processing.
- the existing jieba word segmentation module may be used to perform word segmentation on the above clauses to obtain each word segment marked with a third type label.
- Step 330 Convert each participle labeled with the third type label to a corresponding third value according to a preset rule, and store each participle labeled with the third type label and its corresponding third value in the preset including participle and The dictionary of the corresponding conversion relationship of numerical values.
- each participle marked with the third type label can be converted into a numeric value one by one according to preset rules (such as random conversion, or a word-to-value conversion rule can be preset according to actual needs) (in this embodiment , May be referred to as a third value), where the third value may be, for example, an integer value, and each participle labeled with a third type label and the third value may be stored in a preset correspondence including the participle and the value The conversion relationship in the dictionary.
- preset rules such as random conversion, or a word-to-value conversion rule can be preset according to actual needs
- each participle marked with the third type label and the third value can be stored in the dictionary in the form of a file or a data table, and the content can be as shown in Table 1 above.
- each clause labeled with a third type label is converted into a clause in a numeric vector format to form a third numeric vector clause set.
- each clause composed of a plurality of word segments labeled with a third type label can be converted into a clause in a numeric vector format and form a numeric vector segment
- the sentence set in this embodiment, may be referred to as a third numerical vector clause set.
- Step 350 Select a third preset number of clauses from the third numerical vector clause set as the third training data.
- the training set for training and the test set for testing can be determined before the deep learning model training.
- a preset number can be selected from the third numerical vector clause set (in this
- the third preset number of clauses may be referred to as the third training data, and the remaining (eg, 10%) clauses may be used as the third test data.
- Step 360 Train the third training data through the deep learning model to obtain the third deep learning model.
- the deep learning model may be a long-short-term memory network (LSTM, Long Short-Term Memory) machine learning model.
- LSTM Long Short-Term Memory
- some key parameters may be set, such as embedding size (embedding size), Hidden layer size (hidden layer size), batch training size (batch size), full sample number of cycles (num epochs), forgetting rate (dropout), activation function (activation, using sigmoid for variable types), loss function (loss, Use binary-crossentropy) and other parameters for different types of problems.
- the third training data can be input to the LSTM machine learning model for training, so as to obtain the third deep learning model after training.
- the third test data can also be input into the third deep learning model for testing to obtain the accuracy rate of the third deep learning model.
- the accuracy of our third deep learning on the third test data can reach about 85%.
- the accuracy rate is lower than the predetermined standard (such as 70%)
- a good third-depth model can score clauses more accurately and determine sentence types according to needs.
- the opinion clauses of each industry in the securities research report to be analyzed determined in S04 can be scored using the above-mentioned third deep learning model to obtain the first sentence for each industry opinion category in the securities research report to be analyzed.
- Three scores, and based on the third score, the sentiment trend of the industry opinion clause in the securities research report to be analyzed can be determined, that is, the bullish industry opinion clause or the bearish industry opinion clause.
- the third score may be the score considered by the model, and the score may be regarded as a probability (between 0 and 1). For example, if the model believes that it is more biased towards the opinions of bullish industries, the score It is closer to 1 (such as 0.85), and the model believes that the clauses that are more biased towards the view category of the bearish industry, the score is closer to 0 (such as 0.3).
- the third score can be compared with a preset third score threshold, and the degree of bullish or bearish emotion can be further judged according to the comparison result, for example, a score with a third score greater than the third score threshold Sentences are judged as industry opinion clauses with high bullishness or low bearish sentiment, and clauses with a third score not greater than the third score threshold are judged as industry opinion clauses with low bullish or high bearish sentiment.
- the second score threshold can be set between 0.5-0.6, preferably set to 0.6, that is to say, clauses with a third score greater than 0.6 can be determined to have higher bullish sentiment or bearish sentiment Low industry opinion clauses; clauses with a third score not greater than 0.6 can be determined as industry opinion clauses with low bullish sentiment or high bearish sentiment.
- a pre-trained third deep learning model is used to obtain a third category score for determining the sentiment trend of opinion clauses in various industries in a smart scoring manner, which is used to analyze the opinion clauses in various industries.
- the opinion clause of the bullish industry or the opinion clause of the bearish industry, and the degree of bullish bearish sentiment on the one hand, can greatly save manpower and improve the efficiency of analysis; on the other hand, can improve the accuracy of the analysis results.
- S06 Determine the overall sentiment trend of the securities research report to be analyzed based on the second score and the third score of the opinion clauses of various industries in the securities research report to be analyzed.
- a second score is obtained for each industry opinion clause in the securities research report to be analyzed
- a third score is obtained for each industry opinion category clause in the securities research report to be analyzed.
- the product of the second score and the third score of each industry opinion clause in the securities research report to be analyzed can be calculated as the first product value of each industry opinion category clause; then all The sum of the first product value of the industry opinion class clause is used as the first sum value; then the sum of the second scores of all industry opinion class clauses is calculated as the second sum value; next, the first sum value is divided by The second sum value to obtain the overall sentiment score of the securities research report to be analyzed (which can also be understood as the rise and fall score). Since the second score and the third score are between 0 and 1, the calculated securities to be analyzed The overall sentiment score of the research report is also between 0 and 1.
- the preset score threshold can be set to 0.5-0.6 according to actual experience, preferably set to 0.6.
- the overall sentiment trend of the securities research report to be analyzed can be determined to be bullish, and if the overall sentiment score is not higher than the preset score threshold, the securities research report to be analyzed can be determined The overall sentiment trend is bearish.
- the second score (which can be used to determine whether it is a more obvious industry viewpoint) and the third score (which can be used to determine each Industry sentiment trends) through comprehensive calculation, the overall sentiment score of the securities research report to be analyzed is obtained, and the overall sentiment score of the securities research report to be analyzed and the predetermined index are compared to obtain the overall securities research report to be analyzed Emotional trend, which can get more objective and accurate results of emotional trend analysis.
- the core opinions in the securities research report can also be extracted.
- the opinion clauses of various industries in the securities research report to be analyzed can be divided into two parts according to their corresponding third scores, including the part with the third score greater than 0.6 and the part with the third score not greater than 0.6 , And both parts can be sorted in positive order (scores from large to small).
- the industry opinion category clauses consistent with the overall sentiment trend of the securities research report can be determined according to the third score of the industry category category clauses.
- the industry opinion category clause with the third score greater than 0.6 can be determined as the industry opinion category that is consistent with the overall sentiment trend of the securities research report to be analyzed Sentence; if the overall sentiment trend of the securities research report to be analyzed is bearish, the industry opinion clause with the third score not greater than 0.6 can be determined as the industry opinion category clause consistent with the overall sentiment trend of the securities research report to be analyzed .
- the preset number of industry opinion categories with the highest third score can be selected Clause as the core opinion of the securities research report to be analyzed and output; if the overall sentiment trend of the securities research report to be analyzed is bearish, the preset number of industry opinion categories with the lowest third score can be selected as the securities to be analyzed
- the core opinions of the research report are output.
- the preset number can be set according to actual needs. In this embodiment, for example, it can be set to two.
- the securities research report to be analyzed can be first sentence-processed, and then each clause can be scored using a pre-trained first deep learning model to obtain the first score of each clause, and Determine whether the clause is an opinion clause according to the first score, and then use the pre-trained second deep learning model to score the judged opinion clause to obtain the second score of each opinion clause, and according to the first score Two scores determine whether each opinion clause is an industry opinion clause. Next, the judged industry opinion clause is scored using a pre-trained third deep learning model to obtain the third opinion clause of each industry.
- the above process of intelligently scoring and analyzing using the deep learning model can not only save a lot of manpower, but also improve the analysis efficiency and accuracy of the analysis results.
- FIGS. 2 to 4 show the use depth An optional structural block diagram of a securities research report analysis device of a learning model.
- the securities research report analysis device using a deep learning model is divided into one or more program modules, and one or more program modules are stored in a storage medium, and It is executed by one or more processors to complete this application.
- the program module referred to in this application refers to a series of computer program instruction segments that can perform specific functions. It is more suitable than the program itself to describe the execution process of a securities research report analysis device using a deep learning model in a storage medium. The following description will specifically introduce The function of each program module in this embodiment.
- the securities research report analysis device 20 using the deep learning model may include:
- the receiving module 21 can receive the input securities research report to be analyzed
- Sentence module 22 can be used to perform sentence processing on the securities research report to be analyzed to obtain each clause in the securities research report to be analyzed;
- the first scoring module 23 can be used to score each clause in the securities research report to be analyzed using a pre-trained first deep learning model to obtain a first score for each clause in the securities research report to be analyzed, Based on the first score, the opinion clause is determined from the clauses in the securities research report to be analyzed;
- the second scoring module 24 can be used to score each viewpoint class in the securities research report to be analyzed using a pre-trained second deep learning model to obtain a score for each opinion class in the securities research report to be analyzed The second score, and based on the second score, determine the industry opinion clause from each opinion clause in the securities research report to be analyzed;
- the third scoring module 25 can be used to score the perspective clauses of various industries in the securities research report to be analyzed using a pre-trained third deep learning model to obtain the perspective categories for each industry in the securities research report to be analyzed.
- the third score of the sentence, the third score is used to determine the sentiment trend of each industry opinion clause in the securities research report to be analyzed;
- the sentiment trend determination module 26 can be used to determine the overall sentiment trend of the securities research report to be analyzed according to the second score and the third score of the opinion clauses of various industries in the securities research report to be analyzed.
- clause module 22 may specifically include:
- Sentence unit 221 can be used to perform sentence processing on the securities research report to be analyzed according to preset symbols, to obtain each clause in the securities research report to be analyzed;
- the word segmentation unit 222 may be used to perform word segmentation processing on each clause in the securities research report to be analyzed to obtain each word segment in the securities research report to be analyzed;
- the numerical value determining unit 223 may be used to determine the numerical value corresponding to each partial word in the securities research report to be analyzed based on a preset dictionary including the corresponding conversion relationship between the partial word and the numerical value;
- the conversion unit 224 can be used to convert each clause in the securities research report to be analyzed into a clause in a numeric vector format according to the determination result.
- the training process of the first deep learning model may include the following steps:
- Step 110 a first data set is determined, the first data set includes a plurality of clauses obtained by pre-sentencing the abstract section of the securities research report of a preset number of articles, each of which has a pre-marked
- the first type label includes opinion class and non-opinion class;
- Step 120 Perform segmentation processing on each clause labeled with the first type label to obtain each segmentation labeled with the first type label;
- Step 130 Convert each participle marked with the first type label to a corresponding first value according to a preset rule, and store each participle marked with the first type label and its corresponding first value in a preset including participle and The dictionary of the corresponding conversion relationship of numerical values;
- Step 140 According to the dictionary, convert each clause labeled with the first type label into a numeric vector format clause to form a first numeric vector clause set;
- Step 150 Select a first preset number of clauses from the first numerical vector clause set as the first training data
- Step 160 Train the first training data through the deep learning model to obtain the first deep learning model.
