WO2008145055A1 - The method for obtaining restriction word information, optimizing output and the input method system - Google Patents

Abstract

The method for obtaining restriction word information includes the follow steps, obtaining characteristic information based on the target word, and judging whether the characteristic information is in accordance with the preset condition. If being suitable, the information ensures that the target word is a restriction word, and a related restriction information is recorded. The restriction information is used for a restricting sequence when the word is singly outputted. The inventive embodiment, by presetting the word bank including the input and output method of restriction word information, judges whether the output candidate item accords with the preset condition, based on the result, judges whether the candidate item with restriction word information is displayed and outputted. Accordingly user can obtain more effective output without increasing the operation, the character output process of the input system is optimized greatly, and the intelligence of the input system is also improved.

Description

 Method for obtaining restricted word information, method for optimizing output and input method system The present application claims to be submitted to the Chinese Patent Office on May 25, 2007, and the application number is 200710099644. 0, the invention name is "method for obtaining restricted word information, optimized output" The priority of the method and the input method system of the Chinese Patent Application, the entire contents of which is incorporated herein by reference. Technical field

 The present invention relates to the field of computer character input data processing, and more particularly to a method and apparatus for acquiring restricted word information, a method for updating an input method vocabulary, a method for optimizing output, and an input method system.

Background technique

 With the popularization and development of computer technology and Internet technology, users of different professional fields, different interests and usage habits are increasingly demanding the intelligence of the input method system.

 In the prior art, techniques for using the Internet corpus statistics to filter and input the input lexicon have emerged. The resulting Internet vocabulary can contain many new words that are not available through previous closed corpus information (such as modern Chinese dictionaries, news, newspapers, etc.), which can greatly improve people's input efficiency. However, it is precisely because of the complexity of the Internet corpus that some of the words that are derived from the word frequency statistics have some deficiencies in linguistics or the use of input habits.

 For example, for the pinyin encoded string "liangjiang" entered by the user, generally available candidates include "two rivers,", "goods,", etc., candidates with Internet thesaurus may also include "quantity" because " The amount of "this word" appears quite high in Internet pages, but it generally appears at the junction of multiple words in a sentence (used to express a link relationship), for example, "the amount of passengers will exceed." The word "quantity" is included in the input lexicon, which can increase the intelligence of the input method (to achieve a higher intelligent group effect), in some cases can improve user input efficiency, but because of the amount , the word rarely appears in the case of a separate word, which may also cause trouble for the user input, increase the number of candidates that the user needs to select, and reduce the input efficiency.

 Therefore, a technical problem that is urgently needed to be solved by those skilled in the art is: how to find such words with linguistics or habitual defects in the vocabulary, and impose restrictions on the input process to further improve the input efficiency.

Summary of the invention The technical problem to be solved by the present invention is to provide a method and apparatus for acquiring restriction word information, which can find words having linguistics or usage habits from a large number of vocabularies, thereby improving the user's input experience.

 Another object of the present invention is to provide a method for updating an input method vocabulary, a method for optimizing an output, and an input method system, which can restrict certain words in some cases in an actual input process, thereby It can achieve the purpose of improving the intelligence of the input method without increasing the user's operation.

 In order to solve the above technical problem, the present invention discloses a method for acquiring restriction word information, which may specifically include: acquiring a target word; acquiring feature information corresponding to the target word; determining whether the feature information or its corresponding calculation result is consistent The preset condition, if it is met, determines that the target word is a restriction word and records related restriction information, and the restriction information is used to limit the ordering when the word is separately output.

 Preferably, the feature information is: a feature value of the word at the beginning of the target word in the default corpus as a prefix of the word, and a feature value of the word at the end of the target word in the preset corpus as a suffix; The preset condition for determining is: whether at least one of the feature values is in a preset range.

 Preferably, the feature information is: a feature value of a linguistic collocation relationship of each single word and/or a multi-word word included in the target word in a preset corpus; : Whether at least one of the above feature values exists in the preset range.

 Preferably, the feature information is: an attribute value separately input by the user in the input method application; the preset condition for determining is: whether the feature value belongs to a preset range.

 Preferably, the feature information includes: a feature value of the word at the beginning of the target word in the default corpus as a prefix; the word at the end of the target word is used as a suffix feature value in the default corpus; The common word frequency of the target word; the preset condition for the judgment is: whether the ratio of the at least one feature value to the target word common word frequency in the feature value belongs to a preset range.

