CN104199813B - Pseudo-feedback-based personalized machine translation system and method - Google Patents
Pseudo-feedback-based personalized machine translation system and method Download PDFInfo
- Publication number
- CN104199813B CN104199813B CN201410491100.9A CN201410491100A CN104199813B CN 104199813 B CN104199813 B CN 104199813B CN 201410491100 A CN201410491100 A CN 201410491100A CN 104199813 B CN104199813 B CN 104199813B
- Authority
- CN
- China
- Prior art keywords
- translation
- module
- preliminary
- postedit
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Landscapes
- Machine Translation (AREA)
Abstract
The invention relates to a pseudo-feedback-based personalized machine translation system and method. The existing traditional machine translation methods are unavailable for the obtaining of high-quality personalized translation systems, and the various translation demands of users cannot be met. The pseudo-feedback-based personalized machine translation system comprises a phrase table filter module, an input module, an initial translation module, a pseudo-feedback search module, a phrase table sorting module and a decoder module. The pseudo-feedback-based personalized machine translation method includes the steps: an inputting step, a user inputs a translation task S; an initial translation step, an initial machine translation result T' of the translation task is obtained with the initial translation module; a pseudo-feedback search step, the pseudo-feedback search module is used to search to obtain initial translation results and standard translations R of similar translation instances; a phrase table sorting step, a trained universal post-editing model is turned into a personalized post-editing model, and filtering is performed to obtain an optimized post-editing model; a decoder module decoding step, the optimized personalized post-editing model is used to decode the initial machine translation result T' of the translation task so as to obtain an optimal final translation result. The pseudo-feedback-based personalized machine translation system and method is applicable to the field of machine translation.
Description
Technical field
The present invention relates to a kind of personalization machine translation system and method, belong to machine translation field.
Background technology
Recently as developing rapidly for machine translation mothod, quality of its translation has had carrying largely
Rise, the obstacle that current some general translation on line services have been able to help people to break through language goes to read and understand that some are normal
Across the text of language.But further the quality of hoisting machine translation but encounters very big difficulty.On the one hand, because existing
Some statistical machine translation technology major defects are, if personalized translation is completed, it is necessary to substantial amounts of field feedback, and
Statistics training modeling is carried out in these data, the Machine Translation Model of property one by one is realized.And the use needed for these training
The acquisition of family feedback information is very difficult, and existing method cannot effectively utilize these feedback informations, so as to cannot obtain
Obtain high-quality personalized translation system.Although field feedback can be utilized by traditional postedit, due to can
It is less using user data, cause the advantage for counting postedit model to be difficult to bring into play.On the other hand, traditional machine translation
The optimization aim of method is normally based on open field, without being aimed at what specific translation duties were carried out.Although depositing
In the research for domain-adaptive problem, but still fall within for professional colony, and face extensive various machine translation and use
Family is especially for internet online user, it is impossible to meet the various translation demands of user.So further improving machine translation
Quality is that we want a technical problem urgently to be resolved hurrily.
The content of the invention
High-quality personalized translation system cannot be obtained the invention aims to solve traditional machine translation method
System, causes to meet the problem of the various translation demands of user, and propose it is a kind of can improve mechanical translation quality based on puppet
The personalization machine translation system and interpretation method of feedback.
A kind of personalization machine translation system based on pseudo- feedback, the translation system includes：
For the phrase table filtering module that each the general postedit model phrase table to development set data is filtered；
Input module for obtaining the translation duties S of user input；
Preliminary machine translation result for being obtained translation duties to being translated after user input translation duties S
T ', the source language sentence in the translation instance storehouse provided local system translate and obtains the preliminary translation of the sentence T's of translation instance
Preliminary translation module；
For in the translation instance storehouse of local system word alignment form, retrieval to obtain the preliminary translation of similar translation example
The pseudo- feedback searching module of result and standard translation translation R；
Classified for the phrase table to the postedit model after training and obtained the phrase of personalized postedit model
Table sort module；
The preliminary translation result of the similar translation example for being obtained to the retrieval of pseudo- feedback searching module is decoded, and is obtained
To the decoder module of final translation result.