- the training process of the second deep learning model may include the following steps:
- Step 210 Determine a second data set, where the second data set includes a plurality of clauses obtained by pre-sentencing the abstract section of the securities research report of a preset number of articles, each of which has a pre-marked
- the first type of label, the first type of label includes opinion class and non-opinion class, and the clause marked with the opinion class tag also has the pre-labeled second type tag, and the second type tag includes industry opinion class and non-industry opinion class;
- Step 220 Perform segmentation processing on each clause labeled with the second type label to obtain each segmentation labeled with the second type label;
- Step 230 Convert each participle marked with the second type label to a corresponding second value according to a preset rule, and store each participle marked with the second type label and its corresponding second value in the preset including participle and The dictionary of the corresponding conversion relationship of numerical values;
- Step 240 According to the dictionary, convert each clause labeled with the second type label into a clause in a numeric vector format to form a second numeric vector clause set;
- Step 250 Select a second preset number of clauses from the second numerical vector clause set as the second training data
- Step 260 Train the second training data through the deep learning model to obtain the second deep learning model.
- the training process of the third deep learning model may include the following steps:
- Step 310 Determine a third data set, where the third data set includes a plurality of clauses obtained by pre-sentencing the abstract section of the securities research report of a preset number of articles, each of which has a pre-marked
- the first type of label includes opinion class and non-opinion class
- the clause marked with the opinion class tag also has the pre-labeled second type tag
- the second type tag includes industry opinion class and non-industry opinion class
- the clause marked with the industry opinion label also has the pre-labeled third type label
- the third type label includes the bullish category and the bearish category
- Step 320 Perform segmentation processing on each clause labeled with the third type label to obtain each segmentation labeled with the third type label;
- Step 330 Convert each participle labeled with the third type label to a corresponding third value according to a preset rule, and store each participle labeled with the third type label and its corresponding third value in the preset including participle and The dictionary of the corresponding conversion relationship of numerical values;
- each clause labeled with the third type label is converted into a clause in a numeric vector format to form a third numeric vector clause set;
- Step 350 select a third preset number of clauses from the third numerical vector clause set as the third training data
- Step 360 Train the third training data through the deep learning model to obtain the third deep learning model.
- the deep learning model is a machine learning model for long- and short-term memory networks.
- the emotional trend determination module 26 may specifically include:
- the first calculation unit 261 is used to calculate the product of the second score and the third score of each industry opinion category clause in the securities research report to be analyzed as the first product value of each industry opinion category clause;
- the second calculation unit 262 is used to calculate the sum of the first product values of all industry opinion clauses as the first sum value
- the third calculation unit 263 is used to calculate the sum of the second scores of all industry opinion clauses as the second sum value
- the fourth calculation unit 264 is used to divide the first sum value by the second sum value to obtain the overall sentiment score of the securities research report to be analyzed;
- the judging unit 265 is used to judge whether the overall sentiment score of the securities research report to be analyzed is higher than a preset score threshold;
- the sentiment trend determination unit 266 can be used to determine the overall sentiment trend of the securities research report to be analyzed as bullish when the judgment result of the judgment unit 265 is yes, and determine the securities research report to be analyzed when the judgment result of the judgment unit 265 is no The overall emotional trend is bearish.
- analysis device of the securities research report may further include:
- the industry opinion clause determination module is used to determine the industry opinion category clause consistent with the overall sentiment trend of the securities research report according to the third score of each industry opinion category in the securities research report to be analyzed;
- the selection module is used to select a predetermined number of industry opinion clauses with the highest or lowest third score from the determination results, and output as the core opinion of the securities research report.
- the securities research report to be analyzed can be sentence-processed first, and then each clause can be scored using a pre-trained first deep learning model to obtain the first score of each clause, and Determine whether the clause is an opinion clause according to the first score, and then use the pre-trained second deep learning model to score the judged opinion clause to obtain the second score of each opinion clause, and according to the first score Two scores determine whether each opinion clause is an industry opinion clause. Next, the judged industry opinion clause is scored using a pre-trained third deep learning model to obtain the third opinion clause of each industry.