Preferably, the feature information includes: a feature value of a linguistic collocation relationship of each single word and/or a multi-word included in the target word in a preset corpus; and a general word frequency of the target word; The preset condition for judging is: whether at least one feature value exists in the above feature value The ratio of the general word frequency to the target word belongs to a preset range.

 Preferably, the feature information is: a feature value that the target word is input by the user in the input method application; and a general word frequency of the target word; the preset condition for the judgment is: the feature value and the target word Whether the ratio of the general word frequency belongs to the preset range.

 Preferably, the feature information is: user-sorted position information of the target words in each candidate word encoded for the same input; and original sorted position information of the target words in each candidate word encoded for the same input; The user ranking information is related to a feature value that the target word is separately input by the user in the input method application; the original sorting information is related to a general word frequency of the target word; and the preset condition for determining is: the user Whether the difference between the sort position information and the original sort position information belongs to a preset range.

 Preferably, before the feature information obtaining step, the method further includes: an optimal selection step of the target word. Preferably, the restriction information includes: a weight of the restriction word that is separately output in each preset scenario.

 Preferably, the restriction information further includes: a linguistic allocation parameter of the restriction word in a preset corpus; the linguistic collocation parameter is used to limit the ordering of the word when the intelligent group word is output.

 Preferably, the method may further include: generating a vocabulary or vocabulary, the vocabulary or vocabulary including the restricted words and related restriction information; or generating a vocabulary, the vocabulary including the Limit words and their associated restrictions, as well as generic terms.

 According to another embodiment of the present invention, a method for acquiring restricted word information includes: acquiring a target word; obtaining a linguistic collocation parameter of the target word in a preset corpus; determining the linguistic collocation parameter Whether the preset condition is met, if yes, the restriction information of the target word is recorded, and the restriction information includes a corresponding linguistic collocation parameter; the restriction information is used to limit the ordering of the word intelligent group word output.

 Preferably, the linguistic collocation parameter is a general parameter; or the linguistic collocation parameter includes a sub-parameter for each preset scene.

According to another embodiment of the present invention, a method for updating a thesaurus includes: acquiring a target word; acquiring feature information corresponding to the target word; determining whether the feature information or its corresponding calculation result meets a pre-determination Setting a condition, if yes, determining that the target word is a restriction word and recording related restriction information, the restriction information is used to limit the ordering when the word is output alone, and/or, Used to limit the ordering of the word intelligent group word output; add the limit word and its related restriction information to the existing vocabulary of the input method.

 Preferably, the adding is: determining whether the restricted word already exists in the original thesaurus, and if so, recording only relevant restriction information into the existing thesaurus of the input method; or, adding For example, the restriction word and its related restriction information are directly recorded into the existing vocabulary of the input method, and if the vocabulary is repeated, the original vocabulary is overwritten; or, the adding is: The related restriction information is stored as a restricted vocabulary, and the existing vocabulary of the restricted vocabulary and the input method is used for collaborative completion of candidate ordering.

 Preferably, the restriction word has restriction information in each preset scenario.

 According to another embodiment of the present invention, an apparatus for acquiring restriction word information is further disclosed, including:

 a target word obtaining unit, configured to acquire a target word;

 a feature information acquiring unit, configured to acquire feature information corresponding to the target word;

 a restriction information acquiring unit, configured to determine whether the feature information or a corresponding calculation result thereof meets a preset condition, and if yes, determine that the target word is a restriction word and record related restriction information, where the restriction information is used to limit the word Sorting when outputting separately, and/or, is used to limit the ordering of the word smart group word output.

 According to another embodiment of the present invention, a method for optimizing an output is disclosed, including: receiving user input information, and converting the input information; obtaining an output candidate; determining whether an output candidate meets application restriction information Pre-conditions; if yes, extract the restriction information corresponding to the output candidate, and sort the candidates according to the restriction information.

 Preferably, the preset condition of the application restriction information is: whether the output candidate is a separately output word; or the preset condition of the application restriction information is: whether the output candidate belongs to a smart group word situation .

 Preferably, the limiting information is obtained by: acquiring a target word; acquiring feature information corresponding to the target word; determining whether the feature information or its corresponding calculation result meets a preset condition, and if yes, Target word record related restriction information.

Preferably, when it is required to determine whether the output candidate is a separately output word, the following steps are performed: determining whether an output candidate includes only one element, and the length is greater than one. Outputting characters; the elements are words stored in a preset vocabulary; if so, it is determined that the output candidates are words that are output separately.

