CN110147550A - Pronunciation character fusion method neural network based - Google Patents
Pronunciation character fusion method neural network based Download PDFInfo
- Publication number
- CN110147550A CN110147550A CN201910327655.2A CN201910327655A CN110147550A CN 110147550 A CN110147550 A CN 110147550A CN 201910327655 A CN201910327655 A CN 201910327655A CN 110147550 A CN110147550 A CN 110147550A
- Authority
- CN
- China
- Prior art keywords
- text
- data
- pronunciation
- neural network
- vector
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/205—Parsing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/232—Orthographic correction, e.g. spell checking or vowelisation
-
- 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
- G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- General Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Data Mining & Analysis (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Machine Translation (AREA)
- Electrically Operated Instructional Devices (AREA)
Abstract
Present invention discloses a kind of pronunciation character fusion methods neural network based, include the following steps, S1, dataset acquisition step, obtain text data;S2, data prediction step, pre-process text data, remove the noise for including in text data, text data is then converted to phonetic symbol or phonetic text;S3, text vector step, the text term vector and pronunciation vector of training text data, and be vectorization data by text data and corresponding phonetic symbol or phonetic text conversion;Vectorization data are input in two neural networks the sequence information for being trained, distinguishing learning text, finally merge the text part of speech feature information learnt and pronunciation character information by S4, neural metwork training and fusion steps.By the present invention in that being enriched the expression of text with the pronunciation character of text and the mode that blends it with text part of speech feature, being improved the result of natural language processing task.
Description
Technical field
The present invention relates to a kind of Feature fusions, and in particular to a kind of pronunciation character fusion side neural network based
Method belongs to natural language processing field.
Background technique
With flourishing for internet, especially in recent years, largely such as microblogging, wechat, the social media of electric business
Appearance, internet is by leaps and bounds to have strided forward the Web2.0 epoch, its own has also been transformed to " interaction from " read formula internet "
Formula internet ".Netizens have been not only the channel for being regarded as obtaining information to the cognition of internet, even more as biography
The platform broadcast information, share oneself viewpoint and emotion.Daily, hundreds of millions of netizens can issue the letter of simultaneously spread index grade
Breath, in the information of these magnanimity, the overwhelming majority expresses the viewpoint and hobby of publisher itself.In nowadays this big number
According to epoch, these viewpoints are all undoubtedly extremely valuable wealth, wherein include different people to different society field and
The subjectivity view of phenomenon.Based on this, either businessman or social organization are again or individual increasingly payes attention on network
Text information is carried out analysis by the comment of user and has become one particularly important development of natural language processing field
Direction.
In current most of text analyzing model, the input of model is only the semantic meaning representation of word, they ignore
The bulk information for including in text, can not accurately indicate urtext, therefore single features are only used only and carry out nature language
The effect is unsatisfactory for speech processing.And in language, pronunciation often also contains the emotion information of text.If it is possible to
Pronunciation character is collected processing, forms abundant in content Deep Semantics feature, then natural language processing will be improved
Effect.
In conclusion a kind of fusion method of pronunciation character how is proposed on the basis of existing technology, thus fully
Using text information, also just become the common goal in research of those skilled in that art.
Summary of the invention
In view of the prior art, there are drawbacks described above, and the purpose of the present invention is to propose to a kind of pronunciation characters neural network based
Fusion method includes the following steps:
S1, dataset acquisition step obtain text data;
S2, data prediction step, pre-process text data collected, remove the noise for including in text data,
Obtained text data is then converted into phonetic symbol text or phonetic text;
S3, text vector step, the text term vector and pronunciation vector of training text data, and will be acquired in previous step
Text data and corresponding phonetic symbol text or phonetic text conversion be vectorization data;
Vectorization data obtained in previous step are input to two neural networks by S4, neural metwork training and fusion steps
In be trained, respectively learning text sequence information, finally by the text part of speech feature information and pronunciation character that learns letter
Breath is merged.