A kind of personalization machine based on pseudo- feedback translates the interpretation method of translation system, in user input translation duties S
Before, train general using statistical method using the translation instance preliminary translation of the sentence T and standard translation translation R in translation memory
Postedit model, completes the training process of general postedit model；The personalization machine interpretation method passes through following steps reality
It is existing：
Step one, phrase table filtering mold process：Using phrase table filtering module to development set data each it is general after compile
Model phrase table is collected to be filtered；
Based on the result after filtering using default-weight to each sentence D in development set dataiDecoded, produced n-
Best translation results；Then, n-best translation results are combined；Finally, using MERT instruments to the n-best after combination
Translation result integrally adjusts ginseng, additionally it is possible to realize characteristic parameter optimization process；
Step 2, input process：Be input into translation duties S to input module by user；
Step 3, preliminary translation process：The preliminary translation process is defeated with user before including user input translation duties S
Enter two parts after translation duties S；
Before user input translation duties S, the transcription platform built using the machine translation system of local system incite somebody to action this
The source language sentence in the translation instance storehouse that ground system is provided tentatively is translated, and obtains the preliminary translation of the sentence T of translation instance；
Meanwhile, after the translation duties S by input module acquisition user input, obtained using the translation of preliminary translation module
The preliminary machine translation result T ' of translation duties；
Step 4, pseudo- feedback searching process：According to the preliminary translation of the sentence T of translation instance obtained in step 3, local
In the translation instance storehouse of word alignment form, the inspection of cosine similarity is carried out with original language bag of words using pseudo- feedback searching module
Rope, obtains the preliminary translation result and standard translation translation R of similar translation example, and from the preliminary translation knot of similar translation instance
Most like preceding 900-1100 is selected in the retrieval result of fruit and standard translation translation R；
Wherein, the cosine similarity CS is calculated according to the vector space model with original language bag of words as unit, institute
The computational methods for stating cosine similarity CS are：
Wherein, Vec (Sexample) it is the source language sentence subvector of translation instance, Vec (Sinput) it is translation duties vector,
Vec(Sinput)·Vec(Sexample) it is two inner products of vector, | | | | it is the norm of vector；
Step 5, phrase table assorting process：According to the most like preceding 900-1100 similar translation reality that step 4 is selected
The preliminary translation result and standard translation translation R of example, using phrase table sort module by training after general postedit model
Phrase table is categorized as contributing to the positive phrase for lifting translation quality and the passive phrase for incorporating final translation result noise, makes
General postedit model after training becomes personalized postedit model, then by the positive phrase in personalized postedit model and
The preliminary translation result and standard translation of the similar translation example that passive phrase goes out with pseudo- feedback searching procedural retrieval in step 4
Translation R contrast, the passive phrase is filtered out from personalized postedit model phrase table, so as to obtain one optimization
Property postedit model；
Step 6, decoder module decoding process：Using in step 5 optimize personalized postedit model as translation mould
Type, the preliminary machine of the translation duties obtained to step 3 using traditional machine translation coding/decoding method using decoder module is turned over
Translate result T ' to be decoded, obtain the final translation result of goodization.
Beneficial effects of the present invention are：The present invention is to carrying out similar turning in translation instance storehouse using pseudo- feedback searching module
Translate example to be retrieved, then general postedit phrase is classified by phrase table sort module, filter out passive postedit
Phrase, selects postedit rule and obtains the personalized postedit model for optimizing, so that the quality of hoisting machine translation.In addition,
Application characteristic parameter optimization process during by building in preliminary translation process postedit model, and in characteristic parameter optimization process
In, for the development set data for giving, the carrying out to being input into decodes respectively, then carries out overall tune and joins, with effectively optimization ginseng
Number, the benefit of lifting system performance.Particularly, retrieval is being concentrated in local translation instance database data using pseudo- feedback searching module
During, obtain the parallel sentence similar with the preliminary translation result of sentence to be translated obtained by user input to come replace feedback
Information, is difficult to obtain field feedback this problem so as to solve.
In addition, the inventive method make use of feedback information well, effectively postedit is set up on initial translation model
Model, translation result and the translation knot of Google that personalization machine translation system and method based on pseudo- feedback of the invention are obtained
Fruit is contrasted, and its translation quality improves 19.5%；The translation result of the machine translation system trained with Moses instruments enters
Row contrast, its translation quality improves 14.1%
Brief description of the drawings
Fig. 1 is translation flow schematic diagram of the invention.