 According to another embodiment of the present invention, an input method system is further provided, including an input interface unit and a display unit, the input method system further comprising:

 a vocabulary, the vocabulary includes restriction information for the vocabulary; the restriction information is used to limit the ordering when the word is outputted separately, and/or, and is used to limit the ordering when the word intelligent group word is output;

 a candidate obtaining unit, configured to obtain an output candidate according to the input information of the user; a determining unit, configured to determine whether an output candidate meets a preset condition of the application restriction information; and a candidate sorting unit, configured to meet the preset condition And extracting restriction information corresponding to the output candidate, and sorting each candidate according to the restriction information.

 Preferably, the preset condition of the application restriction information is: whether the output candidate is a separately output word; or the preset condition of the application restriction information is: whether the output candidate belongs to a smart group word situation .

 Preferably, the determining unit further includes: a subunit for determining whether an output candidate includes only one element; wherein, the element is a word stored in a preset vocabulary; and, for determining the output candidate Whether the length of the item is greater than one sub-unit of the output character; and, for determining that the output candidate is a sub-unit of the word that is output separately when the two judgment conditions are met.

 Preferably, the input interface unit, the display unit, and the vocabulary of the input method system are located in the same computing device; or the input interface unit and the display unit of the input method system are located in the first computing device, and the vocabulary is located in the second In the computing device, the input method system acquires corresponding information from the second computing device according to the information input by the user, and displays the corresponding word in the first computing device.

 Compared with the prior art, the embodiment of the invention has the following advantages:

The embodiment of the invention presets an input method vocabulary including the restriction word information, and when the user inputs, determines whether the output candidate meets the preset condition of the application restriction information; and further controls the candidate with the restriction word information according to whether the result is consistent Whether the item is displayed and output, so that it can be output more efficiently without increasing the user's operation (for example, in practice, the limit word "quantity" will not be displayed in the candidate when it is output separately. In other cases, participation in group words) greatly optimizes the character output process of the input method system and improves the intelligence of the input method system. DRAWINGS

 1 is a flow chart of steps of Embodiment 1 of a method for acquiring restriction word information according to the present invention; FIG. 2 is a flow chart of steps of Embodiment 2 of a method for acquiring restriction word information according to the present invention; FIG. 3 is an update of the present invention; FIG. 4 is a structural block diagram of an embodiment of an apparatus for obtaining restriction information according to the present invention; FIG. 5 is a flow chart of steps of an embodiment of a method for optimizing output according to the present invention;

 6 is a schematic diagram of a word grid of a pinyin network segmentation method;

 Figure 7 is a block diagram showing the structure of an embodiment of an input method system.

detailed description

 The above described objects, features and advantages of the present invention will become more apparent from the aspects of the appended claims.

 Referring to FIG. 1, a method embodiment 1 for acquiring restriction word information is shown, which may include: Step 101: Acquire a target word;

 The process of obtaining the target word can be obtained from the Internet, that is, directly obtained from the Internet corpus (for example, an Internet web page collection or a search keyword set, etc.), and can also be obtained from an existing vocabulary, and the present invention It is not necessary to be limited as long as a target word set can be obtained; as for the range size of the set, those skilled in the art can set according to actual needs.

 Preferably, for the obtained target word set, an optimization step may be further included, and some attributes of the target word are used to remove some words to further narrow the scope. For example, words from which the Internet word frequency or the lexical vocabulary frequency is less than or equal to a preset threshold are removed from the set; words that are not subject to the qualifier (e.g., general vocabulary in the dictionary) are removed from the set. Of course, the optimization step described above can also be completed in the process of acquiring the target word set.

 Step 102: Acquire feature information corresponding to the target word;

 Step 103: Determine whether the feature information or its corresponding calculation result meets a preset condition, and if yes, determine that the target word is a restriction word and record related restriction information, where the restriction information is used to limit the single output of the word. Sort the candidates.

For example, the restriction words "quantity", "previous", etc. do not appear in the candidate when outputted separately, but there is no limit when outputting with other words intelligent group words. Specific example: When typing In the case of "liangjiang", the first candidate based on the pre-output of the word frequency information is "quantity", but since it has the restriction information mark, it is removed from the candidate and is not displayed; when "lvkeliangjiangchaoguo" is input , the output candidate "The passenger volume will exceed", at this time the word "quantity" does not need to be restricted.