Preferably, dataset acquisition step described in S1, specifically includes: downloading publicly-owned data set or crawls network using crawler
Text is formed multiple data sets based on the data downloaded or crawled, and summarizes whole set of data, forms textual data
According to.
Preferably, data prediction step described in S2, specifically includes:
S21, pending data collection is chosen from text data and executes subsequent step;
S22, the text data for judging that pending data is concentrated then are held for Chinese text or English text if Chinese text
Row S23 is then jumped if English data and is executed S26;
S23, noise removal process is carried out to the Chinese text in data set to be processed, the noise removal process includes format mark
Standardization, removal additional character and punctuation mark;
S24, the Chinese text that noise removal process is completed is segmented and stop words is gone to handle;
S25, pinyin marking is carried out to the Chinese text that participle is completed and stop words is gone to handle, obtains the spelling of urtext data
Sound text;
S26, noise removal process is carried out to the English text in data set to be processed, the noise removal process includes format mark
Standardization, removal additional character and punctuation mark;
S27, speech reduction processing is carried out to the English text that noise removal process is completed;
S28, phonetic symbol mark is carried out to the English text that speech reduction processing is completed, obtains the phonetic symbol text of urtext data
This.
Preferably, when carrying out word segmentation processing described in S24, participle tool is stammerer participle, SnowNLP or THULAC;
When carrying out going stop words to handle described in S24, deactivating vocabulary is that Harbin Institute of Technology deactivates vocabulary, Baidu deactivates vocabulary or Sichuan University
Machine intelligence laboratory deactivates dictionary.
Preferably, when carrying out the mark of phonetic symbol described in S28, matching method mark is carried out using english dictionary or uses crawler
It is labeled using network English dictionary.
Preferably, text vector step described in S3, specifically includes:
Pretreated text data is passed through in S31, selection;
S32, training text term vector and pronunciation vector;
S33, using text term vector obtained in S32 and pronunciation vector, by selected in S31 by pretreated text
Data and corresponding phonetic symbol text or phonetic text conversion are the data of matrix form, and the data of the matrix form include text
Term vector matrix and pronunciation vector matrix.
Preferably, text term vector described in S32 is the text term vector resource shared on network, or uses the big of downloading
Scale corpus cooperates experimental data set, utilizes the resulting text term vector of Word2Vec or GloVe training;It is sent out described in S32
Sound vector is trained using Word2Vec or GloVe;The text term vector is identical as the dimension of both pronunciation vectors.
Preferably, neural metwork training and fusion steps described in S4, specifically includes:
S41, the text term vector matrix for reading text data, and by the text term vector Input matrix read to neural network
In be trained;
S42, read text data pronunciation vector matrix, and by the pronunciation vector matrix read be input in neural network into
Row training;
S43, to passing through two neural metwork trainings obtained text part of speech feature information and pronunciation character information in S41 and S42
Carry out fusion treatment.
Preferably, neural network used in S41 and S42 is RNN, CNN, LSTM or Bi-LSTM;Institute in S41 and S42
The parameter of both neural networks used is identical.
Preferably, fusion treatment described in S43, amalgamation mode are splicing, are added or averagely merge.
Compared with prior art, advantages of the present invention is mainly reflected in the following aspects:
Pronunciation character fusion method neural network based of the invention, by using the pronunciation character of text and by itself and text
The mode that part of speech feature blends enriches the expression of text, improves including text classification and natural language analysis
The result of a variety of natural language processing tasks.
Meanwhile the present invention also provides reference for other relevant issues in same domain, can be opened up on this basis
Extension is stretched, and is applied in other related art schemes in terms of natural language processing, has very wide application prospect.
Just attached drawing in conjunction with the embodiments below, the embodiment of the present invention is described in further detail, so that of the invention
Technical solution is more readily understood, grasps.