- the above process of intelligently scoring and analyzing using the deep learning model can not only save a lot of manpower, but also improve the analysis efficiency and accuracy of the analysis results.
- This embodiment also provides a computer device, such as a smartphone, a tablet computer, a notebook computer, a desktop computer, a rack server, a blade server, a tower server, or a rack server (including an independent server, or A server cluster composed of multiple servers), etc.
- the computer device 50 of this embodiment includes at least but not limited to: a memory 51 and a processor 52 that can be communicatively connected to each other through a system bus, as shown in FIG. 5.
- FIG. 5 only shows the computer device 50 having the components 51-52, but it should be understood that it is not required to implement all the components shown, and more or fewer components may be implemented instead.
- the memory 51 (ie, readable storage medium) includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), Read only memory (ROM), electrically erasable programmable read only memory (EEPROM), programmable read only memory (PROM), magnetic memory, magnetic disk, optical disk, etc.
- the memory 51 may be an internal storage unit of the computer device 50, such as a hard disk or memory of the computer device 50.
- the memory 51 may also be an external storage device of the computer device 50, for example, a plug-in hard disk equipped on the computer device 50, a smart memory card (Smart Media, Card, SMC), and a secure digital (Secure Digital, SD) card, flash card (Flash Card), etc.
- the memory 51 may also include both the internal storage unit of the computer device 50 and its external storage device.
- the memory 51 is generally used to store the operating system and various application software installed in the computer device 50, such as the program code of the securities research report analysis apparatus using the deep learning model in Embodiment 2.
- the memory 51 can also be used to temporarily store various types of data that have been output or are to be output.
- the processor 52 may be a central processing unit (CPU), controller, microcontroller, microprocessor, or other data processing chip in some embodiments.
- the processor 52 is generally used to control the overall operation of the computer device 50.
- the processor 52 is used to run program codes or process data stored in the memory 51, such as a securities research report analysis device using a deep learning model.
- This embodiment also provides a computer-readable storage medium, such as flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), only Read memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, server, App store, etc., on which computer programs are stored, When the program is executed by the processor, the corresponding function is realized.
- the computer-readable storage medium of this embodiment is used for a securities research report analysis apparatus using a deep learning model, and when executed by a processor, implements the securities research report analysis method using a deep learning model of Embodiment 1.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Machine Translation (AREA)
Abstract
一种利用深度学习模型的证券研报分析方法,包括:接收待分析证券研报(S01);对该证券研报进行分句得到各分句(S02);将各分句采用第一深度学习模型进行评分得到各分句的第一评分并基于第一评分确定观点类分句(S03);将各观点类分句采用第二深度学习模型进行评分得到各观点类分句的第二评分并基于第二评分确定行业观点类分句(S04);将各行业观点类分句采用第三深度学习模型进行评分得到各行业观点类分句的第三评分(S05);根据各行业观点类分句的第二、第三评分确定该证券研报的整体情感趋势(S06)。利用深度学习模型进行智能化评分的方式实现对证券研报的分析,可解决针对证券研报分析方案的效率及准确率较低的问题,提高对证券研报分析的效率及准确率。
Description
相关申请的交叉引用
本申请申明享有2018年10月22日递交的申请号为CN2018112287617、名称为“利用深度学习模型的证券研报分析方法及装置”的中国专利申请的优先权,该中国专利申请的整体内容以参考的方式结合在本申请中。
本申请涉及计算机技术领域,具体涉及一种利用深度学习模型的证券研报分析方法及装置。
证券研究报告,也可简称为证券研报,是指相关研究人员(比如证券公司内的研究人员等)对证券及相关产品的价值、或者影响其市场价格的因素进行分析,所作出的研究报告。
对证券研报进行分析,可以及时了解证券研报中关于行业、政策、投资可行性等方面情况,但目前主要还是通过人工方式对证券研报进行阅读分析,以获得作者的核心观点等有用信息,此种方式需要耗费大量人力,效率及准确率都较低。此外,也有通过情感词典进行分析的方案,比如通过与情感词典中关键语库进行比对,以对证券研报中的语句进行情感分析,此种方式相对割裂了语句的前后关系,很难对语句进行比较完善准确的分析,尤其在针对具有因果、转折类等关联关系的语句时,分析的准确率更为不理想。
相关技术中针对证券研报分析方案的效率及准确率较低的问题,目前尚未提出有效的解决方案。
发明内容
本申请的目的在于提供一种利用深度学习模型的证券研报分析方法、装置、计算机设备及可读存储介质,进而在一定程度上克服上述现有技术中存在的问题,可提高对证券研报分析的效率及准确率。
本申请是通过下述技术方案来解决上述技术问题:
根据本申请的一个方面,提供了一种利用深度学习模型的证券研报分析方法,包括如下步骤:
S01,接收输入的待分析证券研报;
S02,对所述待分析证券研报进行分句处理,得到待分析证券研报中的各分句;
S03,将待分析证券研报中的各分句采用预先训练好的第一深度学习模型进行评分,以得到针对待分析证券研报中的各分句的第一评分,并基于第一评分从待分析证券研报中的各分句中确定观点类分句;
S04,将待分析证券研报中的各观点类分句采用预先训练好的第二深度学习模型进行评分,以得到针对待分析证券研报中的各观点类分句的第二评分,并基于第二评分从待分析证券研报中的各观点类分句中确定行业观点类分句;
S05,将待分析证券研报中的各行业观点类分句采用预先训练好的第三深度学习模型进行评分,以得到针对待分析证券研报中的各行业观点类分句的第三评分,所述第三评分用于确定待分析证券研报中的各行业观点类分句的情感趋势;
S06,根据待分析证券研报中的各行业观点类分句的第二评分及第三评分确定待分析证券研报的整体情感趋势。
为了实现上述目的,本申请还提供一种利用深度学习模型的证券研报分析装置,包括:
接收模块,用于接收输入的待分析证券研报;
分句模块,用于对待分析证券研报进行分句处理,得到待分析证券研报中的各分句;
第一评分模块,用于将待分析证券研报中的各分句采用预先训练好的第一深度学习模型进行评分,以得到针对待分析证券研报中的各分句的第一评分,并基于第一评分从待分析证券研报中的各分句中确定观点类分句;
第二评分模块,用于将待分析证券研报中的各观点类分句采用预先训练好的第二深度学习模型进行评分,以得到针对待分析证券研报中的各观点类分句的第二评分,并基于第二评分从待分析证券研报中的各观点类分句中确定行业观点类分句;
第三评分模块,用于将待分析证券研报中的各行业观点类分句采用预先训练好的第三深度学习模型进行评分,以得到针对待分析证券研报中的各行业观点类分句的第三评分,所述第三评分用于确定待分析证券研报中的各行业观点类分句的情感趋势;
情感趋势确定模块,用于根据待分析证券研报中的各行业观点类分句的第二评分及第三评分确定待分析证券研报的整体情感趋势。
为了实现上述目的,本申请还提供一种计算机设备,包括存储器、处理器以及存储在存储器上并可在处理器上运行的计算机程序,所述处理器执行所述程序时实现利用深度学习模型的证券研报分析方法的以下步骤:
S01,接收输入的待分析证券研报;
S02,对所述待分析证券研报进行分句处理,得到待分析证券研报中的各分句;
