CN105573992B - Real-time translation method and apparatus - Google Patents
Real-time translation method and apparatus Download PDFInfo
- Publication number
- CN105573992B CN105573992B CN201510933733.5A CN201510933733A CN105573992B CN 105573992 B CN105573992 B CN 105573992B CN 201510933733 A CN201510933733 A CN 201510933733A CN 105573992 B CN105573992 B CN 105573992B
- Authority
- CN
- China
- Prior art keywords
- translation
- chinese character
- result
- character string
- candidate
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/40—Processing or translation of natural language
- G06F40/58—Use of machine translation, e.g. for multi-lingual retrieval, for server-side translation for client devices or for real-time translation
Abstract
The invention discloses a real-time translation method and apparatus. The method comprises the following steps: step 1, obtaining a text input by a user and a corresponding pinyin string in real time; step 2, segmenting the pinyin string according to all possibilities by taking pinyin of a single Chinese character as a unit, taking candidate Chinese characters corresponding to the pinyin as nodes, forming a grid structure by connecting lines of the candidate Chinese characters, scoring a Chinese character string composed of any segmentation path in the grid by using a language model to obtain candidate results, and selecting an optimal Chinese character string from the candidate results; and step 3, calling a translation algorithm to translate the optimal Chinese character string, and outputting a translation result. Through the real-time translation method and apparatus, the technical problem that the user needs to occasionally perform switching between the text and an online dictionary during real-time translation of the input text is solved.
Description
Technical field
The present invention relates to a kind of real time translating method and device, more particularly to a kind of real time translation side to input text
Method and device.
Background technology
With the development of global economic integration, cross-cultural, transnational between exchange increasingly frequent, exchange way also becomes
In diversified, multi-lingual opinion on public affairs, internationalization, the impact that language difference produces also seems more and more prominent, turning between different language
Translate and seem increasingly important.
Currently, when people need cypher text, often by means of some online dictionaries or translation engine,
But the working environment of these resources and user is often detached from, user needs handoff scenario frequently, such as:In word
Input text, however it is necessary that be switched to webpage to search translation up.So make translation become efficiency very low, work together very
Inconvenient.
Certainly, also some translation software occur in that screen word-selecting instant translation function at present, can be user-friendly, but
Nonetheless, need to the text instant translation of input when it is still desirable to user stop current inputting, by mobile for mouse to being wanted
On the words intercepting, instant translation result then just occurs, this user is still felt inconvenience.
Content of the invention
For overcoming the shortcomings of above-mentioned prior art presence, the main purpose of the present invention is to provide a kind of real time translating method
And device, which solve user's needs switching between text and online dictionary frequently when to input text real time translation, keep away
Exempt to leave the technical problem that operation interface consults other media.
For reaching above and other purpose, the present invention proposes a kind of real time translating method, comprises the steps:
Step one, the text of user in real input and corresponding pinyin string;
Step 2, by pinyin string in units of individual Chinese character phonetic according to being possible to carry out cutting, corresponding with phonetic
Each candidate Chinese character is node, and between candidate Chinese character, line forms network, for any one cutting route in grid
The Chinese character string constituting, is given a mark using language model, obtains candidate result, select in candidate result one optimum
Chinese character string;
Step 3, calls translation computing that described optimum Chinese character string is translated, and translation result is carried out defeated
Go out.
Further, also included before step 3 in described step 2 obtain optimum Chinese character string carry out pre-
Process, the step identifying the constituent of Chinese character string of described optimum.
Further, also comprised the steps before step 3:
Translate record being viewed locally whether once having, if it has, then direct returning result.
Further, in step 3, call translation computing that the part not having to translate is translated after setting time,
And translation result is returned local.
Further, the step calling translation computing after setting time includes:
User creates a thread t in main thread after tapping a keyboard, and records the ID of t;
Timer is set in described thread t, after n millisecond, checks the ID whether ID phase with main thread record of thread t
Deng;
If equal, illustrate that t is up-to-date thread, then access translating server, if unequal, illustrate that main thread is created again
Build new thread, have updated ID, t has not been up-to-date, then exit.
Further, in step 2, described candidate result is preserved in the buffer according to marking height sequence, described
Before candidate result preserves in the buffer according to marking height sequence, inquire about local cache, if there is current input in caching
Described pinyin string, then by caching in result be discharged to first, by described candidate result according to marking height be discharged in caching
Result after.