Specific embodiment
Specific embodiment one：
The personalization machine translation system based on pseudo- feedback of present embodiment, the translation system includes：
For the phrase table filtering module that each the general postedit model phrase table to development set data is filtered；
Input module for obtaining the translation duties S of user input；
Preliminary machine translation result for being obtained translation duties to being translated after user input translation duties S
T ', the source language sentence in the translation instance storehouse provided local system translate and obtains the preliminary translation of the sentence T's of translation instance
Preliminary translation module；
For in the translation instance storehouse of local system word alignment form, retrieval to obtain the preliminary translation of similar translation example
The pseudo- feedback searching module of result and standard translation translation R；
Classified for the phrase table to the postedit model after training and obtained the phrase of personalized postedit model
Table sort module；
The preliminary translation result of the similar translation example for being obtained to the retrieval of pseudo- feedback searching module is decoded, and is obtained
To the decoder module of final translation result.
Specific embodiment two：
From unlike specific embodiment one, the personalization machine based on pseudo- feedback translates system described in present embodiment
System, the phrase table filtering module is contained in the phrase table sort module.
Specific embodiment three：
The interpretation method of the personalization machine translation system based on pseudo- feedback of present embodiment, appoints in user input translation
Before business S, trained using statistical method using the translation instance preliminary translation of the sentence T and standard translation translation R in translation memory
General postedit model, completes the training process of general postedit model；The personalization machine interpretation method is by following step
It is rapid to realize：
Step one, phrase table filtering mold process：Using phrase table filtering module to development set data each it is general after compile
Model phrase table is collected to be filtered；
Based on the result after filtering using default-weight to each sentence D in development set dataiDecoded, produced n-
Best translation results；Then, n-best translation results are combined；Finally, using MERT instruments to the n-best after combination
Translation result integrally adjusts ginseng, additionally it is possible to realize characteristic parameter optimization process；
Step 2, input process：Be input into translation duties S to input module by user；
Step 3, preliminary translation process：The preliminary translation process is defeated with user before including user input translation duties S
Enter two parts after translation duties S；
Before user input translation duties S, the transcription platform built using the machine translation system of local system incite somebody to action this
The source language sentence in the translation instance storehouse that ground system is provided tentatively is translated, and obtains the preliminary translation of the sentence T of translation instance；
Meanwhile, after the translation duties S by input module acquisition user input, obtained using the translation of preliminary translation module
The preliminary machine translation result T ' of translation duties；
Step 4, pseudo- feedback searching process：According to the preliminary translation of the sentence T of translation instance obtained in step 3, local
In the translation instance storehouse of word alignment form, the inspection of cosine similarity is carried out with original language bag of words using pseudo- feedback searching module
Rope, obtains the preliminary translation result and standard translation translation R of similar translation example, and from the preliminary translation knot of similar translation instance
Most like preceding 900-1100 is selected in the retrieval result of fruit and standard translation translation R；
Wherein, the cosine similarity CS is calculated according to the vector space model with original language bag of words as unit, institute
The computational methods for stating cosine similarity CS are：
Wherein, Vec (Sexample) it is the source language sentence subvector of translation instance, Vec (Sinput) it is translation duties vector,
Vec(Sinput)·Vec(Sexample) it is two inner products of vector, | | | | it is the norm of vector；
Step 5, phrase table assorting process：According to the most like preceding 900-1100 similar translation reality that step 4 is selected
The preliminary translation result and standard translation translation R of example, using phrase table sort module by training after general postedit model
Phrase table is categorized as contributing to the positive phrase for lifting translation quality and the passive phrase for incorporating final translation result noise, makes
General postedit model after training becomes personalized postedit model, then by the positive phrase in personalized postedit model and
The preliminary translation result and standard translation of the similar translation example that passive phrase goes out with pseudo- feedback searching procedural retrieval in step 4
Translation R contrast, the passive phrase is filtered out from personalized postedit model phrase table, so as to obtain one optimization
Property postedit model；
Step 6, decoder module decoding process：Using in step 5 optimize personalized postedit model as translation mould
Type, the preliminary machine of the translation duties obtained to step 3 using traditional machine translation coding/decoding method using decoder module is turned over
Translate result T ' to be decoded, obtain the final translation result of goodization.