S03,将待分析证券研报中的各分句采用预先训练好的第一深度学习模型进行评分,以得到针对待分析证券研报中的各分句的第一评分,并基于第一评分从待分析证券研报中的各分句中确定观点类分句;
S04,将待分析证券研报中的各观点类分句采用预先训练好的第二深度学习模型进行评分,以得到针对待分析证券研报中的各观点类分句的第二评分,并基于第二评分从待分析证券研报中的各观点类分句中确定行业观点类分句;
S05,将待分析证券研报中的各行业观点类分句采用预先训练好的第三深度学习模型进行评分,以得到针对待分析证券研报中的各行业观点类分句的第三评分,所述第三评分用于确定待分析证券研报中的各行业观点类分句的情感趋势;
S06,根据待分析证券研报中的各行业观点类分句的第二评分及第三评分确定待分析证券研报的整体情感趋势。
为了实现上述目的,本申请还提供一种计算机可读存储介质,其上存储有计算机程序,所述程序被处理器执行时实现利用深度学习模型的证券研报分析方法的以下步骤:
S01,接收输入的待分析证券研报;
S02,对所述待分析证券研报进行分句处理,得到待分析证券研报中的各分句;
S03,将待分析证券研报中的各分句采用预先训练好的第一深度学习模型进行评分,以得到针对待分析证券研报中的各分句的第一评分,并基于第一评分从待分析证券研报中的各分句中确定观点类分句;
S04,将待分析证券研报中的各观点类分句采用预先训练好的第二深度学习模型进行评分,以得到针对待分析证券研报中的各观点类分句的第二评分,并基于第二评分从待分析证券研报中的各观点类分句中确定行业观点类分句;
S05,将待分析证券研报中的各行业观点类分句采用预先训练好的第三深度学习模型进行评分,以得到针对待分析证券研报中的各行业观点类分句的第三评分,所述第三评分用于确定待分析证券研报中的各行业观点类分句的情感趋势;
S06,根据待分析证券研报中的各行业观点类分句的第二评分及第三评分确定待分析证 券研报的整体情感趋势。
本申请提供的利用深度学习模型的证券研报分析方法、装置、计算机设备及可读存储介质,可先将待分析证券研报进行分句处理,再将各分句采用预先训练好的第一深度学习模型进行评分以得到各分句的第一评分,并根据第一评分判断分句是否为观点类分句,然后将判断出的观点类分句采用预先训练好的第二深度学习模型进行评分以得到各观点类分句的第二评分,并根据第二评分判断各观点类分句是否为行业观点类分句,接下来将判断出的行业观点类分句采用预先训练好的第三深度学习模型进行评分以得到各行业观点类分句的第三评分,并根据第三评分判断各行业观点类分句的情感趋势,最后根据各行业观点类分句的第二评分及第三评分确定该待分析证券研报的整体情感趋势。通过上述方案,可基于预先训练好的深度学习模型通过评分的方式挑选出观点类分句、行业观点类分句以及确定行业观点类分句的情感趋势,并最终通过客观的评分确定该证券研报的整体情感趋势,由此,上述利用深度学习模型智能评分并分析的过程,不但可大大节省人力,而且可提高分析效率及分析结果的准确率。
图1是根据本申请实施例的利用深度学习模型的证券研报分析方法的一种可选的流程示意图;
图2是根据本申请实施例的利用深度学习模型的证券研报分析装置的一种可选的程序模块示意图;
图3是根据本申请实施例的利用深度学习模型的证券研报分析装置的另一种可选的程序模块示意图;
图4是根据本申请实施例的利用深度学习模型的证券研报分析装置的又一种可选的程序模块示意图;
图5是根据本申请实施例的计算机设备的一种可选的硬件架构示意图。
为了使本申请的目的、技术方案及优点更加清楚明白,以下结合附图及实施例,对本申请进行进一步详细说明。应当理解,此处所描述的具体实施例仅用以解释本申请,并不用于限定本申请。基于本申请中的实施例,本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例,都属于本申请保护的范围。
实施例1
下面结合附图对本申请提供的利用深度学习模型的证券研报分析方法进行说明。
图1为本申请利用深度学习模型的证券研报分析方法的一种可选的流程示意图,如图1所示,该方法可以包括以下步骤:
S01,接收输入的待分析证券研报。
在本实施例中,比如可以一篇证券研报为单位进行其核心观点的提取,当接收到一篇待分析的证券研报后,可先对该篇证券研报的格式进行判断。具体的,可判断该篇证券研报的格式是否为文本类型,比如“.txt”格式、“.doc”格式等。若判断结果为否,也就是说该证券研报的格式不是文本类型,则可将该证券研报的格式转换为文本类型,比如该证券研报的格式为便携文件格式(PDF格式),则可使用现有的“PDFParser”等工具将PDF格式的证券研报转换txt格式的证券研报。以此,可统一接收到的待分析证券研报的格式,以更方便且更高效的对待分析证券研报的内容进行读取。
S02,对待分析证券研报进行分句处理,得到待分析证券研报中的各分句。
在本实施例中,可根据预置类型的符号对待分析证券研报进行分句处理,比如可根据 逗号“,”、句号“。”、点号“.”、破折号“—”、中括号“[]、【】”、分号“;”等符号,对待分析证券研报进行分句处理,以得到待分析证券研报中的各分句。
在得到待分析证券研报中的各分句后,可进一步采用jieba分词模块对各分句进行分词处理,以得到待分析证券研报中的各分词。
在得到待分析证券研报中的各分词后,可基于在预置的包括分词与数值的对应转换关系的字典,确定待分析证券研报中各分词所对应的数值(在本实施例中为整数型数值)。然后,再根据各分词与其对应数值的确定结果,将待分析证券研报中每个由多个分词组成的分句转换成数值向量格式的分句。以此,可在后续步骤中,将数值向量格式的分句输入深度学习模型中,以便深度学习模型对各分句进行评分并可根据评分对分句的类型进行判断。
S03,将待分析证券研报中的各分句采用预先训练好的第一深度学习模型进行评分,以得到针对待分析证券研报中的各分句的第一评分,并基于第一评分从待分析证券研报中的各分句中确定观点类分句。
首先,先对第一深度学习模型的训练过程进行说明,该训练过程可包括如下步骤:
步骤110,确定第一数据集。
比如,预先挑选预置篇数(比如300篇)的证券研报,比如可挑选一定比例的观点比较明显、尤其是行业观点比较明显、情感趋势比较清晰的证券研报,还可挑选一定比例的观点比较模糊、尤其是行业观点比较模糊、情感趋势不太清晰的证券研报等,然后可对这些证券研报的摘要部分进行分句处理以得到多个分句,然后以人工方式对每个分句进行第一类型标签的标注,该第一类型标签可包括观点类和非观点类。也就是说,经第一类型标签标注后,在第一数据集中的分句可包括观点类分句和非观点类分句。
步骤120,将标注有第一类型标签的各分句进行分词处理,得到标注有第一类型标签的各分词。比如,可使用现有jieba分词模块对第一数据集中标注有第一类型标签的各分句进行分词处理,以得到标注有第一类型标签的各分词。
步骤130,根据预置规则将标注有第一类型标签的各分词转换为对应的第一数值,并将标注有第一类型标签的各分词与其对应的第一数值存储于预置的包括分词与数值的对应转换关系的字典中。
具体的,可根据预置规则(比如随机转换,或者可根据实际需求预先设定词语到数值的转换规则等)将标注有第一类型标签的各分词一一转换为数值(在本实施例中,可称为第一数值),其中,第一数值比如可为integer整数型的数值,并可将标注有第一类型标签的各分词与第一数值存储于预置的包括分词与数值的对应转换关系的字典中。
在本实施例中,可将标注有第一类型标签的各分词与第一数值的对应转换关系以文件或者数据表的形式保存于字典中,内容可如表1所示。
表1分词与数值的对应转换关系
| 分词 | 数值 |
| 分词1 | 2 |
| 分词2 | 175 |
| 分词3 | 50 |
| ... | ... |
步骤140,根据字典,将标注有第一类型标签的各分句转换成数值向量格式的分句,形成第一数值向量分句集合。
也即,可根据上述包括分词与数值的对应转换关系的字典,将每个标注有第一类型标签的由多个分词组成的分句,转换成数值向量格式的分句,并形成数值向量分句集合,在本实施例中,可称为第一数值向量分句集合。
步骤150,从第一数值向量分句集合中选取第一预置数量的分句作为第一训练数据。
通常在进行深度学习模型训练之前,可先确定用于训练的训练集和用于测试的测试集,在本实施例中,可从上述第一数值向量分句集合中选取预置数量(在本实施例中,可称为 第一预置数量,比如占比90%)的分句作为第一训练数据,剩余(比如占比10%)的分句则可作为第一测试数据。
步骤160,将第一训练数据经深度学习模型进行训练,以获得第一深度学习模型。
在本实施例中,该深度学习模型可为长短期记忆网络(LSTM,Long Short-Term Memory)机器学习模型,在训练之前,可对一些关键参数进行设置,比如嵌入层大小(embedding size)、隐藏层大小(hidden layer size)、批次训练大小(batch size)、全样本循环次数(num epochs)、遗忘率(dropout)、激活函数(activation,分类型变量使用sigmoid)、损失函数(loss,分类型问题使用binary-crossentropy)等参数。然后,可将第一训练数据输入LSTM机器学习模型进行训练,以在训练完毕后获得第一深度学习模型。
此外,在训练完毕获得第一深度学习模型后,还可将第一测试数据输入第一深度学习模型进行测试,以得到该第一深度学习模型的准确率,经多次试验,我们的第一深度学习模型在第一测试数据上的准确率可达到85%左右。在准确率低于预定标准(比如70%)时,可通过调整参数、对训练集数据进行调整等方式重新进行训练,以得到符合预定标准的准确率的第一深度学习模型,从而保证利用训练好的第一深度模型可更为准确的对分句进行评分并判断出句子类型。
由此,可将待分析证券研报中的各分句采用上述第一深度学习模型进行评分以得到针对待分析证券研报中的各分句的第一评分,并可基于第一评分从待分析证券研报中的各分句中确定观点类分句。
在本实施例中,该第一评分可为该模型认为的评分,该评分可视为概率(在0至1之间),比如模型认为比较偏向于观点类的分句,则评分更接近于1(比如0.75),模型认为比较偏向于非观点类的分句,则评分更接近于0(比如0.22)。
在得到第一评分后,可根据该第一评分与预先设置的第一分数阈值进行比较,并根据比较结果判断出观点类分句,比如可将第一评分大于第一分数阈值的分句判定为观点类分句,将第一评分不大于第一分数阈值的分句判定为非观点分句。在本实施例中,可将该第一分数阈值设置为0.5,也就是说,第一评分不大于0.5的分句均可确定为非观点类分句,不再参与后续步骤;第一评分大于0.5的分句均可确定为观点类分句,可以继续参与后续步骤。
以此,采用预先训练好的第一深度学习模型,以智能评分的方式分析出分句的类型(观点类分句或非观点类分句),一方面,可大大节省人力,提高分析效率;另一方面,可提高分析结果的准确率。
S04,将待分析证券研报中的各观点类分句采用预先训练好的第二深度学习模型进行评分,以得到针对待分析证券研报中的各观点类分句的第二评分,并基于第二评分从待分析证券研报中的各观点类分句中确定行业观点类分句。
首先,先对第二深度学习模型的训练过程进行说明,该训练过程可包括如下步骤:
步骤210,确定第二数据集。
比如,预先挑选预置篇数(比如300篇)证券研报,比如可挑选一定比例的观点比较明显、尤其是行业观点比较明显、情感趋势比较清晰的证券研报,还可挑选一定比例的观点比较模糊、尤其是行业观点比较模糊、情感趋势不太清晰的证券研报等,然后可对证券研报的摘要部分进行分句处理以得到多个分句,然后以人工方式对每个分句进行第一类型标签的标注,该第一类型标签可包括观点类和非观点类。也就是说,经第一类型标签标注后,第二数据集中的分句可包括观点类分句和非观点类分句。
然后,再以人工方式对每个观点类分句进行第二类型标签的标注,该第二类型标签可包括行业观点类和非行业观点类。也就是说,经第二类型标签标注后,在第二数据集中的分句可包括观点类分句和非观点类分句,且观点类分句中还包括行业观点类分句和非行业观点类分句。
步骤220,将标注有第二类型标签的各分句进行分词处理,得到标注有第二类型标签 的各分词。
也就是说,将第二数据集中的标注有第二类型标签的各分句(也即第二数据集中的行业观点类分句和非行业观点类分句)进行分词处理,比如,可使用现有jieba分词模块对上述各分句进行分词处理,以得到标注有第二类型标签的各分词。
步骤230,根据预置规则将标注有第二类型标签的各分词转换为对应的第二数值,并将标注有第二类型标签的各分词与其对应的第二数值存储于预置的包括分词与数值的对应转换关系的字典中。
具体的,可根据预置规则(比如随机转换,或者可根据实际需求预先设定词语到数值的转换规则等)将标注有第二类型标签的各分词一一转换为数值(在本实施例中,可称为第二数值),其中,第一数值比如可为integer整数型的数值,并可将标注有第二类型标签的各分词与第二数值存储于预置的包括分词与数值的对应转换关系的字典中。
在本实施例中,可将标注有第二类型标签的各分词与第二数值的对应转换关系以文件或者数据表的形式保存于字典中,内容可如上述表1所示。