Detailed description of the invention
Fig. 1 is the overall flow schematic diagram of the method for the present invention;
Fig. 2 is the flow diagram of data prediction step in the present invention;
Fig. 3 is the flow diagram of text vectorization step in the present invention;
Fig. 4 is the flow diagram of neural metwork training and fusion steps in the present invention.
Specific embodiment
The present invention is directed in current most of text analyzing models and inputs only using only part of speech feature, has ignored in text
This problem of the other information for including, proposes a kind of pronunciation character fusion method neural network based, and the present invention is using
It joined pronunciation character while semantic term vector and indicated with vector, take full advantage of the information of text, form fusion
There is the Deep Semantics feature of text term vector and the vector that pronounces.Just method of the invention is illustrated in conjunction with attached drawing below.
As shown in Figure 1, a kind of pronunciation character fusion method neural network based, includes the following steps:
S1, dataset acquisition step obtain text data.
S2, data prediction step, pre-process text data collected, and include in removal text data makes an uproar
Obtained text data is then converted to phonetic symbol text or phonetic text by sound.
S3, text vector step, the text term vectors of training text data and pronunciation vector, and by institute in previous step
Obtained text data and corresponding phonetic symbol text or phonetic text conversion is vectorization data.
Vectorization data obtained in previous step are input to two nerves by S4, neural metwork training and fusion steps
It is trained in network, the sequence information of difference learning text, it is finally that the text part of speech feature information learnt and pronunciation is special
Reference breath is merged.
Dataset acquisition step described in S1, specifically includes: it downloads publicly-owned data set or crawls network text using crawler, with
Multiple data sets are formed based on the data downloaded or crawled, and summarize whole set of data, form text data.
As shown in Fig. 2, data prediction step described in S2, specifically includes:
S21, pending data collection is chosen from text data and executes subsequent step.
S22, judge text data that pending data is concentrated for Chinese text or English text, if Chinese text,
Then execute S23;If English data, then jumps and execute S26.
S23, noise removal process is carried out to the Chinese text in data set to be processed, the noise removal process includes lattice
Formula standardization, removal additional character and punctuation mark etc..
S24, the Chinese text that noise removal process is completed is segmented and stop words is gone to handle.Carrying out described point
When word processing, participle tool is stammerer participle, SnowNLP or THULAC etc.;When going stop words to handle described in the progress, stop words
Table is that Harbin Institute of Technology deactivates vocabulary, Baidu deactivates vocabulary or Sichuan University's machine intelligence laboratory deactivates dictionary etc..
S25, pinyin marking is carried out to the Chinese text that participle is completed and stop words is gone to handle, obtains urtext data
Phonetic text, the step for can be realized by the library the Python such as xpinyin, pypinyin.
S26, noise removal process is carried out to the English text in data set to be processed, the noise removal process includes lattice
Formula standardization, removal additional character and punctuation mark etc..
S27, speech reduction processing is carried out to the English text that noise removal process is completed, it is right in subsequent step to facilitate
The phonetic symbol of English text marks, and the speech reduction processing can be realized by the WordNet dictionary in the library NLTK in Python.
S28, phonetic symbol mark is carried out to the English text that speech reduction processing is completed, obtains the phonetic symbol of urtext data
Text.When carrying out phonetic symbol mark, it english dictionary can be used to carry out matching method mark or utilize network English using crawler
Dictionary is labeled.
As shown in figure 3, text vector step described in S3, specifically includes:
Pretreated text data is passed through in S31, selection.
S32, training text term vector and pronunciation vector.
The text term vector is the text term vector resource shared on network, Chinese with good grounds Wiki corpus training
Text term vector, the English Glove text term vector series for having Stamford training;Or matched using the large-scale corpus collection of downloading
It closes experimental data set, utilize the resulting text term vector of Word2Vec or GloVe training;The pronunciation vector uses Word2Vec
Or GloVe is trained.The text term vector is identical as the dimension of both pronunciation vectors.