For reaching above-mentioned purpose, the present invention also provides a kind of real time translator, including:
Character string acquisition module, the text for user in real input and corresponding pinyin string;
Syllable text conversion module, for just pinyin string in units of individual Chinese character phonetic according to being possible to carry out cutting,
With phonetic each candidate Chinese character corresponding as node, between candidate Chinese character, line forms network, for any in grid
Article one, the Chinese character string that cutting route is constituted, is given a mark using language model, obtains candidate result, selects in candidate result
Select an optimum Chinese character string;
Translation module, calls translation computing that this optimum Chinese character string is translated, and translation result is carried out defeated
Go out.
Further, this real time translator also includes pretreatment module, the optimum that described Syllable text conversion module is obtained
Chinese character string pre-processed, identify described optimum Chinese character string constituent.
Further, described real time translator also includes translation cache module, calls translation for described translation module
Being viewed locally whether once having translation record before computing, if it has, then direct returning result.
Further, described real time translator also includes calling control module, calls for controlling after setting time
Described translation module carries out translating computing, and translation is returned local.
Compared with prior art, a kind of real time translating method of the present invention and device can make user need not leave work at present ring
Border, directly passes through input and obtains translation, it is to avoid leave the trouble that operation interface consults other media,
Thus improve efficiency and improving Consumer's Experience, can accurately be translated content, the present invention is permissible simultaneously
It is being viewed locally whether once there was translation record, if it has, then direct returning result, accelerating the speed of real time translation.
Brief description
Fig. 1 is a kind of flow chart of steps of real time translating method of the present invention;
Fig. 2 is the Syllable text conversion process schematic of the present invention;
Fig. 3 is a kind of configuration diagram of real time translator of the present invention.
Specific embodiment
Below by way of specific instantiation and embodiments of the present invention are described with reference to the drawings, those skilled in the art can
Understand further advantage and effect of the present invention by content disclosed in the present specification easily.The present invention also can be by other different
Instantiation implemented or applied, the every details in this specification also can be based on different viewpoints and application, without departing substantially from
Carry out various modification and change under the spirit of the present invention.
Fig. 1 is a kind of flow chart of steps of real time translating method of the present invention.As shown in figure 1, a kind of real time translation of the present invention
Method, comprises the steps:
Step 101, the text of user in real input and corresponding pinyin string;
Step 102, by pinyin string in units of individual Chinese character phonetic according to being possible to carry out cutting, corresponding with phonetic
Each candidate Chinese character is node, and between candidate Chinese character, line forms network, for any one cutting route in grid
The Chinese character string constituting, is given a mark using language model, obtains candidate result, select in candidate result one optimum
Chinese character string simultaneously records in local cache.
As shown in Fig. 2 illustrating, by pinyin string according to being possible to carry out cutting, form network, such as:
Tiananmen can be cut into tian ' an ' men it is also possible to be cut into ti ' an ' an ' men;For any in grid
Article one, cutting route, is given a mark using language model, finally lists candidate according to score height.
Assume now using two gram language model (bigram language model), so-called language model can be understood as
On the premise of continuous N-1 character string occurs, the Probability p of n-th character appearance (wn | w1, w2 ..., wn-1), N=2 it is simply that
Binary model;Add to character string first<s>Character representation character start of string,</s>Character representation terminates, then two paths become
For:
<s> tian an men </s>
<s> ti an an men </s>
First paths are given a mark, calculates:
p(tian|<s>)*p(an|tian)*p(men|an)*p(</s>|men)
Second paths are in the same manner:
p(ti|<s>)*p(an|ti)*p(an|an)*p(men|an)*p(</s>|men)
Wherein, the probability of p (wn | wn-1) obtains from corpus.This is most basic calculation, the present invention
It is not limited.
After user have selected candidate kanji character string, pinyin-Hanzi character string be recorded in local cache.
As illustrated in fig. 2, it is assumed that 4 phonetics are shu ru fang fa, and assume that shu corresponds to a candidate
" defeated ", ru corresponds to three candidates " such as entering you ", and fang corresponds to two candidates " side of putting ", and corresponding four candidates of fa " send out method
Penalize ", each node of corresponding upper in figure respectively, paths are a Chinese character string, such as:Input method, method is put in input.
Step 103, pre-processes to the Chinese character string of optimum obtaining in described step 102, identify described in
The constituent of excellent Chinese character string.