步骤240,根据字典,将标注有第二类型标签的各分句转换成数值向量格式的分句,形成第二数值向量分句集合。
也即,可根据上述包括分词与数值的对应转换关系的字典,将每个标注有第二类型标签的由多个分词组成的分句,转换成数值向量格式的分句,并形成数值向量分句集合,在本实施例中,可称为第二数值向量分句集合。
步骤250,从第二数值向量分句集合中选取第二预置数量的分句作为第二训练数据。
通常在进行深度学习模型训练之前,可先确定用于训练的训练集和用于测试的测试集,在本实施例中,可从上述第二数值向量分句集合中选取预置数量(在本实施例中,可称为第二预置数量,比如占比90%)的分句作为第二训练数据,剩余(比如占比10%)的分句则可作为第二测试数据。
步骤260,将第二训练数据经深度学习模型进行训练,以获得第二深度学习模型。
在本实施例中,该深度学习模型可为长短期记忆网络(LSTM,Long Short-Term Memory)机器学习模型,在训练之前,可对一些关键参数进行设置,比如嵌入层大小(embedding size)、隐藏层大小(hidden layer size)、批次训练大小(batch size)、全样本循环次数(num epochs)、遗忘率(dropout)、激活函数(activation,分类型变量使用sigmoid)、损失函数(loss,分类型问题使用binary-crossentropy)等参数。然后,可将第二训练数据输入LSTM机器学习模型进行训练,以在训练完毕后获得第二深度学习模型。
此外,在训练完毕获得第二深度学习模型后,还可将第二测试数据输入第二深度学习模型进行测试,以得到该第二深度学习模型的准确率。经多次试验,我们的第二深度学习在第二测试数据上的准确率可达到85%左右。在准确率低于预定标准(比如70%)时,可通过调整参数、对训练集数据进行调整等方式重新进行训练,以得到符合预定标准的准确率的第二深度学习模型,从而保证利用训练好的第二深度模型可更为准确的对分句进行评分并判断出句子类型。
由此,可将由S03中判定的待分析证券研报中的各观点类分句采用上述第二深度学习模型进行评分,以得到针对待分析证券研报中的各观点类分句的第二评分,并可基于第二评分从待分析证券研报中的观点类分句中确定行业观点类分句。
在本实施例中,该第二评分可为该模型认为的评分,该评分可视为概率(在0至1之间),比如模型认为比较偏向于行业观点类的分句,则评分更接近于1(比如0.85),模型认为比较偏向于非行业观点类的分句,则评分更接近于0(比如0.3)。在得到第二评分后,可根据该第二评分与预先设置的第二分数阈值进行比较,并根据比较结果判断出观点类分句,比如可将第二评分大于第二分数阈值的分句判定为行业观点类分句,第二评分不大于第二分数阈值的分句判定为非行业观点分句。在本实施例中,可将该第二分数阈值设置为0.5,也就是说,第二评分不大于0.5的分句均可确定为非行业观点类分句,不再参与后续 步骤;第二评分大于0.5的分句均可确定为行业观点类分句,可以继续参与后续步骤。
以此,采用预先训练好的第二深度学习模型,以智能评分的方式分析出观点类分句的具体类型(行业观点类分句或非行业观点类分句),一方面,可大大节省人力,提高分析效率;另一方面,可提高分析结果的准确率。
S05,将待分析证券研报中的各行业观点类分句采用预先训练好的第三深度学习模型进行评分,以得到针对待分析证券研报中的各行业观点类分句的第三评分,所述第三评分用于确定待分析证券研报中的各行业观点类分句的情感趋势。
首先,先对第三深度学习模型的训练过程进行说明,该训练过程可包括如下步骤:
步骤310,确定第三数据集。
比如,我们可预先精心挑选预置篇数(比如300篇)证券研报,比如可挑选一定比例的观点比较明显、尤其是行业观点比较明显、情感趋势比较清晰的证券研报,还可挑选一定比例的观点比较模糊、尤其是行业观点比较模糊、情感趋势不太清晰的证券研报等,然后可对证券研报的摘要部分进行分句处理以得到多个分句,然后以人工方式对每个分句进行第一类型标签的标注,该第一类型标签可包括观点类和非观点类。也就是说,经第一类型标签标注后,第一数据集中的分句可包括观点类分句和非观点类分句。
然后,再以人工方式对每个观点类分句进行第二类型标签的标注,该第二类型标签可包括行业观点类和非行业观点类。也就是说,经第二类型标签标注后,第二数据集中的分句可包括观点类分句和非观点类分句,同时观点类分句中还包括行业观点类分句和非行业观点类分句。
接下来,再以人工方式对每个行业观点类分句进行第三类型标签的标注,该第三类型标签可包括看涨类和看跌类。也就是说,经第三类型标签标注后,第三数据集中的分句可包括观点类分句和非观点类分句,且观点类分句中还包括行业观点类分句和非行业观点类分句,且行业观点类分句中还包括看涨类行业观点类分句和看跌类行业观点类分句。
步骤320,将标注有第三类型标签的各分句进行分词处理,得到标注有第三类型标签的各分词。
也就是说,将第三数据集中标注有第三类型的各分句(也即第三数据集中的看涨类行业观点类分句和看跌类行业观点类分句)进行分词处理。比如,可使用现有jieba分词模块对上述各分句进行分词处理,以得到标注有第三类型标签的各分词。
步骤330,根据预置规则将标注有第三类型标签的各分词转换为对应的第三数值,并将标注有第三类型标签的各分词与其对应的第三数值存储于预置的包括分词与数值的对应转换关系的字典中。
具体的,可根据预置规则(比如随机转换,或者可根据实际需求预先设定词语到数值的转换规则等)将标注有第三类型标签的各分词一一转换为数值(在本实施例中,可称为第三数值),其中,第三数值比如可为integer整数型的数值,并可将标注有第三类型标签的各分词与第三数值存储于预置的包括分词与数值的对应转换关系的字典中。
在本实施例中,可将标注有第三类型标签的各分词与第三数值的对应转换关系以文件或者数据表的形式保存于字典中,内容可如上述表1所示。
步骤340,根据所述字典,将标注有第三类型标签的各分句转换成数值向量格式的分句,形成第三数值向量分句集合。
也即,可根据上述包括分词与数值的对应转换关系的字典,将每个标注有第三类型标签的由多个分词组成的分句,转换成数值向量格式的分句,并形成数值向量分句集合,在本实施例中,可称为第三数值向量分句集合。
步骤350,从第三数值向量分句集合中选取第三预置数量的分句作为第三训练数据。
通常在进行深度学习模型训练之前,可先确定用于训练的训练集和用于测试的测试集,在本实施例中,可从上述第三数值向量分句集合中选取预置数量(在本实施例中,可称为第三预置数量,比如占比90%)的分句作为第三训练数据,剩余(比如占比10%)的分句 则可作为第三测试数据。
步骤360,将第三训练数据经深度学习模型进行训练,以获得第三深度学习模型。
在本实施例中,该深度学习模型可为长短期记忆网络(LSTM,Long Short-Term Memory)机器学习模型,在训练之前,可对一些关键参数进行设置,比如嵌入层大小(embedding size)、隐藏层大小(hidden layer size)、批次训练大小(batch size)、全样本循环次数(num epochs)、遗忘率(dropout)、激活函数(activation,分类型变量使用sigmoid)、损失函数(loss,分类型问题使用binary-crossentropy)等参数。然后,可将第三训练数据输入LSTM机器学习模型进行训练,以在训练完毕后获得第三深度学习模型。
此外,在训练完毕获得第三深度学习模型后,还可将第三测试数据输入第三深度学习模型进行测试,以得到该第三深度学习模型的准确率。经多次试验,我们的第三深度学习在第三测试数据上的准确率可达到85%左右。在准确率低于预定标准(比如70%)时,可通过调整参数、对训练集数据进行调整等方式重新进行训练,以得到符合预定标准的准确率的第三深度学习模型,从而保证利用训练好的第三深度模型可更为准确的对分句进行评分并根据需求判断出句子类型。
由此,可将由S04中判定的待分析证券研报中的各行业观点类分句采用上述第三深度学习模型进行评分,以得到针对待分析证券研报中的各行业观点类分句的第三评分,并可基于第三评分确定待分析证券研报中的行业观点类分句的情感趋势,也即看涨类行业观点类分句还是看跌类行业观点类分句。
在本实施例中,该第三评分可为该模型认为的评分,该评分可视为概率(在0至1之间),比如模型认为比较偏向于看涨类行业观点类的分句,则评分更接近于1(比如0.85),模型认为比较偏向于看跌类行业观点类的分句,则评分更接近于0(比如0.3)。在得到第三评分后,可根据该第三评分与预先设置的第三分数阈值进行比较,并根据比较结果进一步判断看涨或看跌情绪的程度,比如可将第三评分大于第三分数阈值的分句判定为看涨情绪较高或看跌情绪较低的行业观点类分句,将第三评分不大于第三分数阈值的分句判定为看涨情绪较低或看跌情绪较高的行业观点类分句。在本实施例中,可将该第二分数阈值设置为0.5-0.6之间,优选设置为0.6,也就是说,第三评分大于0.6的分句均可确定为看涨情绪较高或看跌情绪较低的行业观点类分句;第三评分不大于0.6的分句均可确定为看涨情绪较低或看跌情绪较高的行业观点类分句。
以此,采用预先训练好的第三深度学习模型,以智能评分的方式得出用于确定各行业观点类分句的情感趋势的第三类评分,以用来分析各行业观点类分句是看涨类行业观点类分句还是看跌类行业观点类分句,以及看涨看跌情绪的程度,一方面,可大大节省人力,提高分析效率;另一方面,可提高分析结果的准确率。
S06,根据待分析证券研报中的各行业观点类分句的第二评分及第三评分确定待分析证券研报的整体情感趋势。
在本实施例中,在S04中得到待分析证券研报中的各行业观点类分句的第二评分,以及在S05中得到待分析证券研报中的各行业观点类分句的第三评分后,可结合这两个评分对待分析证券研报的整体情感趋势进行评估。
在具体实现时,可先计算待分析证券研报中的每个行业观点类分句的第二评分与第三评分的积,作为每个行业观点类分句的第一积值;再计算所有行业观点类分句的第一积值的和,作为第一和值;然后计算所有行业观点类分句的第二评分的和,作为第二和值;接下来,将第一和值除以第二和值,以得到待分析证券研报的整体情感评分(也可以理解为涨跌评分),由于第二评分和第三评分都处于0至1之间,因此,计算得到的待分析证券研报的整体情感得分也处于0至1之间。
在计算得到待分析证券研报的整体情感评分之后,可继续判断该整体情感评分是否高于预置评分阈值,该预置评分阈值比如可根据实际经验设置为0.5-0.6之间,优选设置为0.6。
若整体情感评分高于上述预置评分阈值,则可确定该待分析证券研报的整体情感趋势 为看涨,若整体情感评分不高于上述预置评分阈值,则可确定该待分析证券研报的整体情感趋势为看跌。
以此,可通过对深度学习模型评出的待分析证券研报中的各行业观点类分句的第二评分(可用于确定是否为较为明显的行业观点)及第三评分(可用于确定各行业观点的情感趋势)进行综合计算的方式,得到待分析证券研报的整体情感评分,并通过该待分析证券研报的整体情感评分与预定指标的比较结果,得到待分析证券研报的整体情感趋势,从而可得到更为客观且准确的情感趋势分析结果。
此外,在确定待分析证券研报的整体情感趋势之后,还可对待分析证券研报中的核心观点进行提取。在具体实现时,比如可将待分析证券研报中的各行业观点类分句根据其对应的第三评分分为两部分,包括第三评分大于0.6的部分和第三评分不大于0.6的部分,且两部分均可以正序(分值从大到小)进行排序。由此,可根据各行业观点类分句的第三评分确定与证券研报的整体情感趋势一致的行业观点类分句。
也就是说,若待分析证券研报的整体情感趋势为看涨,则可将第三评分大于0.6的行业观点类分句,确定为与待分析证券研报的整体情感趋势一致的行业观点类分句;若待分析证券研报的整体情感趋势为看跌,则可将第三评分不大于0.6的行业观点类分句,确定为与待分析证券研报的整体情感趋势一致的行业观点类分句。
在确定了与待分析证券研报的整体情感趋势一致的行业观点类分句之后,若待分析证券研报的整体情感趋势为看涨,可选取第三评分最高的预设个数的行业观点类分句,作为待分析证券研报的核心观点并输出;若待分析证券研报的整体情感趋势为看跌,可选取第三评分最低的预设个数的行业观点类分句,作为待分析证券研报的核心观点并输出。其中该预设个数可根据实际需求进行设置,在本实施例中,比如可设置为2个。
也就是说,若待分析证券研报的整体情感趋势为看涨,则可在上述第三评分大于0.6的部分中选取2个第三评分最高(即排序为第一、第二)的行业观点类分句,作为该待分析证券研报的核心观点并输出;若待分析证券研报的整体情感趋势为看跌,则可在上述第三评分不大于0.6的部分中选取2个第三评分最低(即排序为倒数第一、第二)的行业观点类分句,作为该待分析证券研报的核心观点并输出。
以此,可根据客观的评分结果选取出与待分析证券研报的整体情感趋势一致且最代表核心观点的分句,从而可保证选取结果的准确性,以便于用户准确了解该待分析证券研报的核心观点。