S33, using text term vector obtained in S32 and pronunciation vector, by selected in S31 by pretreated
Text data and corresponding phonetic symbol text or phonetic text conversion are the data of matrix form, and the data of the matrix form include
Text term vector matrix and pronunciation vector matrix.
As shown in figure 4, neural metwork training and fusion steps described in S4, specifically include:
S41, the text term vector matrix for reading text data, and by the text term vector Input matrix read to neural network
In be trained.
S42, the pronunciation vector matrix for reading text data, and the pronunciation vector matrix read is input to neural network
In be trained.
It should be noted that neural network used in S41 and S42 is RNN, CNN, LSTM or Bi-LSTM etc.;S41
And the parameter of both neural networks used in S42 is identical.
S43, to passing through the obtained text part of speech feature information of two neural metwork trainings and pronunciation character in S41 and S42
Information carries out fusion treatment.Fusion treatment described herein, amalgamation mode are splicing, addition or average fusion etc..
In conclusion pronunciation character fusion method neural network based of the invention, special by using the pronunciation of text
Levy and by the mode that it is blended with text part of speech feature, enrich the expression of text, improve including text classification and nature
The result of a variety of natural language processing tasks including language analysis.
Meanwhile the present invention also provides reference for other relevant issues in same domain, can be opened up on this basis
Extension is stretched, and is applied in other related art schemes in terms of natural language processing, has very wide application prospect.
It is obvious to a person skilled in the art that invention is not limited to the details of the above exemplary embodiments, Er Qie
In the case where without departing substantially from spirit and essential characteristics of the invention, the present invention can be realized in other specific forms.Therefore, no matter
From the point of view of which point, the present embodiments are to be considered as illustrative and not restrictive, and the scope of the present invention is by appended power
Benefit requires rather than above description limits, it is intended that all by what is fallen within the meaning and scope of the equivalent elements of the claims
Variation is included within the present invention, and any reference signs in the claims should not be construed as limiting the involved claims.
In addition, it should be understood that although this specification is described in terms of embodiments, but not each embodiment is only wrapped
Containing an independent technical solution, this description of the specification is merely for the sake of clarity, and those skilled in the art should
It considers the specification as a whole, the technical solutions in the various embodiments may also be suitably combined, forms those skilled in the art
The other embodiments being understood that.
Claims (10)
1. a kind of pronunciation character fusion method neural network based, which comprises the steps of:
S1, dataset acquisition step obtain text data;
S2, data prediction step, pre-process text data collected, remove the noise for including in text data,
Obtained text data is then converted into phonetic symbol text or phonetic text;
S3, text vector step, the text term vector and pronunciation vector of training text data, and will be acquired in previous step
Text data and corresponding phonetic symbol text or phonetic text conversion be vectorization data;
Vectorization data obtained in previous step are input to two neural networks by S4, neural metwork training and fusion steps
In be trained, respectively learning text sequence information, finally by the text part of speech feature information and pronunciation character that learns letter
Breath is merged.
2. pronunciation character fusion method neural network based according to claim 1, which is characterized in that data described in S1
Collect acquisition step, specifically includes: downloading publicly-owned data set or crawl network text using crawler, with the number downloaded or crawled
Multiple data sets are formed based on, and summarize whole set of data, form text data.
3. pronunciation character fusion method neural network based according to claim 2, which is characterized in that data described in S2
Pre-treatment step specifically includes:
S21, pending data collection is chosen from text data and executes subsequent step;
S22, the text data for judging that pending data is concentrated then are held for Chinese text or English text if Chinese text
Row S23 is then jumped if English data and is executed S26;
S23, noise removal process is carried out to the Chinese text in data set to be processed, the noise removal process includes format mark
Standardization, removal additional character and punctuation mark;
S24, the Chinese text that noise removal process is completed is segmented and stop words is gone to handle;
S25, pinyin marking is carried out to the Chinese text that participle is completed and stop words is gone to handle, obtains the spelling of urtext data
Sound text;
S26, noise removal process is carried out to the English text in data set to be processed, the noise removal process includes format mark
Standardization, removal additional character and punctuation mark;
S27, speech reduction processing is carried out to the English text that noise removal process is completed;
S28, phonetic symbol mark is carried out to the English text that speech reduction processing is completed, obtains the phonetic symbol text of urtext data
This.