 The restriction words and their restriction information obtained in this embodiment may be directly stored in a separate vocabulary (or vocabulary), for example, generating a vocabulary (or vocabulary) dedicated to storing the qualifiers and The related restriction information; may also generate an input method vocabulary together with the general words, for example, generate a vocabulary, the vocabulary includes the restriction words and related restriction information, and general words; It is added to the existing vocabulary of the input method.

 The restriction information may be in the manner of identification (for example, the restriction word in the lexicon is marked with 0 or 1), or may be a specific numerical value (for example, two decimal places from 0 to 1), for The ordering of the candidates is adjusted, of course, not showing is an extreme situation. The obtained restriction words and their restriction information can be manually changed by the user according to actual needs, or it is feasible to automatically update and modify by the server.

 In the present embodiment, the corresponding judgment conditions may differ depending on the obtained feature information. Steps 102 and 103 will be described below by way of a plurality of examples. The preset corpus may be any corpus; the eigenvalues may be obtained by statistics, or may be directly obtained according to experience or existing knowledge; the eigenvalues may be various values, such as probability or frequency. It should be noted that the feature information and the judgment conditions described below are merely examples, and those skilled in the art can set more complicated feature information and judgment conditions as needed, and the present invention does not limit this.

 example 1

 The feature information is: a feature value of the word at the beginning of the target word in the default corpus as a prefix of the word, and a feature value of the word at the end of the target word in the default corpus as a suffix;

 The preset condition for determining is: whether at least one of the feature values is present in the preset range. That is, if one of the initial feature value or the last feature value is within the preset range, then the target word can be determined as a restricted word.

For example, for the word "quantity", the word "quantity" rarely appears at the beginning of the word, if the word "quantity" If the first appearance frequency is less than or equal to the preset threshold, then the "quantity" will be determined as a limit word. Of course, for a word composed of three or more words, it is also possible to judge the feature value of a word located at a certain position in the word at the same position in the word in the default corpus.

 Example 2

 The feature information is: a feature value of a linguistic collocation relationship of each single word and/or multi-word included in the target word in a preset corpus;

 The preset condition for determining is: whether at least one of the feature values is present in the preset range.

 The linguistic collocation relationship may include a collocation parameter of a word and a word, a collocation parameter of a word and a part of speech, a matching parameter of a part of speech and a part of speech, and the like. Those skilled in the art can select or apply the above various matching relationships according to actual needs.

 For example, for "is a play, the word, "yes" followed by a verb, such a collocation is rarely seen in linguistics, so the collocation feature of its collocation feature value (ie "yes + verb" can be obtained). Value) Less than or equal to the preset threshold, you can determine "is play" as a limit word.

 Example 3

 The feature information is: a feature value that the target word inputs by the user in the input method application; the preset condition for determining is: whether the feature value belongs to a preset range.

 The user input alone may be a user's statistical value, or may be a user group or may be obtained by monitoring user input behavior.

 For example, for the word "is play," the user rarely enters the word separately, so when the statistical feature value (eg, the input frequency value alone) is less than or equal to the preset threshold, then it can be determined that "is playing" In the following examples, in order to further improve the determination accuracy of the restriction words, a general word frequency is introduced in the judgment condition, and the general word frequency can be the Internet word frequency or the word library word frequency. The similarities of the examples are not described here. For details, please refer to the above.

 Example 4

The feature information includes: a word at the beginning of the word in the target word is used in a preset corpus The feature value of the prefix; the single word of the target word at the end of the word in the default corpus as the eigenvalue of the suffix; and the general word frequency of the target word;

 The preset condition for the judgment is: whether the ratio of the at least one feature value to the target word common word frequency among the above feature values belongs to a preset range.

 Example 5

 The feature information includes: a feature value of a linguistic collocation relationship of each single word and/or a multi-word word included in the target word in a preset corpus; and a general word frequency of the target word;

 The preset condition for the judgment is: whether the ratio of the at least one feature value to the target word common word frequency among the above feature values belongs to a preset range.

 Example 6

 The feature information is: user-sorted position information of the target word in each candidate word encoded for the same input; and original sorted position information of the target word in each candidate word encoded for the same input; wherein, the user The sorting information is related to the feature value separately input by the user in the input method application; the original sorting information is related to the universal word frequency of the target word; in a simple case, the user sorting information and the user vocabulary can be considered The information is related, and the original ranking information is related to the system vocabulary information.