根据本实施例的各个实施方式,可先将待分析证券研报进行分句处理,再将各分句采用预先训练好的第一深度学习模型进行评分以得到各分句的第一评分,并根据第一评分判断分句是否为观点类分句,然后将判断出的观点类分句采用预先训练好的第二深度学习模型进行评分以得到各观点类分句的第二评分,并根据第二评分判断各观点类分句是否为行业观点类分句,接下来将判断出的行业观点类分句采用预先训练好的第三深度学习模型进行评分以得到各行业观点类分句的第三评分,并根据第三评分判断各行业观点类分句的情感趋势,最后根据各行业观点类分句的第二评分及第三评分确定该待分析证券研报的整体情感趋势。通过上述方案,可基于预先训练好的深度学习模型通过评分的方式挑选出观点类分句、行业观点类分句以及确定行业观点类分句的情感趋势,并最终通过客观的评分确定该证券研报的整体情感趋势,由此,上述利用深度学习模型智能评分并分析的过程,不但可大大节省人力,而且可提高分析效率及分析结果的准确率。
实施例2
基于上述实施例1中提供的利用深度学习模型的证券研报分析方法,本实施例中提供一种利用深度学习模型的证券研报分析装置,具体地,图2至4示出了该利用深度学习模型的证券研报分析装置的可选的结构框图,该利用深度学习模型的证券研报分析装置被分割成一个或多个程序模块,一个或者多个程序模块被存储于存储介质中,并由一个或多个 处理器所执行,以完成本申请。本申请所称的程序模块是指能够完成特定功能的一系列计算机程序指令段,比程序本身更适合描述利用深度学习模型的证券研报分析装置在存储介质中的执行过程,以下描述将具体介绍本实施例各程序模块的功能。
如图2所示,该利用深度学习模型的证券研报分析装置20可包括:
接收模块21,可接收输入的待分析证券研报;
分句模块22,可用于对待分析证券研报进行分句处理,得到待分析证券研报中的各分句;
第一评分模块23,可用于将待分析证券研报中的各分句采用预先训练好的第一深度学习模型进行评分,以得到针对待分析证券研报中的各分句的第一评分,并基于第一评分从待分析证券研报中的各分句中确定观点类分句;
第二评分模块24,可用于将待分析证券研报中的各观点类分句采用预先训练好的第二深度学习模型进行评分,以得到针对待分析证券研报中的各观点类分句的第二评分,并基于第二评分从待分析证券研报中的各观点类分句中确定行业观点类分句;
第三评分模块25,可用于将待分析证券研报中的各行业观点类分句采用预先训练好的第三深度学习模型进行评分,以得到针对待分析证券研报中的各行业观点类分句的第三评分,所述第三评分用于确定待分析证券研报中的各行业观点类分句的情感趋势;
情感趋势确定模块26,可用于根据待分析证券研报中的各行业观点类分句的第二评分及第三评分确定待分析证券研报的整体情感趋势。
进一步的,参看图3所示,分句模块22,可具体包括:
分句单元221,可用于根据预置类型的符号对待分析证券研报进行分句处理,得到待分析证券研报中的各分句;
分词单元222,可用于对待分析证券研报中的各分句进行分词处理,得到待分析证券研报中的各分词;
数值确定单元223,可用于基于预置的包括分词与数值的对应转换关系的字典,确定待分析证券研报中的各分词对应的数值;
转换单元224,可用于根据确定结果,将待分析证券研报中的各分句转换成数值向量格式的分句。
在本实施例中,第一深度学习模型的训练过程,可包括如下步骤:
步骤110,确定第一数据集,所述第一数据集包括预先对预置篇数的证券研报的摘要部分进行分句处理后得到的多个分句,其中每个分句具有预先标注的第一类型标签,第一类型标签包括观点类和非观点类;
步骤120,将标注有第一类型标签的各分句进行分词处理,得到标注有第一类型标签的各分词;
步骤130,根据预置规则将标注有第一类型标签的各分词转换为对应的第一数值,并将标注有第一类型标签的各分词与其对应的第一数值存储于预置的包括分词与数值的对应转换关系的字典中;
步骤140,根据字典,将标注有第一类型标签的各分句转换成数值向量格式的分句,形成第一数值向量分句集合;
步骤150,从第一数值向量分句集合中选取第一预置数量的分句作为第一训练数据;
步骤160,将第一训练数据经深度学习模型进行训练,以获得第一深度学习模型。
在本实施例中,第二深度学习模型的训练过程,可包括如下步骤:
步骤210,确定第二数据集,其中,第二数据集包括预先对预置篇数的证券研报的摘要部分进行分句处理后得到的多个分句,其中每个分句具有预先标注的第一类型标签,第一类型标签包括观点类和非观点类,且标注有观点类标签的分句同时具有预先标注的第二类型标签,第二类型标签包括行业观点类和非行业观点类;
步骤220,将标注有第二类型标签的各分句进行分词处理,得到标注有第二类型标签 的各分词;
步骤230,根据预置规则将标注有第二类型标签的各分词转换为对应的第二数值,并将标注有第二类型标签的各分词与其对应的第二数值存储于预置的包括分词与数值的对应转换关系的字典中;
步骤240,根据字典,将标注有第二类型标签的各分句转换成数值向量格式的分句,形成第二数值向量分句集合;
步骤250,从第二数值向量分句集合中选取第二预置数量的分句作为第二训练数据;
步骤260,将第二训练数据经深度学习模型进行训练,以获得第二深度学习模型。
在本实施例中,第三深度学习模型的训练过程,可包括如下步骤:
步骤310,确定第三数据集,其中,第三数据集包括预先对预置篇数的证券研报的摘要部分进行分句处理后得到的多个分句,其中每个分句具有预先标注的第一类型标签,第一类型标签包括观点类和非观点类,且标注有观点类标签的分句同时具有预先标注的第二类型标签,第二类型标签包括行业观点类和非行业观点类,且标注有行业观点类标签的分句同时具有预先标注的第三类型标签,第三类型标签包括看涨类和看跌类;
步骤320,将标注有第三类型标签的各分句进行分词处理,得到标注有第三类型标签的各分词;
步骤330,根据预置规则将标注有第三类型标签的各分词转换为对应的第三数值,并将标注有第三类型标签的各分词与其对应的第三数值存储于预置的包括分词与数值的对应转换关系的字典中;
步骤340,根据字典,将标注有第三类型标签的各分句转换成数值向量格式的分句,形成第三数值向量分句集合;
步骤350,从第三数值向量分句集合中选取第三预置数量的分句作为第三训练数据;
步骤360,将第三训练数据经深度学习模型进行训练,以获得第三深度学习模型。
进一步的,深度学习模型为长短期记忆网络机器学习模型。
此外,参看图4所示,情感趋势确定模块26,可具体包括:
第一计算单元261,用于计算待分析证券研报中的每个行业观点类分句的第二评分与第三评分的积,作为每个行业观点类分句的第一积值;
第二计算单元262,用于计算所有行业观点类分句的第一积值的和,作为第一和值;
第三计算单元263,用于计算所有行业观点类分句的第二评分的和,作为第二和值;
第四计算单元264,用于将第一和值除以第二和值,得到待分析证券研报的整体情感评分;
判断单元265,用于判断待分析证券研报的整体情感评分是否高于预置评分阈值;
情感趋势确定单元266,可用于在判断单元265的判断结果为是时,确定待分析证券研报的整体情感趋势为看涨,在判断单元265的判断结果为否时,确定待分析证券研报的整体情感趋势为看跌。
进一步的,该证券研报的分析装置,还可包括:
行业观点类分句确定模块,用于根据待分析证券研报中的各行业观点类分句的第三评分确定与证券研报的整体情感趋势一致的行业观点类分句;
选取模块,用于从确定结果中选取第三评分最高或最低的预设个数的行业观点类分句,作为证券研报的核心观点并输出。
关于上述实施例中的装置,其中各个单元、模块执行操作的具体方式已经在有关该方法的实施例中进行了详细描述,此处将不做详细阐述说明。
通过本实施例的各个实施方式,可先将待分析证券研报进行分句处理,再将各分句采用预先训练好的第一深度学习模型进行评分以得到各分句的第一评分,并根据第一评分判断分句是否为观点类分句,然后将判断出的观点类分句采用预先训练好的第二深度学习模型进行评分以得到各观点类分句的第二评分,并根据第二评分判断各观点类分句是否为行 业观点类分句,接下来将判断出的行业观点类分句采用预先训练好的第三深度学习模型进行评分以得到各行业观点类分句的第三评分,并根据第三评分判断各行业观点类分句的情感趋势,最后根据各行业观点类分句的第二评分及第三评分确定该待分析证券研报的整体情感趋势。通过上述方案,可基于预先训练好的深度学习模型通过评分的方式挑选出观点类分句、行业观点类分句以及确定行业观点类分句的情感趋势,并最终通过客观的评分确定该证券研报的整体情感趋势,由此,上述利用深度学习模型智能评分并分析的过程,不但可大大节省人力,而且可提高分析效率及分析结果的准确率。
实施例3
本实施例还提供一种计算机设备,如可以执行程序的智能手机、平板电脑、笔记本电脑、台式计算机、机架式服务器、刀片式服务器、塔式服务器或机柜式服务器(包括独立的服务器,或者多个服务器所组成的服务器集群)等。如图5所示,本实施例的计算机设备50至少包括但不限于:可通过系统总线相互通信连接的存储器51、处理器52,如图5所示。需要指出的是,图5仅示出了具有组件51-52的计算机设备50,但是应理解的是,并不要求实施所有示出的组件,可以替代的实施更多或者更少的组件。
本实施例中,存储器51(即可读存储介质)包括闪存、硬盘、多媒体卡、卡型存储器(例如,SD或DX存储器等)、随机访问存储器(RAM)、静态随机访问存储器(SRAM)、只读存储器(ROM)、电可擦除可编程只读存储器(EEPROM)、可编程只读存储器(PROM)、磁性存储器、磁盘、光盘等。在一些实施例中,存储器51可以是计算机设备50的内部存储单元,例如该计算机设备50的硬盘或内存。在另一些实施例中,存储器51也可以是计算机设备50的外部存储设备,例如该计算机设备50上配备的插接式硬盘,智能存储卡(Smart Media Card,SMC),安全数字(Secure Digital,SD)卡,闪存卡(Flash Card)等。当然,存储器51还可以既包括计算机设备50的内部存储单元也包括其外部存储设备。本实施例中,存储器51通常用于存储安装于计算机设备50的操作系统和各类应用软件,例如实施例2的利用深度学习模型的证券研报分析装置的程序代码等。此外,存储器51还可以用于暂时地存储已经输出或者将要输出的各类数据。
处理器52在一些实施例中可以是中央处理器(Central Processing Unit,CPU)、控制器、微控制器、微处理器、或其他数据处理芯片。该处理器52通常用于控制计算机设备50的总体操作。本实施例中,处理器52用于运行存储器51中存储的程序代码或者处理数据,例如利用深度学习模型的证券研报分析装置等。
实施例4
本实施例还提供一种计算机可读存储介质,如闪存、硬盘、多媒体卡、卡型存储器(例如,SD或DX存储器等)、随机访问存储器(RAM)、静态随机访问存储器(SRAM)、只读存储器(ROM)、电可擦除可编程只读存储器(EEPROM)、可编程只读存储器(PROM)、磁性存储器、磁盘、光盘、服务器、App应用商城等等,其上存储有计算机程序,程序被处理器执行时实现相应功能。本实施例的计算机可读存储介质用于利用深度学习模型的证券研报分析装置,被处理器执行时实现实施例1的利用深度学习模型的证券研报分析方法。
上述本申请实施例序号仅仅为了描述,不代表实施例的优劣。
通过以上的实施方式的描述,本领域的技术人员可以清楚地了解到上述实施例方法可借助软件加必需的通用硬件平台的方式来实现,当然也可以通过硬件,但很多情况下前者是更佳的实施方式。
以上仅为本申请的优选实施例,并非因此限制本申请的专利范围,凡是利用本申请说明书及附图内容所作的等效结构或等效流程变换,或直接或间接运用在其他相关的技术领域,均同理包括在本申请的专利保护范围内。
Claims (20)
- 一种利用深度学习模型的证券研报分析方法,其特征在于,包括如下步骤:S01,接收输入的待分析证券研报;S02,对待分析证券研报进行分句处理,得到待分析证券研报中的各分句;S03,将待分析证券研报中的各分句采用预先训练好的第一深度学习模型进行评分,以得到针对待分析证券研报中的各分句的第一评分,并基于第一评分从待分析证券研报中的各分句中确定观点类分句;S04,将待分析证券研报中的各观点类分句采用预先训练好的第二深度学习模型进行评分,以得到针对待分析证券研报中的各观点类分句的第二评分,并基于第二评分从待分析证券研报中的各观点类分句中确定行业观点类分句;S05,将待分析证券研报中的各行业观点类分句采用预先训练好的第三深度学习模型进行评分,以得到针对待分析证券研报中的各行业观点类分句的第三评分,所述第三评分用于确定待分析证券研报中的各行业观点类分句的情感趋势;S06,根据待分析证券研报中的各行业观点类分句的第二评分及第三评分确定待分析证券研报的整体情感趋势。
- 根据权利要求1所述的利用深度学习模型的证券研报分析方法,其特征在于,S02对待分析证券研报进行分句处理,得到待分析证券研报中的各分句,包括:根据预置类型的符号对待分析证券研报进行分句处理,得到待分析证券研报中的各分句;对待分析证券研报中的各分句进行分词处理,得到待分析证券研报中的各分词;基于预置的包括分词与数值的对应转换关系的字典,确定待分析证券研报中的各分词对应的数值;根据确定结果,将待分析证券研报中的各分句转换成数值向量格式的分句。