4. pronunciation character fusion method neural network based according to claim 3, it is characterised in that: carrying out S24
Described in word segmentation processing when, participle tool is stammerer participle, SnowNLP or THULAC;Gone at stop words described in S24
When reason, deactivating vocabulary is that Harbin Institute of Technology deactivates vocabulary, Baidu deactivates vocabulary or Sichuan University's machine intelligence laboratory deactivates dictionary.
5. pronunciation character fusion method neural network based according to claim 3, it is characterised in that: carrying out S28
Described in phonetic symbol mark when, carry out matching method mark using english dictionary or marked using network English dictionary using crawler
Note.
6. pronunciation character fusion method neural network based according to claim 3, which is characterized in that text described in S3
Vectorization step, specifically includes:
Pretreated text data is passed through in S31, selection;
S32, training text term vector and pronunciation vector;
S33, using text term vector obtained in S32 and pronunciation vector, by selected in S31 by pretreated text
Data and corresponding phonetic symbol text or phonetic text conversion are the data of matrix form, and the data of the matrix form include text
Term vector matrix and pronunciation vector matrix.
7. pronunciation character fusion method neural network based according to claim 6, it is characterised in that: described in S32
Text term vector is the text term vector resource shared on network, or cooperates experimental data using the large-scale corpus collection of downloading
Collection utilizes the resulting text term vector of Word2Vec or GloVe training;Pronunciation vector described in S32 using Word2Vec or
GloVe is trained;The text term vector is identical as the dimension of both pronunciation vectors.
8. pronunciation character fusion method neural network based according to claim 6, which is characterized in that nerve described in S4
Network training and fusion steps, specifically include:
S41, the text term vector matrix for reading text data, and by the text term vector Input matrix read to neural network
In be trained;
S42, read text data pronunciation vector matrix, and by the pronunciation vector matrix read be input in neural network into
Row training;
S43, to passing through two neural metwork trainings obtained text part of speech feature information and pronunciation character information in S41 and S42
Carry out fusion treatment.
9. pronunciation character fusion method neural network based according to claim 8, it is characterised in that: in S41 and S42
Used neural network is RNN, CNN, LSTM or Bi-LSTM;The parameter of both neural networks used in S41 and S42
It is identical.
10. pronunciation character fusion method neural network based according to claim 8, which is characterized in that described in S43
Fusion treatment, amalgamation mode are splicing, are added or averagely merge.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910327655.2A CN110147550A (en) | 2019-04-23 | 2019-04-23 | Pronunciation character fusion method neural network based |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910327655.2A CN110147550A (en) | 2019-04-23 | 2019-04-23 | Pronunciation character fusion method neural network based |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110147550A true CN110147550A (en) | 2019-08-20 |
Family
ID=67593857
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910327655.2A Pending CN110147550A (en) | 2019-04-23 | 2019-04-23 | Pronunciation character fusion method neural network based |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110147550A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111078887A (en) * | 2019-12-20 | 2020-04-28 | 厦门市美亚柏科信息股份有限公司 | Text classification method and device |
CN111191463A (en) * | 2019-12-30 | 2020-05-22 | 杭州远传新业科技有限公司 | Emotion analysis method and device, electronic equipment and storage medium |
CN112652291A (en) * | 2020-12-15 | 2021-04-13 | 携程旅游网络技术(上海)有限公司 | Speech synthesis method, system, device and storage medium based on neural network |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108874765A (en) * | 2017-05-15 | 2018-11-23 | 阿里巴巴集团控股有限公司 | Term vector processing method and processing device |