 The preset condition for determining is: whether the difference between the user sorting position information and the original sorting position information belongs to a preset range.

 Example

 The feature information is: a feature value that the target word is input by the user in the input method application; and a general word frequency of the target word;

 The preset condition for determining is: whether the ratio of the feature value to the target word common word frequency belongs to a preset range.

 A specific implementation process of the specific description example 7 is as follows:

 Count the general word frequency of each word f- web;

 Counting the frequency of each word being entered separately in the input record of the user group f-user;

 Calculate alpha = f_user/f_web and define the word whose alpha is much smaller than the normal level as a limit word;

Or, calculate alpha = f_user/f_web , ^! Capture alpha far less than normal and f-user Words whose value is below a certain threshold are determined as limit words;

 Alternatively, calculate alpha = f-user/f_web, and identify words with alpha far below normal and f-web values greater than a certain threshold as restricted words.

 Among them, alpha is the calculation result, f-web is the general word frequency information of a word, and f-user is the characteristic word frequency information of the word.

 Specifically, for all target vocabularies, the corresponding alpha values can be calculated and sorted according to the alpha value from small to large. For words whose alpha value is at top, such as the first 5%, and the word frequency is higher, such as greater than 10000, it is considered a restriction.

 It should be noted that the determination conditions in the above respective examples may also be used in combination. In summary, those skilled in the art can set various determination methods as needed, which cannot be listed here. In a preferred embodiment of the present invention, the restriction information may include: a weight of the restriction word that is separately output in each preset scenario. That is, the restriction word can have restriction information in different application scenarios, and does not have only one general restriction information. For example, the current program of the input method determines the application scenario of the user, and when the user inputs in word, the limit information value limited to the preset scene (for example, the working term environment) is called.

 Further, the restriction information may further include: a linguistic collocation parameter of the restriction word in the preset corpus; the linguistic collocation parameter is used to limit the ordering of the word when the intelligent group word is output. That is to say, for certain restriction words, when they are output separately, they need to be restricted, and when they are outputted by intelligent group words, they also need to be restricted. For example, for the previous one, the word needs to be limited when it is output separately, and it should not appear in the candidate as much as possible, but for the "previous,, and "in" intelligent group word output, it should also be based on the collocation relationship. Restrictions, "previous" and "in" such collocations do not appear in the candidates as much as possible.

 The restriction information may include all linguistic allocation parameters (for example, part-of-speech matching parameters) of the word in the preset corpus, or may only save the required matching parameters. For example, set a threshold value for the limit output. If a collocation parameter is less than or equal to the threshold value, the linguistic collocation parameter is saved.

It should be noted that the preset corpus information may be Internet corpus information and/or user input. Enter the corpus information. The Internet corpus information may be obtained by crawling a massive webpage from a web spider through a spider; the user input recording corpus may include direct information and indirect information, for example, a character record input by a user may be used as direct information. The character distribution statistics input by the user can be used as indirect information. Of course, the preset corpus information may also be set by a person skilled in the art according to needs or experience, and the present invention does not need to be limited thereto. Referring to FIG. 2, an embodiment 2 of the method for acquiring the restriction word information is shown, which may include: Step 201: Acquire a target word;

 Step 202: Obtain a linguistic collocation parameter of the target word in a preset corpus;

 Step 203: Determine whether the linguistic collocation parameter meets a preset condition, and if yes, record the restriction information of the target word, the restriction information includes a corresponding linguistic collocation parameter, and the restriction information is used to limit the word intelligence. Sorting when group words are output.

 For example, the value of the collocation parameter of the "previous" and the position word is very low, and the collocation parameter is recorded in the restriction information of "up", and if the intelligent grouping is performed, the candidate is "previous" and orientation. The collocation of words removes the candidate.

 For another example, if the matching parameter of the "speaking" and the verb is less than the predetermined threshold, and the matching parameter is recorded in the restriction information of "speaking", if a candidate is a combination of "speaking" and a verb, then "talking" is The sequence of intelligent group words is removed.

 Preferably, the linguistic collocation parameter may be a general parameter; or the linguistic collocation parameter may also include sub-parameters for each preset scene. The linguistic collocation parameters may include collocation parameters of words and words, collocation parameters of words and part of speech, collocation parameters of part of speech and part of speech, and the like. The performance value of the linguistic collocation parameter may be adjacent co-occurrence frequency, co-occurrence probability or connection strength value, etc., and the values may be obtained from any preset corpus, or may be based on existing experience or knowledge. Get it directly.