- 根据权利要求1所述的利用深度学习模型的证券研报分析方法,其特征在于,第一深度学习模型的训练过程,包括如下步骤:步骤110,确定第一数据集,所述第一数据集包括预先对预置篇数的证券研报的摘要部分进行分句处理后得到的多个分句,其中每个分句具有预先标注的第一类型标签,第一类型标签包括观点类和非观点类;步骤120,将标注有第一类型标签的各分句进行分词处理,得到标注有第一类型标签的各分词;步骤130,根据预置规则将标注有第一类型标签的各分词转换为对应的第一数值,并将标注有第一类型标签的各分词与其对应的第一数值存储于预置的包括分词与数值的对应转换关系的字典中;步骤140,根据所述字典,将标注有第一类型标签的各分句转换成数值向量格式的分句,形成第一数值向量分句集合;步骤150,从第一数值向量分句集合中选取第一预置数量的分句作为第一训练数据;步骤160,将第一训练数据经深度学习模型进行训练,以获得第一深度学习模型。
- 根据权利要求1所述的利用深度学习模型的证券研报分析方法,其特征在于,第二深度学习模型的训练过程,包括如下步骤:步骤210,确定第二数据集,所述第二数据集包括预先对预置篇数的证券研报的摘要部分进行分句处理后得到的多个分句,其中每个分句具有预先标注的第一类型标签,第一类型标签包括观点类和非观点类,且标注有观点类标签的分句同时具有预先标注的第二类型标签,第二类型标签包括行业观点类和非行业观点类;步骤220,将标注有第二类型标签的各分句进行分词处理,得到标注有第二类型标签 的各分词;步骤230,根据预置规则将标注有第二类型标签的各分词转换为对应的第二数值,并将标注有第二类型标签的各分词与其对应的第二数值存储于预置的包括分词与数值的对应转换关系的字典中;步骤240,根据所述字典,将标注有第二类型标签的各分句转换成数值向量格式的分句,形成第二数值向量分句集合;步骤250,从第二数值向量分句集合中选取第二预置数量的分句作为第二训练数据;步骤260,将第二训练数据经深度学习模型进行训练,以获得第二深度学习模型。
- 根据权利要求1所述的利用深度学习模型的证券研报分析方法,其特征在于,第三深度学习模型的训练过程,包括如下步骤:步骤310,确定第三数据集,所述第三数据集包括预先对预置篇数的证券研报的摘要部分进行分句处理后得到的多个分句,其中每个分句具有预先标注的第一类型标签,第一类型标签包括观点类和非观点类,且标注有观点类标签的分句同时具有预先标注的第二类型标签,第二类型标签包括行业观点类和非行业观点类,且标注有行业观点类标签的分句同时具有预先标注的第三类型标签,第三类型标签包括看涨类和看跌类;步骤320,将标注有第三类型标签的各分句进行分词处理,得到标注有第三类型标签的各分词;步骤330,根据预置规则将标注有第三类型标签的各分词转换为对应的第三数值,并将标注有第三类型标签的各分词与其对应的第三数值存储于预置的包括分词与数值的对应转换关系的字典中;步骤340,根据所述字典,将标注有第三类型标签的各分句转换成数值向量格式的分句,形成第三数值向量分句集合;步骤350,从第三数值向量分句集合中选取第三预置数量的分句作为第三训练数据;步骤360,将第三训练数据经深度学习模型进行训练,以获得第三深度学习模型。
- 根据权利要求3或4或5所述的利用深度学习模型的证券研报分析方法,其特征在于,所述深度学习模型为长短期记忆网络机器学习模型。
- 根据权利要求1所述的利用深度学习模型的证券研报分析方法,其特征在于,S06根据待分析证券研报中的各行业观点类分句的第二评分及第三评分确定待分析证券研报的整体情感趋势,包括:计算待分析证券研报中的每个行业观点类分句的第二评分与第三评分的积,作为每个行业观点类分句的第一积值;计算所有行业观点类分句的第一积值的和,作为第一和值;计算所有行业观点类分句的第二评分的和,作为第二和值;将第一和值除以第二和值,得到待分析证券研报的整体情感评分;判断所述待分析证券研报的整体情感评分是否高于预置评分阈值;若是,则确定待分析证券研报的整体情感趋势为看涨,若否,则确定待分析证券研报的整体情感趋势为看跌。
- 根据权利要求1所述的利用深度学习模型的证券研报分析方法,其特征在于,还包括:根据待分析证券研报中的各行业观点类分句的第三评分确定与证券研报的整体情感趋势一致的行业观点类分句;从确定结果中选取第三评分最高或最低的预设个数的行业观点类分句,作为待分析证券研报的核心观点并输出。
- 一种利用深度学习模型的证券研报分析装置,其特征在于,包括:接收模块,用于接收输入的待分析证券研报;分句模块,用于对所述待分析证券研报进行分句处理,得到待分析证券研报中的各分 句;第一评分模块,用于将待分析证券研报中的各分句采用预先训练好的第一深度学习模型进行评分以得到针对待分析证券研报中的各分句的第一评分,并基于第一评分从待分析证券研报中的各分句中确定观点类分句;第二评分模块,用于将待分析证券研报中的各观点类分句采用预先训练好的第二深度学习模型进行评分以得到针对待分析证券研报中的各观点类分句的第二评分,并基于第二评分从待分析证券研报中的各观点类分句中确定行业观点类分句;第三评分模块,用于将待分析证券研报中的各行业观点类分句采用预先训练好的第三深度学习模型进行评分以得到针对待分析证券研报中的各行业观点类分句的第三评分,所述第三评分用于确定待分析证券研报中的各行业观点类分句的情感趋势;情感趋势确定模块,用于根据待分析证券研报中的各行业观点类分句的第二评分及第三评分确定待分析证券研报的整体情感趋势。
- 一种计算机设备,所述计算机设备包括存储器、处理器以及存储在存储器上并可在处理器上运行的计算机程序,其特征在于,所述处理器执行所述程序时实现利用深度学习模型的证券研报分析方法的以下步骤:S01,接收输入的待分析证券研报;S02,对待分析证券研报进行分句处理,得到待分析证券研报中的各分句;S03,将待分析证券研报中的各分句采用预先训练好的第一深度学习模型进行评分,以得到针对待分析证券研报中的各分句的第一评分,并基于第一评分从待分析证券研报中的各分句中确定观点类分句;S04,将待分析证券研报中的各观点类分句采用预先训练好的第二深度学习模型进行评分,以得到针对待分析证券研报中的各观点类分句的第二评分,并基于第二评分从待分析证券研报中的各观点类分句中确定行业观点类分句;S05,将待分析证券研报中的各行业观点类分句采用预先训练好的第三深度学习模型进行评分,以得到针对待分析证券研报中的各行业观点类分句的第三评分,所述第三评分用于确定待分析证券研报中的各行业观点类分句的情感趋势;S06,根据待分析证券研报中的各行业观点类分句的第二评分及第三评分确定待分析证券研报的整体情感趋势。
- 根据权利要求10所述的计算机设备,其特征在于,第一深度学习模型的训练过程,包括如下步骤:步骤110,确定第一数据集,所述第一数据集包括预先对预置篇数的证券研报的摘要部分进行分句处理后得到的多个分句,其中每个分句具有预先标注的第一类型标签,第一类型标签包括观点类和非观点类;步骤120,将标注有第一类型标签的各分句进行分词处理,得到标注有第一类型标签的各分词;步骤130,根据预置规则将标注有第一类型标签的各分词转换为对应的第一数值,并将标注有第一类型标签的各分词与其对应的第一数值存储于预置的包括分词与数值的对应转换关系的字典中;步骤140,根据所述字典,将标注有第一类型标签的各分句转换成数值向量格式的分句,形成第一数值向量分句集合;步骤150,从第一数值向量分句集合中选取第一预置数量的分句作为第一训练数据;步骤160,将第一训练数据经深度学习模型进行训练,以获得第一深度学习模型。
- 根据权利要求10所述的计算机设备,其特征在于,第二深度学习模型的训练过程,包括如下步骤:步骤210,确定第二数据集,所述第二数据集包括预先对预置篇数的证券研报的摘要部分进行分句处理后得到的多个分句,其中每个分句具有预先标注的第一类型标签,第一 类型标签包括观点类和非观点类,且标注有观点类标签的分句同时具有预先标注的第二类型标签,第二类型标签包括行业观点类和非行业观点类;步骤220,将标注有第二类型标签的各分句进行分词处理,得到标注有第二类型标签的各分词;步骤230,根据预置规则将标注有第二类型标签的各分词转换为对应的第二数值,并将标注有第二类型标签的各分词与其对应的第二数值存储于预置的包括分词与数值的对应转换关系的字典中;步骤240,根据所述字典,将标注有第二类型标签的各分句转换成数值向量格式的分句,形成第二数值向量分句集合;步骤250,从第二数值向量分句集合中选取第二预置数量的分句作为第二训练数据;步骤260,将第二训练数据经深度学习模型进行训练,以获得第二深度学习模型。
- 根据权利要求10所述的计算机设备,其特征在于,第三深度学习模型的训练过程,包括如下步骤:步骤310,确定第三数据集,所述第三数据集包括预先对预置篇数的证券研报的摘要部分进行分句处理后得到的多个分句,其中每个分句具有预先标注的第一类型标签,第一类型标签包括观点类和非观点类,且标注有观点类标签的分句同时具有预先标注的第二类型标签,第二类型标签包括行业观点类和非行业观点类,且标注有行业观点类标签的分句同时具有预先标注的第三类型标签,第三类型标签包括看涨类和看跌类;步骤320,将标注有第三类型标签的各分句进行分词处理,得到标注有第三类型标签的各分词;步骤330,根据预置规则将标注有第三类型标签的各分词转换为对应的第三数值,并将标注有第三类型标签的各分词与其对应的第三数值存储于预置的包括分词与数值的对应转换关系的字典中;步骤340,根据所述字典,将标注有第三类型标签的各分句转换成数值向量格式的分句,形成第三数值向量分句集合;步骤350,从第三数值向量分句集合中选取第三预置数量的分句作为第三训练数据;步骤360,将第三训练数据经深度学习模型进行训练,以获得第三深度学习模型。
- 根据权利要求10所述的计算机设备,其特征在于,S06根据待分析证券研报中的各行业观点类分句的第二评分及第三评分确定待分析证券研报的整体情感趋势,包括:计算待分析证券研报中的每个行业观点类分句的第二评分与第三评分的积,作为每个行业观点类分句的第一积值;计算所有行业观点类分句的第一积值的和,作为第一和值;计算所有行业观点类分句的第二评分的和,作为第二和值;将第一和值除以第二和值,得到待分析证券研报的整体情感评分;判断所述待分析证券研报的整体情感评分是否高于预置评分阈值;若是,则确定待分析证券研报的整体情感趋势为看涨,若否,则确定待分析证券研报的整体情感趋势为看跌。
- 根据权利要求10所述的计算机设备,其特征在于,还包括:根据待分析证券研报中的各行业观点类分句的第三评分确定与证券研报的整体情感趋势一致的行业观点类分句;从确定结果中选取第三评分最高或最低的预设个数的行业观点类分句,作为待分析证券研报的核心观点并输出。
- 一种计算机可读存储介质,其上存储有计算机程序,其特征在于:所述程序被处理器执行时实现利用深度学习模型的证券研报分析方法的以下步骤:S01,接收输入的待分析证券研报;S02,对待分析证券研报进行分句处理,得到待分析证券研报中的各分句;S03,将待分析证券研报中的各分句采用预先训练好的第一深度学习模型进行评分,以得到针对待分析证券研报中的各分句的第一评分,并基于第一评分从待分析证券研报中的各分句中确定观点类分句;S04,将待分析证券研报中的各观点类分句采用预先训练好的第二深度学习模型进行评分,以得到针对待分析证券研报中的各观点类分句的第二评分,并基于第二评分从待分析证券研报中的各观点类分句中确定行业观点类分句;S05,将待分析证券研报中的各行业观点类分句采用预先训练好的第三深度学习模型进行评分,以得到针对待分析证券研报中的各行业观点类分句的第三评分,所述第三评分用于确定待分析证券研报中的各行业观点类分句的情感趋势;S06,根据待分析证券研报中的各行业观点类分句的第二评分及第三评分确定待分析证券研报的整体情感趋势。
- 根据权利要求16所述的计算机可读存储介质,其特征在于,第一深度学习模型的训练过程,包括如下步骤:步骤110,确定第一数据集,所述第一数据集包括预先对预置篇数的证券研报的摘要部分进行分句处理后得到的多个分句,其中每个分句具有预先标注的第一类型标签,第一类型标签包括观点类和非观点类;步骤120,将标注有第一类型标签的各分句进行分词处理,得到标注有第一类型标签的各分词;步骤130,根据预置规则将标注有第一类型标签的各分词转换为对应的第一数值,并将标注有第一类型标签的各分词与其对应的第一数值存储于预置的包括分词与数值的对应转换关系的字典中;步骤140,根据所述字典,将标注有第一类型标签的各分句转换成数值向量格式的分句,形成第一数值向量分句集合;步骤150,从第一数值向量分句集合中选取第一预置数量的分句作为第一训练数据;步骤160,将第一训练数据经深度学习模型进行训练,以获得第一深度学习模型。