Specific embodiment four：
From turning over for the personalization machine system that pseudo- feedback is based on unlike specific embodiment three, described in present embodiment
Method is translated, decoding process described in step 6 utilizes formula：Treatment translation duties
Preliminary machine translation result T ' obtains the final translation result of goodization；In formula, P (T " | T ') is the translation of general postedit model
Probability, P (S | T ", T ') be in general postedit model using phrase to (T ", T ') to give input translation duties S just
Step machine translation sentence T ' carries out the probability of postedit model translation, and it is 1 or 0 to define its probable value, then by following two
Method obtain P (S | T ", T ') value：
1) when the phrase in the personalized postedit model of optimization is to (PT,PR) in two phrases respectively with translation duties
Preliminary machine translation result T ' and standard translation translation R in when thering is at least one phrase to match, P (S | T ", T ') probability
Value takes 1, otherwise takes 0；Or,
2) when the phrase in the personalized postedit model of optimization is to (PT,PR) in phrase PRIn standard translation translation R
When thering is at least one phrase to match, P (S | T ", T ') probable value take 1, otherwise take 0.
Specific embodiment five：
From the personalization machine system that pseudo- feedback is based on unlike specific embodiment three or four, described in present embodiment
Interpretation method, when carrying out described in step 4 pseudo- feedback searching process, from the preliminary translation result and standard of similar translation instance
Most like first 1000 are selected in the retrieval result of translation translation R.
Using IWSLT2012 Olympics as user input translation duties, using this translation duties data test
The personalization machine translation system and method based on pseudo- feedback of present invention design, the training of the translation duties offer of user input
Data are the spoken field of tourism, cover the concrete applications such as traffic, food and drink, stadiums, commercial affairs under Olympic Games application background
Occasion, altogether comprising 52,603 pairs of Chinese-English bilingual sentences right, specially 495,638 Chinese words and 527,599 English words, by it
As the local translation instance storehouse of the personalization of user.Employ including 2,057 pairs of Chinese-English bilinguals sentence to development set and including 998
To Chinese-English bilingual sentence to test set；Preliminary translation module has used Google's translation on line system, from Google's translation on line system
The translation result of above-mentioned language material is crawled, translation quality evaluation criterion uses BLEU-4, the test result that will be obtained is directly and paddy
Song translation result is contrasted.Meanwhile, the machine translation system that will be trained using the Moses instruments increased income is used as the second group pair
According to experiment, contrasted.
Evaluation criterion, the personalization machine translation system fed back based on puppet of present invention design and side are scored at BLEU-4
The translation result that method is obtained is contrasted with the translation result of Google's translation on line system, and its translation quality improves 19.5%；With
The translation result of the machine translation system that Moses instruments are trained is contrasted, and its translation quality improves 14.1%, test knot
Fruit is as shown in table 1：
Table 1：Personalized translation result based on pseudo- feedback is contrasted with the translation quality of other systems translation result.
Claims (5)
1. it is a kind of based on the pseudo- personalization machine translation system fed back, it is characterised in that the translation system includes：
For the phrase table filtering module that each the general postedit model phrase table to development set data is filtered；
Input module for obtaining the translation duties S of user input；
Preliminary machine translation result T ' for being obtained translation duties to being translated after user input translation duties S is right
The source language sentence in the translation instance storehouse that local system is provided translate and obtains tentatively turning over for the preliminary translation of the sentence T of translation instance
Translate module；
For in the translation instance storehouse of local system word alignment form, retrieval to obtain the preliminary translation result of similar translation example
With the pseudo- feedback searching module of standard translation translation R；
The phrase table classified for the phrase table to the postedit model after training and obtained personalized postedit model divides
Generic module；
The preliminary translation result of the similar translation example for being obtained to the retrieval of pseudo- feedback searching module is decoded, and is obtained most
The decoder module of whole translation result.
2. according to claim 1 based on the pseudo- personalization machine translation system fed back, it is characterised in that the phrase table mistake
Filter module is contained in the phrase table sort module.