It should be noted that, by using the above-mentioned selection step, the qualified restriction words can be removed from the sequence of the intelligent group words, thereby reducing the search space in the intelligent group words and improving the efficiency of the intelligent group words. Referring to FIG. 3, an embodiment of a method for updating an input method vocabulary is shown, which may specifically include: Step 301: Obtain a target word.

 Step 302: Acquire feature information corresponding to the target word;

 Step 303: Determine whether the feature information or its corresponding calculation result meets a preset condition. If yes, determine that the target word is a restriction word and record related restriction information, where the restriction information is used to limit the single output of the word. Sorting, and/or, used to limit the ordering of the word smart group word output;

 Step 304: Add the restriction word and its related restriction information to an existing vocabulary of the input method. This embodiment can be applied to: the server obtains the restriction word information, and then updates it to the existing vocabulary of the input method in time. The updated restriction information may include the restriction information obtained by the foregoing embodiments of FIG. 2 and FIG. 3, that is, may include information for limiting the ordering when the word is output separately, and may also include sorting for limiting the output of the word intelligent group words. Information; the two can exist separately or together. For example, the restriction information may include: a weight of the restriction word that is separately outputted in each preset scenario.

 Of course, you can also add the restriction information to the server-side lexicon on the server side, and then update the new vocabulary as a whole. The specific update transmission method will not be detailed here.

 The addition described in step 304 can be in various ways, for example,

 The adding is: determining whether the restricted word already exists in the original thesaurus, and if so, recording only the relevant restriction information into the existing thesaurus of the input method;

 Or the adding is: directly recording the restriction word and its related restriction information into an existing vocabulary of the input method, and if the vocabulary is repeated, overwriting the original vocabulary;

 Alternatively, the adding is: storing the restricted words and their related restriction information as an independent restricted vocabulary, and the restricted vocabulary and the input lexicon are used to collaboratively perform candidate ordering. Referring to FIG. 4, an apparatus for acquiring a restriction word information is shown, which may specifically include: a target word obtaining unit 401, configured to acquire a target word;

 The feature information acquiring unit 402 is configured to acquire feature information corresponding to the target word;

The restriction information obtaining unit 403 is configured to determine whether the feature information or its corresponding calculation result meets a preset condition, and if yes, determine that the target word is a restriction word and record related restriction information, where the restriction information is used to limit the Sorting when the words are output separately, and/or, used to limit the Sorting the word smart group word output. Referring to FIG. 5, an embodiment of a method for optimizing an output is shown. Specifically, the method may include: Step 501: Receive user input information, and convert the input information.

 The input information may include an encoded character string, and may also include handwritten input information as well as voice input information, since these input methods also require the use of the thesaurus for candidate ordering. That is, the present invention can be applied to input method platforms of various input methods, including keyboard symbols, handwritten information, and voice input. Since the information conversion process in these input methods is a well-known technique, it will not be described here.

 For example, when the user enters, the input method system splits the encoded string entered by the user. A brief description is given to the example of splitting a pinyin coded string. Generally, a pinyin coded string is divided into multiple segments, for example, for a pinyin encoded string.

"fangan" can be divided into "fang" an", or it can be divided into "fan'gan" and so on. Of course, the method of segmentation may be any method in the prior art, and the present invention does not need to be limited thereto.

 Step 502: Obtain an output candidate option.

 Taking a pinyin network segmentation method as an example, the process of obtaining an output candidate according to the segmented coded string is equivalent to the process of automatically converting the input continuous pinyin stream into a corresponding word stream. Specifically, the process is: For a given continuous pinyin stream A, according to a certain pinyin stream segmentation algorithm, it can be divided into a pinyin sequence Al A2 ... Am, where each pinyin Ai corresponds to one Group homophones can be represented by a set of column nodes as Wil Wi2... Wi3. Then, for the Pinyin sequence Al A2 ... Am, the corresponding candidate homophones can be represented by m group column nodes. Obviously, the candidate homophones corresponding to a pinyin sequence constitute a candidate homophone matrix. Connect adjacent nodes with directed edges to form a word grid. The word grid constitutes the state space of the Chinese character input problem. Furthermore, the word conversion problem evolves to search for an optimal path problem in the word grid.