- 根据权利要求16所述的计算机可读存储介质,其特征在于,第二深度学习模型的训练过程,包括如下步骤:步骤210,确定第二数据集,所述第二数据集包括预先对预置篇数的证券研报的摘要部分进行分句处理后得到的多个分句,其中每个分句具有预先标注的第一类型标签,第一类型标签包括观点类和非观点类,且标注有观点类标签的分句同时具有预先标注的第二类型标签,第二类型标签包括行业观点类和非行业观点类;步骤220,将标注有第二类型标签的各分句进行分词处理,得到标注有第二类型标签的各分词;步骤230,根据预置规则将标注有第二类型标签的各分词转换为对应的第二数值,并将标注有第二类型标签的各分词与其对应的第二数值存储于预置的包括分词与数值的对应转换关系的字典中;步骤240,根据所述字典,将标注有第二类型标签的各分句转换成数值向量格式的分句,形成第二数值向量分句集合;步骤250,从第二数值向量分句集合中选取第二预置数量的分句作为第二训练数据;步骤260,将第二训练数据经深度学习模型进行训练,以获得第二深度学习模型。
- 根据权利要求16所述的计算机可读存储介质,其特征在于,第三深度学习模型的训练过程,包括如下步骤:步骤310,确定第三数据集,所述第三数据集包括预先对预置篇数的证券研报的摘要部分进行分句处理后得到的多个分句,其中每个分句具有预先标注的第一类型标签,第一类型标签包括观点类和非观点类,且标注有观点类标签的分句同时具有预先标注的第二类型标签,第二类型标签包括行业观点类和非行业观点类,且标注有行业观点类标签的分句 同时具有预先标注的第三类型标签,第三类型标签包括看涨类和看跌类;步骤320,将标注有第三类型标签的各分句进行分词处理,得到标注有第三类型标签的各分词;步骤330,根据预置规则将标注有第三类型标签的各分词转换为对应的第三数值,并将标注有第三类型标签的各分词与其对应的第三数值存储于预置的包括分词与数值的对应转换关系的字典中;步骤340,根据所述字典,将标注有第三类型标签的各分句转换成数值向量格式的分句,形成第三数值向量分句集合;步骤350,从第三数值向量分句集合中选取第三预置数量的分句作为第三训练数据;步骤360,将第三训练数据经深度学习模型进行训练,以获得第三深度学习模型。
- 根据权利要求16所述的计算机可读存储介质,其特征在于,S06根据待分析证券研报中的各行业观点类分句的第二评分及第三评分确定待分析证券研报的整体情感趋势,包括:计算待分析证券研报中的每个行业观点类分句的第二评分与第三评分的积,作为每个行业观点类分句的第一积值;计算所有行业观点类分句的第一积值的和,作为第一和值;计算所有行业观点类分句的第二评分的和,作为第二和值;将第一和值除以第二和值,得到待分析证券研报的整体情感评分;判断所述待分析证券研报的整体情感评分是否高于预置评分阈值;若是,则确定待分析证券研报的整体情感趋势为看涨,若否,则确定待分析证券研报的整体情感趋势为看跌。
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811228761.7 | 2018-10-22 | ||
| CN201811228761.7A CN109284504A (zh) | 2018-10-22 | 2018-10-22 | 利用深度学习模型的证券研报分析方法及装置 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2020082609A1 true WO2020082609A1 (zh) | 2020-04-30 |
Family
ID=65177794
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/CN2019/070287 Ceased WO2020082609A1 (zh) | 2018-10-22 | 2019-01-03 | 利用深度学习模型的证券研报分析方法及装置 |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN109284504A (zh) |
| WO (1) | WO2020082609A1 (zh) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110110295B (zh) * | 2019-04-04 | 2023-10-20 | 平安科技(深圳)有限公司 | 大样本研报信息提取方法、装置、设备及存储介质 |
| CN110889291B (zh) * | 2019-11-21 | 2023-06-27 | 山东浪潮科学研究院有限公司 | 一种基于深度学习和区块链的研报评价方法 |
| CN112257435B (zh) * | 2020-09-11 | 2024-05-14 | 安徽中科新辰技术有限公司 | 一种基于人工智能分词技术的警情数据治理方法 |
| CN113553829B (zh) * | 2021-07-19 | 2024-12-20 | 中国工商银行股份有限公司 | 一种证券研究报告的观点识别方法及装置 |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106844330A (zh) * | 2016-11-15 | 2017-06-13 | 平安科技(深圳)有限公司 | 文章情感的分析方法和装置 |
| CN107844558A (zh) * | 2017-10-31 | 2018-03-27 | 金蝶软件(中国)有限公司 | 一种分类信息的确定方法以及相关装置 |
| CN108062300A (zh) * | 2016-11-08 | 2018-05-22 | 中移(苏州)软件技术有限公司 | 一种基于中文文本进行情感倾向分析的方法及装置 |
| CN108108355A (zh) * | 2017-12-25 | 2018-06-01 | 北京牡丹电子集团有限责任公司数字电视技术中心 | 基于深度学习的文本情感分析方法和系统 |
Family Cites Families (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090043637A1 (en) * | 2004-06-01 | 2009-02-12 | Eder Jeffrey Scott | Extended value and risk management system |
| CN101667194A (zh) * | 2009-09-29 | 2010-03-10 | 北京大学 | 基于用户评论文本特征的自动摘要方法及其自动摘要系统 |
| WO2012121747A1 (en) * | 2011-03-04 | 2012-09-13 | Ultratick, Inc. | Predicting the performance of a financial instrument |
| CN106407236B (zh) * | 2015-08-03 | 2019-07-23 | 北京众荟信息技术股份有限公司 | 一种面向点评数据的情感倾向性检测方法 |
| CN106384166A (zh) * | 2016-09-12 | 2017-02-08 | 中山大学 | 一种结合财经新闻的深度学习股市预测方法 |
| CN106776581B (zh) * | 2017-02-21 | 2020-01-24 | 浙江工商大学 | 基于深度学习的主观性文本情感分析方法 |
| CN108256098B (zh) * | 2018-01-30 | 2022-02-15 | 中国银联股份有限公司 | 一种确定用户评论情感倾向的方法及装置 |
| CN108647822A (zh) * | 2018-05-10 | 2018-10-12 | 平安科技(深圳)有限公司 | 电子装置、基于研报数据的预测方法和计算机存储介质 |
| CN108647823A (zh) * | 2018-05-10 | 2018-10-12 | 北京航空航天大学 | 基于深度学习的股票数据分析方法和装置 |
-
2018
- 2018-10-22 CN CN201811228761.7A patent/CN109284504A/zh active Pending
-
2019
- 2019-01-03 WO PCT/CN2019/070287 patent/WO2020082609A1/zh not_active Ceased
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108062300A (zh) * | 2016-11-08 | 2018-05-22 | 中移(苏州)软件技术有限公司 | 一种基于中文文本进行情感倾向分析的方法及装置 |
| CN106844330A (zh) * | 2016-11-15 | 2017-06-13 | 平安科技(深圳)有限公司 | 文章情感的分析方法和装置 |
| CN107844558A (zh) * | 2017-10-31 | 2018-03-27 | 金蝶软件(中国)有限公司 | 一种分类信息的确定方法以及相关装置 |
| CN108108355A (zh) * | 2017-12-25 | 2018-06-01 | 北京牡丹电子集团有限责任公司数字电视技术中心 | 基于深度学习的文本情感分析方法和系统 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109284504A (zh) | 2019-01-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Huq et al. | Sentiment analysis on Twitter data using KNN and SVM | |
| CN104050256B (zh) | 基于主动学习的问答方法及采用该方法的问答系统 | |
| Sehgal et al. | Sops: stock prediction using web sentiment | |
| CN113051356A (zh) | 开放关系抽取方法、装置、电子设备及存储介质 | |
| JP6150291B2 (ja) | 矛盾表現収集装置及びそのためのコンピュータプログラム | |
| CN116863495B (zh) | 意图识别方法、装置、电子设备、存储介质 | |
| CN112052682A (zh) | 事件实体联合抽取方法、装置、计算机设备及存储介质 | |
| CN107122416A (zh) | 一种中文事件抽取方法 | |
| CN111090735B (zh) | 一种基于知识图谱的智能问答方法的性能评价方法 | |
| CN109933796A (zh) | 一种公告文本关键信息提取方法及设备 | |
| CN113570380A (zh) | 基于语义分析的业务投诉处理方法、装置、设备及计算机可读存储介质 | |
| CN112163424A (zh) | 数据的标注方法、装置、设备和介质 | |
| WO2020082609A1 (zh) | 利用深度学习模型的证券研报分析方法及装置 | |
| CN114372122A (zh) | 信息的获取方法、计算设备及存储介质 | |
| CN109389418A (zh) | 基于lda模型的供电服务客户诉求识别方法 | |
| CN111198935A (zh) | 模型处理方法、装置、存储介质及电子设备 | |
| CN117807482B (zh) | 海关报关单的分类方法、装置、设备及存储介质 | |
| WO2020082613A1 (zh) | 利用深度学习模型的证券研报核心观点提取方法及装置 | |
| CN112232681A (zh) | 一种面向计算分析类非选择题的智能评卷方法 | |
| WO2019047352A1 (zh) | 基于社交数据的资产配置方法、电子装置及介质 | |
| CN114048325A (zh) | 基于知识图谱的案件处理方法及装置、存储介质 | |
| CN119069138A (zh) | 一种基于大语言模型生成医学领域多样化指令数据的方法 | |
| CN115081441A (zh) | 运维指令意图识别方法、装置、电子设备和存储介质 | |
| CN119739838A (zh) | 多标签生成匹配的rag智能问答方法、装置、设备及介质 | |
| Sonia et al. | Stock price analysis using sentiment analysis of twitter data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| NENP | Non-entry into the national phase |
Ref country code: DE |
|
| 122 | Ep: pct application non-entry in european phase |
Ref document number: 19876224 Country of ref document: EP Kind code of ref document: A1 |