Specifically, in order to accurately translate the word of user's selection, need the Chinese character word string selecting is pre-processed, than
As have selected " Tian An-men ", then be exactly " Tian An-men ", have selected " motion hidden door " it is simply that " motion hidden door ", identifying the inside
Special composition, provides more information for translation process below.
Step 104, calls translation computing that pretreated optimum Chinese character string is translated, and translation result is entered
Row output, it is preferred that simultaneously by translation result(Translation)It is stored in local cache.
It is preferred that before step 103, also comprising the steps:
Translate record being viewed locally whether once having, if it has, then direct returning result.
Specifically, first checked locally whether once there is translation record before calling translation computing, if it has, then not
Call translation service, direct returning result.For example, when user inputs same noun repeatedly, caching can be accelerated accordingly
Speed.
This method is briefly described as by the input through keyboard of client end interface receive user, then by the pinyin string of input
Be converted to Chinese character candidate, when user selects corresponding candidate character, this candidate character be transferred to backstage translation, using multimode
The interpretation method that type merges obtains translation, finally translation is back to client end interface.
It is preferred that in a step 102, described candidate result is preserved in the buffer according to marking height sequence;More preferably,
Inquired about local cache before described candidate result preserves in the buffer according to marking height sequence, if existing current in caching
The described pinyin string of input, then be discharged to first by the result in caching, described candidate result is discharged to slow according to marking height
After the result deposited.
That is, inquiry local cache, if there is the pinyin string of current input in caching(Refer to original character string
Tiananmen, checks whether that user once inputted), then the result in caching is discharged to first, by described cutting result
It is discharged to after buffered results, cutting result is given a mark as procedure described above and sorted from high to low, the result that certainly will and cache
Duplicate removal, if user once inputted, then directly the result of last time is taken out and makes number one, and then other results are pressed
According to fraction size be listed in after as candidate;If the phonetic of not current input, directly cuts foregoing in caching
Point result is listed.
After user have selected candidate character, pinyin-Hanzi be recorded in local cache;
Here it should be noted that, described candidate result according to marking height sequence preserve in the buffer, the capacity of caching is big
I is selected according to demand, is such as set as 1000 records, caching is divided into two parts, a area and b area, a area storage user is
The N of nearly input(≤1000)Data(Temporally), b area stores the most front 1000-N bar note of number of times of user's history input
Record(By number of times), in caching, record storage is a area or b area, after receiving the pinyin-Hanzi string of user's new input:
If all do not had in caching a+b, add a area, if a area record is full, cover time that note at most
Record, record input number of times is 1
If existed in caching b, this record occurrence number+1
If existed in caching a, this record occurrence number+1, and inquire about whether the occurrence number of this record is more than b area
In a certain bar record occurrence number, if there are:If b area record is full, cover that record in b area, otherwise directly
Move to b area.
More excellent, in step 103, the constituent of described optimum Chinese character string includes:Word unit, numeral, the date,
In specific place name, name, mechanism's name, spoken language one or more.
Specifically, Chinese word segmenting, character string cutting is word unit;
Normalization:Numeral, date expression formula are converted into consolidation form, for example:32 → 32,200->200,
March → March, and then facilitate subsequent treatment;
Entity recognition:Name in identification string, place name, mechanism's name, facilitate subsequent treatment;
Spoken identification:Identify the informal written word part of the popular word of user input etc, facilitate subsequent treatment.
It is preferred that at step 104, translation computing can be recalled after setting time, specially user taps a keyboard
Create a thread t in main thread afterwards, and record the ID of t, timer is set in described thread t, after n millisecond, checks line
Whether the ID of journey t is equal with the ID of main thread record, if equal, illustrates that t is up-to-date thread, then access translating server,
If unequal, illustrate that main thread creates new thread again, have updated ID, t is not up-to-date, then exit.
When user taps keyboard every time, candidate word is all it may happen that change, if candidate change once begins to adjust
With translation service, then can increase the burden of server, have a strong impact on Consumer's Experience;The present invention is using the side of delay+asynchronous call
Formula is entered, and after that is, user taps a keyboard, creates a thread t, arranges timer, after n millisecond, thread t is examined in t
Check and see whether oneself is currently up-to-date thread, if it is not, then exiting;If it is, accessing translation service further.