 For example, input a pinyin stream "zheshiyizhipiaoliangdemao", which is divided into pinyin streams to generate "zhe'shi'yi'zhi, piaoliang, de, mao" pinyin sequences. The word grid corresponding to the pinyin sequence is shown in Fig. 6.

Then, query the system's language rule base, perform rule matching, and recursively bind all nodes of adjacent columns that can match a certain language rule into language element nodes to form an element mesh. This element mesh constitutes a new state space for phonetic conversion. By using the Viterbi dynamic programming algorithm, the probability values of the system's binary (Bigram) statistical library and the binary (Bigram) learning library are combined by weighting, and the probability of candidate words in all the words in the element grid is calculated. The word candidate having the greatest probability is output as a result of the phonetic word conversion.

 Of course, it is possible for a person skilled in the art to adopt any method for obtaining the output candidate, and the present invention does not need to be limited thereto.

 Step 503: Determine whether the preset condition of the application restriction information is met;

 Step 504: If yes, extract the restriction information corresponding to the output candidate, and perform sorting display on each candidate according to the restriction information.

 Sorting each candidate according to the restriction information may be implemented by directly setting a presentation position or a sequence, or by modifying a word frequency (including but not limited to weighting, derating, etc.); wherein the most extreme is Remove from candidates without displaying.

 When a word has restriction information that restricts the individual output, the preset condition of the application restriction information may be: whether the output candidate is a separately output word. And the restriction information may be obtained by acquiring the restriction information by: acquiring a target word; acquiring feature information corresponding to the target word; determining whether the feature information or its corresponding calculation result meets a preset condition, if If it matches, the relevant restriction information is recorded for the target word.

 When a word has restriction information for limiting the output of the group word, the preset condition of the application restriction information may be: whether the output candidate belongs to the intelligent group word situation. The restriction information may be obtained by: acquiring a target word; obtaining a linguistic collocation parameter of the target word in a preset corpus; determining whether the linguistic collocation parameter meets a preset condition, and if yes, The restriction information of the target word is recorded, and the restriction information includes a corresponding linguistic collocation parameter, and the restriction information is used to limit the ordering of the word intelligent group word output. Preferably, when it is required to determine whether the output candidate is a separately output word, the following steps can be completed:

For the encoded string input by the user, first obtain all possible output candidates; then, determine whether an output candidate contains only one element, and the length is greater than 1 output character; the element is a word stored in the preset vocabulary Word; if yes, determine that the output candidate is a separate input Out of the word. The judgment of whether or not an element is included may be obtained by querying from the lexicon by means of ID mapping, or by judging the number of included element IDs, it may be determined whether the output candidate contains only one element.

 The one output character may be a different byte length or other length characters in different input method systems. For example, for Chinese, Japanese, or Korean input methods, the one output character is 2 bytes. word. The determination of the length may be determined by reading a length parameter preset in the vocabulary, and the length parameter may be stored in the attribute of the corresponding term for the word ID; or, by directly acquiring the output Judging by the length of the candidate, and using other methods in the prior art are possible, and the present invention is not limited thereto.

 For example, for the case where the user inputs the encoded string "liangjiangzong", after completing the Pinyin network segmentation for the encoded string, the possible candidates are: Liangjiang Total, Volume, Liangjiang, Liangjiang, and so on. Here, it is assumed that each candidate can be expressed as <entry 1, attribute 1>, <entry 2, attribute 2>; or, <ID of the entry 1, attribute 1>, <ID of the entry 2, attribute 2 >.

 For example, for the candidate "two rivers total", it can be expressed as: <two rivers pl>, <total p2>; for the candidate "quantity", it can be expressed as: <quantity will be ql>;

 And for <quantity ql>, it contains only one element, and is greater than 1 output character; continue to judge whether its attribute ql contains a restriction information flag, since it has a restriction information flag (for example, tag is not 0), so The candidate is not output. The length parameter may also be included in the attribute ql.

 The candidate for the final output is: Liangjiang, Liangjiang, Liangjiang.

 For the general case, if a candidate is not output separately, it belongs to the group word output, so the above process can also be used to judge the situation of intelligent group words.