US20180342233A1 (en) * | 2017-05-23 | 2018-11-29 | Beijing Baidu Netcom Science And Technology Co., Ltd. | Method and apparatus for correcting speech recognition error based on artificial intelligence, and storage medium |
-
2019
- 2019-04-23 CN CN201910327655.2A patent/CN110147550A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108874765A (en) * | 2017-05-15 | 2018-11-23 | 阿里巴巴集团控股有限公司 | Term vector processing method and processing device |
US20180342233A1 (en) * | 2017-05-23 | 2018-11-29 | Beijing Baidu Netcom Science And Technology Co., Ltd. | Method and apparatus for correcting speech recognition error based on artificial intelligence, and storage medium |
Non-Patent Citations (2)
Title |
---|
WENHAO ZHU ET AL.: "Improve word embedding using both writing and pronunciation", 《PLOS ONE》 * |
余本功 等: "基于CP-CNN的中文短文本分类研究", 《计算机应用研究》 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111078887A (en) * | 2019-12-20 | 2020-04-28 | 厦门市美亚柏科信息股份有限公司 | Text classification method and device |
CN111078887B (en) * | 2019-12-20 | 2022-04-29 | 厦门市美亚柏科信息股份有限公司 | Text classification method and device |
CN111191463A (en) * | 2019-12-30 | 2020-05-22 | 杭州远传新业科技有限公司 | Emotion analysis method and device, electronic equipment and storage medium |
CN112652291A (en) * | 2020-12-15 | 2021-04-13 | 携程旅游网络技术(上海)有限公司 | Speech synthesis method, system, device and storage medium based on neural network |
CN112652291B (en) * | 2020-12-15 | 2024-04-05 | 携程旅游网络技术(上海)有限公司 | Speech synthesis method, system, equipment and storage medium based on neural network |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Abdullah et al. | SEDAT: sentiment and emotion detection in Arabic text using CNN-LSTM deep learning | |
Deng et al. | Deep learning in natural language processing | |
CN108334891B (en) | Task type intention classification method and device | |
CN108984530A (en) | A kind of detection method and detection system of network sensitive content | |
CN106096664B (en) | A kind of sentiment analysis method based on social network data | |
CN110110054A (en) | A method of obtaining question and answer pair in the slave non-structured text based on deep learning | |
CN109948158A (en) | Emotional orientation analytical method based on environment member insertion and deep learning | |
CN110147550A (en) | Pronunciation character fusion method neural network based | |
CN111797898A (en) | Online comment automatic reply method based on deep semantic matching | |
CN107102980A (en) | The extracting method and device of emotion information | |
Zhang et al. | Effective subword segmentation for text comprehension | |
Savelieva et al. | Abstractive summarization of spoken and written instructions with BERT | |
CN109325120A (en) | A kind of text sentiment classification method separating user and product attention mechanism | |
Mika et al. | Learning to tag and tagging to learn: A case study on wikipedia | |
CN115294427A (en) | Stylized image description generation method based on transfer learning | |
CN111339772B (en) | Russian text emotion analysis method, electronic device and storage medium | |
Sanaullah et al. | A real-time automatic translation of text to sign language. | |
CN112711666B (en) | Futures label extraction method and device | |
Roque | Towards a computational approach to literary text analysis | |
Walha et al. | A Lexicon approach to multidimensional analysis of tweets opinion | |
CN106776568A (en) | Based on the rationale for the recommendation generation method that user evaluates | |
Karpagam et al. | Deep learning approaches for answer selection in question answering system for conversation agents | |
Tarnpradab et al. | Improving online forums summarization via hierarchical unified deep neural network | |
Wei et al. | Comparative studies of AIML | |
CN113704472A (en) | Hate and offensive statement identification method and system based on topic memory network |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190820 |
|
RJ01 | Rejection of invention patent application after publication |