3. the interpretation method of the personalization machine translation system based on pseudo- feedback described in a kind of claim 2, it is characterised in that：
Before user input translation duties S, using translation instance preliminary translation of the sentence T and standard translation translation R in translation memory
General postedit model is trained using statistical method, the training process of general postedit model is completed；The personalization machine is turned over
Method is translated to be realized by following steps：
Step one, phrase table filter process：Using phrase table filtering module to each general postedit models of development set data
Phrase table is filtered；
Based on the result after filtering using default-weight to each sentence D in development set dataiDecoded, produce n-best to turn over
Translate result；Then, n-best translation results are combined；Finally, knot is translated to the n-best after combination using MERT instruments
Fruit is overall to adjust ginseng, additionally it is possible to realize characteristic parameter optimization process；
Step 2, input process：Be input into translation duties S to input module by user；
Step 3, preliminary translation process：Turned over user input before the preliminary translation process includes user input translation duties S
Translate two parts after task S；
Before user input translation duties S, the transcription platform built using the machine translation system of local system will be locally
The source language sentence in the translation instance storehouse that system is provided tentatively is translated, and obtains the preliminary translation of the sentence T of translation instance；
Meanwhile, after the translation duties S by input module acquisition user input, translated using preliminary translation module
The preliminary machine translation result T ' of task；
Step 4, pseudo- feedback searching process：According to the preliminary translation of the sentence T of translation instance obtained in step 3, in local word pair
In the translation instance storehouse of neat form, the retrieval of cosine similarity is carried out with original language bag of words using pseudo- feedback searching module,
Obtain the preliminary translation result and standard translation translation R of similar translation example, and from the preliminary translation result of similar translation instance
It is individual with most like preceding 900-1100 is selected in the retrieval result of standard translation translation R；
Wherein, the cosine similarity CS is calculated according to the vector space model with original language bag of words as unit, described remaining
The computational methods of string similarity CS are：
Wherein, Vec (Sexample) it is the source language sentence subvector of translation instance, Vec (Sinput) it is translation duties vector, Vec
(Sinput)·Vec(Sexample) it is two inner products of vector, | | | | it is the norm of vector；
Step 5, phrase table assorting process：The most like preceding 900-1100 similar translation example selected according to step 4
Preliminary translation result and standard translation translation R, using phrase table sort module by training after general postedit model phrase
Table sort is to contribute to the positive phrase for lifting translation quality and the passive phrase that noise is incorporated to final translation result, makes training
General postedit model afterwards becomes personalized postedit model, then by the positive phrase in personalized postedit model and passiveness
The preliminary translation result and standard translation translation R of the similar translation example that phrase goes out with pseudo- feedback searching procedural retrieval in step 4
Contrast, the passive phrase is filtered out from personalized postedit model phrase table, so as to obtain a personalization for optimization
Postedit model；
Step 6, decoder module decoding process：The personalized postedit model optimized using in step 5 is sharp as translation model
The preliminary machine translation knot of the translation duties obtained to step 3 using traditional machine translation coding/decoding method with decoder module
Fruit T ' is decoded, and obtains the final translation result of goodization.
4. the interpretation method of the personalization machine translation system fed back based on puppet according to claim 3, it is characterised in that：Step
Decoding process described in rapid six utilizes formula：The preliminary machine for processing translation duties is turned over
Translate the final translation result that result T ' obtains goodization；In formula, P (T " | T ') is the translation probability of general postedit model, P (S |
T ", T ') it is that (T ", T ') is turned over to the preliminary machine for giving the translation duties S of input using phrase in general postedit model
Translating sentence T ' carries out the probability of postedit model translation, and it is 1 or 0 to define its probable value, then obtains P by following two methods
(S | T ", T ') value：
1) when the phrase in the personalized postedit model of optimization is to (PT,PR) in two phrases respectively with translation duties just
In step machine translation result T ' and standard translation translation R when thering is at least one phrase to match, P (S | T ", T ') probable value take
1, otherwise take 0；Or,
2) when the phrase in the personalized postedit model of optimization is to (PT,PR) in phrase PRAnd have in standard translation translation R to
When a few phrase matches, P (S | T ", T ') probable value take 1, otherwise take 0.