Concrete grammar is as follows:
Create thread t, the ID of record t in main thread;
T starts timing;
After n millisecond, check whether the ID of oneself is equal with the ID of main thread record;
If equal, illustrate that t is last thread, i.e. up-to-date thread, then continue computing;
If unequal, illustrate that main thread creates new thread again, have updated ID, t is not up-to-date, then exit.
More excellent, in step 104, call the translation operation method in translation computing to be to turn over using multiple after setting time
Translate the mode of Model Fusion, i.e. the pre-processed results according to step 103, turned over by the way of multiple translation models merge
Translate.The mode that this multiple translation model merges includes:
Rule-based translation model:
Because normalized content is limited to numeral, date etc., rule-based interpretative system can be adopted, so permissible
Accurately translated content;
The translation model of Case-based Reasoning:
For entity, spoken vocabulary, the method using Case-based Reasoning is translated.Will identify that the composition conduct coming
Query, enters line retrieval in bilingual case library, if retrieved, returns corresponding translation;
For conventional sentence, phrase, vocabulary translation, translated using the statistical translation model merging.
Fig. 3 is a kind of configuration diagram of real time translator of the present invention.As shown in figure 3, a kind of real time translation of the present invention
Device, including:Character string acquisition module 301, Syllable text conversion module 302, pretreatment module 303 and translation module 304.
Character string acquisition module 301 is used for the text of user in real input and corresponding pinyin string.
Syllable text conversion module 302, for just pinyin string in units of individual Chinese character phonetic according to being possible to be cut
Point, with phonetic each candidate Chinese character corresponding as node, between candidate Chinese character, line forms network, for appointing in grid
The Chinese character string that one cutting route of meaning is constituted, is given a mark using language model, obtains candidate result, in candidate result
An optimum Chinese character string is selected to preserve in the buffer.Preferably, candidate result according to marking height sequence be saved in slow
In depositing.More preferably, inquired about local cache before candidate result preserves in the buffer according to marking height sequence, if in caching
There is the described pinyin string of current input, then the result in caching is discharged to first, described candidate result is high according to marking
After the low result being discharged in caching.
Pretreatment module, pre-processes to the Chinese character string of the optimum that Syllable text conversion module 302 obtains, identifies institute
State the constituent of optimum Chinese character string, wherein, the constituent of optimum Chinese character string includes:Word unit, numeral, day
Phase, specific place name, name, mechanism's name, spoken language in one or more.
Translation module 304, calls translation computing that pretreated optimum Chinese character string is translated, and translation is tied
Fruit is exported, it is preferred that simultaneously by translation result(Translation)It is stored in local cache.In the present invention, translation module 304 is adopted
Translation operation method is by the way of multiple translation models merge
Preferably, the real time translator of the present invention also includes translation cache module 305, for whether being once viewed locally
Through there being translation record, if it has, then direct returning result.
Preferably, the real time translator of the present invention also includes calling control module 306, for calling after setting time
Translation module 304 carries out translating computing, and translation is returned local.
More preferably, call control module 306, call translation module 204 to carry out translating computing after setting time, specifically side
Formula is:User creates a thread t in main thread after tapping a keyboard, and records the ID of t, and it is fixed to arrange in described thread t
When device, after n millisecond, check whether the ID of thread t equal with the ID of main thread record, if equal, illustrate that t is up-to-date
Thread, then access translating server, if unequal, illustrate that main thread creates new thread again, have updated ID, t is not
New, then exit.
In sum, a kind of real time translating method of the present invention and device can make user need not leave current operating environment, directly
Connect input and obtained translation, thus improve efficiency and improving Consumer's Experience, can accurately be translated content, simultaneously
The present invention can be viewed locally whether once there was translation record, if it has, then direct returning result, accelerates real time translation
Speed.
Above-described embodiment only principle of the illustrative present invention and its effect, not for the restriction present invention.Any
Skilled person all can be modified to above-described embodiment and changed without prejudice under the spirit and the scope of the present invention.Therefore,
The scope of the present invention, should be as listed by claims.