5. the interpretation method of the personalization machine translation system fed back based on puppet according to claim 3 or 4, its feature is existed
In：When carrying out pseudo- feedback searching process described in step 4, from the preliminary translation result and standard translation translation R of similar translation instance
Retrieval result in select most like first 1000.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410491100.9A CN104199813B (en) | 2014-09-24 | 2014-09-24 | Pseudo-feedback-based personalized machine translation system and method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410491100.9A CN104199813B (en) | 2014-09-24 | 2014-09-24 | Pseudo-feedback-based personalized machine translation system and method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104199813A CN104199813A (en) | 2014-12-10 |
| CN104199813B true CN104199813B (en) | 2017-05-24 |
Family
ID=52085108
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410491100.9A Active CN104199813B (en) | 2014-09-24 | 2014-09-24 | Pseudo-feedback-based personalized machine translation system and method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104199813B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107301173B (en) * | 2017-06-22 | 2019-10-25 | 北京理工大学 | A kind of automatic post-editing system and method for multi-source neural network remixing mode based on splicing |
| WO2020039808A1 (en) * | 2018-08-24 | 2020-02-27 | 株式会社Ｎｔｔドコモ | Machine translation control device |
| US20210056271A1 (en) * | 2018-08-24 | 2021-02-25 | Ntt Docomo, Inc. | Machine translation control device |
| CN111274827B (en) * | 2020-01-20 | 2021-05-28 | 南京新一代人工智能研究院有限公司 | Suffix translation method based on multi-target learning of word bag |
| CN113807106B (en) * | 2021-08-31 | 2023-03-07 | 北京百度网讯科技有限公司 | Translation model training method and device, electronic equipment and storage medium |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103793368A (en) * | 2012-10-31 | 2014-05-14 | 上海勇金懿信息科技有限公司 | Method for automatically protecting marks in marking language in automatic translation processing |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8494835B2 (en) * | 2008-12-02 | 2013-07-23 | Electronics And Telecommunications Research Institute | Post-editing apparatus and method for correcting translation errors |
-
2014
- 2014-09-24 CN CN201410491100.9A patent/CN104199813B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103793368A (en) * | 2012-10-31 | 2014-05-14 | 上海勇金懿信息科技有限公司 | Method for automatically protecting marks in marking language in automatic translation processing |
Non-Patent Citations (2)
| Title |
|---|
| A Step towards Human-Machine Unification Using Translation Memory and Machine Translation System;Nishtha Jaiswal等;《2011 International Conference on Languages, Literature and Linguistics》;20111228;第26卷;第64-68页 * |
| 一种智能译后编辑器的设计及其实现算法;黄河燕等;《软件学报》;19950305;第6卷(第3期);第129-135页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104199813A (en) | 2014-12-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10191892B2 (en) | Method and apparatus for establishing sentence editing model, sentence editing method and apparatus | |
| CN104199813B (en) | Pseudo-feedback-based personalized machine translation system and method | |
| CN102169495B (en) | Industry dictionary generating method and device | |
| CN110674252A (en) | High-precision semantic search system for judicial domain | |
| CN110276071A (en) | A kind of text matching technique, device, computer equipment and storage medium | |
| CN107102976A (en) | Entertainment newses autocreating technology and system based on microblogging | |
| CN111061861A (en) | XLNET-based automatic text abstract generation method | |
| CN111814477B (en) | Dispute focus discovery method and device based on dispute focus entity and terminal | |
| CN110119443A (en) | A kind of sentiment analysis method towards recommendation service | |
| CN109408806A (en) | A kind of Event Distillation method based on English grammar rule | |
| Wu et al. | Chain of thought prompting elicits knowledge augmentation | |
| CN113407842B (en) | Model training method, theme recommendation reason acquisition method and system and electronic equipment | |
| Deb et al. | Oboyob: A sequential-semantic bengali image captioning engine | |
| CN112949293A (en) | Similar text generation method, similar text generation device and intelligent equipment | |
| Pal et al. | Anubhuti--An annotated dataset for emotional analysis of Bengali short stories | |
| CN116910251A (en) | Text classification method, device, equipment and medium based on BERT model | |
| Parikh et al. | IRLab_DAIICT at SemEval-2020 task 9: Machine learning and deep learning methods for sentiment analysis of code-mixed tweets | |
| Abid | The SADID evaluation datasets for low-resource spoken language machine translation of Arabic dialects | |
| CN116595970A (en) | Sentence synonymous rewriting method and device and electronic equipment | |
| CN114091469B (en) | Network public opinion analysis method based on sample expansion | |
| JP5298834B2 (en) | Example sentence matching translation apparatus, program, and phrase translation apparatus including the translation apparatus | |
| Mezzoudj et al. | Arabic Algerian Oranee dialectal language modelling oriented topic | |
| Arreola et al. | An Embeddings Based Recommendation System for Mexican Tourism. Submission to the REST-MEX Shared Task at IberLEF 2021 | |
| Ghayoomi | Training vs Post-training Cross-lingual Word Embedding Approaches: A Comparative Study | |
| Crespo-Sanchez et al. | A content spectral-based text representation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20200327 Address after: 150001 No. 118 West straight street, Nangang District, Heilongjiang, Harbin Patentee after: Harbin University of technology high tech Development Corporation Address before: 150001 Harbin, Nangang, West District, large straight street, No. 92 Patentee before: HARBIN INSTITUTE OF TECHNOLOGY |
|
| TR01 | Transfer of patent right |