Claims (4)
1. a kind of real time translating method, comprises the steps:
Step one, the text of user in real input and corresponding pinyin string;
Step 2, by pinyin string in units of individual Chinese character phonetic according to being possible to carry out cutting, with phonetic corresponding each
Candidate Chinese character is node, and between candidate Chinese character, line forms network, constitutes for any one cutting route in grid
Chinese character string, given a mark using language model, obtain candidate result, by described candidate result according to marking height sort
Preserve in the buffer, before described candidate result preserves in the buffer according to marking height sequence, inquire about local cache, if
There is the described pinyin string of current input in caching, then the result in caching is discharged to first, by described candidate result, goes
Other results after result in the described caching of weight, are discharged to after the result in caching according to marking height, select in candidate result
One optimum Chinese character string;
Step 3, calls translation computing that described optimum Chinese character string is translated, and translation result is exported;
Also included the Chinese character string of the optimum obtaining in described step 2 is pre-processed before step 3, identify institute
The step stating the constituent of Chinese character string of optimum;
In step 3, call translation computing that the part not have translation is translated after setting time, and by translation result
Return local;
The step calling translation computing after setting time includes:
User creates a thread t in main thread after tapping a keyboard, and records the ID of t;
Timer is set in described thread t, after n millisecond, checks whether the ID of thread t is equal with the ID of main thread record;
If equal, illustrate that t is up-to-date thread, then access translating server, if unequal, illustrate that main thread creates again
New thread, have updated ID, and t is not up-to-date, then exit.
2. as claimed in claim 1 a kind of real time translating method it is characterised in that before step 3, and to described step
Before the Chinese character string of the optimum obtaining in rapid two is pre-processed, also comprise the steps:
Translate record being viewed locally whether once having, if it has, then direct returning result.
3. a kind of real time translator, including:
Character string acquisition module, the text for user in real input and corresponding pinyin string;
Syllable text conversion module, for by pinyin string in units of individual Chinese character phonetic according to being possible to carry out cutting, with phonetic
Each candidate Chinese character corresponding is node, and between candidate Chinese character, line forms network, cuts for any one in grid
The Chinese character string that sub-path is constituted, is given a mark using language model, obtains candidate result, by described candidate result according to beating
Height sequence is divided to preserve in the buffer, before just sequence preserves in the buffer described candidate result according to giving a mark, inquiry is originally
Ground caching, if there is the described pinyin string of current input in caching, the result in caching is discharged to first, by described time
Select in result, other results after result in caching described in duplicate removal, it is discharged to after the result in caching according to marking height, in candidate
An optimum Chinese character string is selected in result;
Translation module, calls translation computing that this optimum Chinese character string is translated, and translation result is exported;
Described real time translator also includes pretreatment module, before described translation module is translated, described sound word is turned
The Chinese character string of the optimum that die change block obtains is pre-processed, and identifies the constituent of described optimum Chinese character string;
Described real time translator also includes calling control module, calls described translation module to enter for controlling after setting time
Row translation computing, and translation is returned local;
The step calling translation computing after setting time includes:
User creates a thread t in main thread after tapping a keyboard, and records the ID of t;
Timer is set in described thread t, after n millisecond, checks whether the ID of thread t is equal with the ID of main thread record;
If equal, illustrate that t is up-to-date thread, then access translating server, if unequal, illustrate that main thread creates again
New thread, have updated ID, and t is not up-to-date, then exit.
4. as claimed in claim 3 a kind of real time translator it is characterised in that:Described real time translator also includes translating
Cache module, calls before translation computing for described translation module, and the Chinese to the optimum that described Syllable text conversion module obtains
Before word character string is pre-processed, translate record being viewed locally whether once having, if it has, then direct returning result.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510933733.5A CN105573992B (en) | 2015-12-15 | 2015-12-15 | Real-time translation method and apparatus |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510933733.5A CN105573992B (en) | 2015-12-15 | 2015-12-15 | Real-time translation method and apparatus |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105573992A CN105573992A (en) | 2016-05-11 |
CN105573992B true CN105573992B (en) | 2017-02-15 |
Family
ID=55884142
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510933733.5A Active CN105573992B (en) | 2015-12-15 | 2015-12-15 | Real-time translation method and apparatus |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105573992B (en) |
Families Citing this family (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106980390A (en) * | 2016-01-18 | 2017-07-25 | 富士通株式会社 | Supplementary translation input method and supplementary translation input equipment |
CN108959274B (en) * | 2018-06-27 | 2022-09-02 | 维沃移动通信有限公司 | Translation method of application program and server |
CN109032375B (en) * | 2018-06-29 | 2022-07-19 | 北京百度网讯科技有限公司 | Candidate text sorting method, device, equipment and storage medium |
CN109921983A (en) * | 2019-03-07 | 2019-06-21 | 上海百途归一智能科技有限公司 | A kind of multilingual chat mobile application |
CN110262674B (en) * | 2019-06-27 | 2023-07-18 | 北京金山安全软件有限公司 | Chinese character input method and device based on pinyin input and electronic equipment |
CN113076760B (en) * | 2020-01-03 | 2024-01-26 | 阿里巴巴集团控股有限公司 | Translation and commodity retrieval method and device, electronic equipment and computer storage medium |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102129423A (en) * | 2010-01-13 | 2011-07-20 | 阿里巴巴集团控股有限公司 | Statement display method and device |
CN102193643A (en) * | 2010-03-15 | 2011-09-21 | 北京搜狗科技发展有限公司 | Word input method and input method system having translation function |
CN104133561A (en) * | 2014-07-31 | 2014-11-05 | 百度在线网络技术(北京)有限公司 | Auxiliary information display method and device based on input method |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9037450B2 (en) * | 2012-12-14 | 2015-05-19 | Microsoft Technology Licensing, Llc | Text overlay techniques in realtime translation |
-
2015
- 2015-12-15 CN CN201510933733.5A patent/CN105573992B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102129423A (en) * | 2010-01-13 | 2011-07-20 | 阿里巴巴集团控股有限公司 | Statement display method and device |
CN102193643A (en) * | 2010-03-15 | 2011-09-21 | 北京搜狗科技发展有限公司 | Word input method and input method system having translation function |
CN104133561A (en) * | 2014-07-31 | 2014-11-05 | 百度在线网络技术(北京)有限公司 | Auxiliary information display method and device based on input method |
Also Published As
Publication number | Publication date |
---|---|
CN105573992A (en) | 2016-05-11 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105573992B (en) | Real-time translation method and apparatus | |
AU2021202694B2 (en) | Facilitating end-to-end communications with automated assistants in multiple languages | |
EP1450267A2 (en) | Methods and systems for language translation | |
WO2008098507A1 (en) | An input method of combining words intelligently, input method system and renewing method | |
CN101458681A (en) | Voice translation method and voice translation apparatus | |
JP2019067433A (en) | Subject provision system | |
JP2020135135A (en) | Dialog content creation assisting method and system | |
CN110516035A (en) | A kind of man-machine interaction method and system of mixing module | |
JP2023062173A (en) | Video generation method and apparatus of the same, and neural network training method and apparatus of the same | |
WO2022148467A1 (en) | Cross-language data enhancement-based word segmentation method and apparatus | |
Yang et al. | Guwen-UNILM: machine translation between ancient and modern Chinese based on pre-trained models | |
KR101616031B1 (en) | Query Translator and Method for Cross-language Information Retrieval using Liguistic Resources from Wikipedia and Parallel Corpus | |
CN116361416A (en) | Speech retrieval method, system and medium based on semantic analysis and high-dimensional modeling | |
JP4643183B2 (en) | Translation apparatus and translation program | |
Ramaswamy et al. | Hierarchical feature-based translation for scalable natural language understanding | |
JPH0652151A (en) | Common starting learning device and kana@(3754/24)japanese syllabary)-kanji@(3754/24)chinese character) conversion device using it | |
JPH09179859A (en) | Device and method for character string conversion | |
JP2017219992A (en) | Translation device and translation system | |
Rahman et al. | A framework for building a natural language interface for bangla | |
JP2017181662A (en) | Voice translating device, voice translating method, and voice translating program | |
CN117077693A (en) | Simultaneous interpretation auxiliary method and system for displaying voice recognition and bilingual terms in real time | |
Ruzicka | 10 KALAM: A Word Analyzer for Sabaic | |
Varghese et al. | A Novel Approach For English-Hindi Cross lingual Summarization | |
Guo | Intelligent English Automatic Translation System Based on Human-Computer Interaction Enhancement Algorithm | |
Wang | Research on Text Simplification Method Based on BERT |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
CP01 | Change in the name or title of a patent holder |
Address after: 100040 middle iron Development Building, No. 20, Shijingshan Road, Shijingshan District, Beijing, 16 layers Patentee after: Chinese translation language through Polytron Technologies Inc Address before: 100040 middle iron Development Building, No. 20, Shijingshan Road, Shijingshan District, Beijing, 16 layers Patentee before: Mandarin Technology (Beijing) Co., Ltd. |
|
CP01 | Change in the name or title of a patent holder |