Of course, when the user inputs only two syllables, it is possible to directly determine that the output is separate without going through the above-described judging process, because the two syllables are generally not in the case of intelligent group words. For example, for a coded string input by the user that does not need to be segmented, it is determined that the obtained output candidate is a separately output word; or, for the output string of the user input, the output candidate corresponding to the single entry in the thesaurus, Determine the word to be output separately. Referring to FIG. 7, an embodiment of an input method system is shown, which may specifically include: an input interface unit 701 and a display unit 702, and;

 The lexicon 703: the vocabulary includes restriction information; wherein the restriction information may be various restriction information as described above; the restriction information may be present in various ways, for example, in a vocabulary manner. In the library, or by marking the corresponding terms in the thesaurus.

 The candidate obtaining unit 704 is configured to: obtain an output candidate according to the input information of the user; the determining unit 705 is configured to determine whether an output candidate meets the preset condition of the application restriction information;

 The candidate sorting unit 706 is configured to: when the preset condition is met, extract the restriction information corresponding to the output candidate, and sort the candidates according to the restriction information.

 The thesaurus 703 may include term information and restriction word information, that is, the word restriction information may be recorded in the existing thesaurus for words that meet the preset conditions. In another preferred case, the vocabulary 703 includes a basic vocabulary and a restricted vocabulary, and the restricted vocabulary is a vocabulary with restricted word information. In this case, the words that meet the preset conditions and the corresponding restriction information can be stored independently as a restricted vocabulary, and the restricted vocabulary and the basic vocabulary constitute the input method vocabulary in this embodiment. Of course, it is also feasible for a person skilled in the art to preset the input method vocabulary by using other methods in the prior art, which is not limited by the present invention.

 Preferably, when a word has restriction information limiting the individual output, the preset condition of the application restriction information may be: whether the output candidate is a separately output word. The determining unit may further include:

 a subunit for determining whether an output candidate includes only one element; wherein the element is a word stored in a preset vocabulary;

 And a subunit for determining whether the length of the output candidate is greater than one output character; and, for determining that the output candidate is a subunit of the separately outputted word when the output candidate meets the two determination conditions.

 When a word has restriction information for limiting the output of the group word, the preset condition of the application restriction information may be: whether the output candidate belongs to the intelligent group word situation. The method of determining may also adopt the foregoing method, and if it does not meet the judgment condition, it belongs to the case of intelligent group words.

The above input method system may be a common input method system, for example, the input method system is lost. The input interface unit, the display unit, and the vocabulary are located in the same computing device; the input method system may be a network input method system, for example, the input interface unit and the display unit of the input method system are located in the first computing device, and the vocabulary is located In the second computing device, the input method system acquires corresponding information from the second computing device according to the information input by the user, and displays the corresponding word candidate in the first computing device.

 Since the foregoing various embodiments are based on the same concept of the present invention, the differences are described with emphasis on each other, and similarities can be found in the corresponding parts of the specification.

 The above provides a method and device for obtaining restriction information information, a method for updating a thesaurus, a method for optimizing output, and an input method system. The description is only used to help understanding. The method of the present invention and its core idea; at the same time, for those skilled in the art, according to the idea of the present invention, there will be changes in the specific implementation manner and the scope of application. It is understood to be a limitation of the invention.

Claims

Rights request  A method for obtaining restriction word information, comprising:  Obtain a target word;  Obtaining corresponding feature information of the target word;  Determining whether the feature information or its corresponding calculation result meets a preset condition, and if so, determining that the target word is a restriction word and recording related restriction information, wherein the restriction information is used to limit the ordering when the word is separately output.  2. The method of claim 1 wherein:  The feature information is: a feature value of the word at the beginning of the target word in the default corpus as a prefix of the word, and a feature value of the word at the end of the target word in the default corpus as a suffix;  The preset condition for determining is: whether at least one of the feature values is present in the preset range.  3. The method of claim 1 wherein:  The feature information is: a feature value of a linguistic collocation relationship of each single word and/or multi-word included in the target word in a preset corpus;  The preset condition for determining is: whether at least one of the feature values is present in the preset range.  4. The method of claim 1 wherein:  The feature information is: a feature value that the target word inputs by the user in the input method application; the preset condition for determining is: whether the feature value belongs to a preset range.  5. The method of claim 1 wherein:  The feature information includes: a feature value of the word at the beginning of the target word in the preset corpus as a prefix; a word at the end of the target word in the default corpus as a feature value of the suffix; and the target word General word frequency  The preset condition for the judgment is: whether the ratio of the at least one feature value to the target word common word frequency among the above feature values belongs to a preset range.  6. The method of claim 1 wherein: The feature information includes: a language of each single word and/or multi-word included in the target word The feature value of the collocation relationship in the preset corpus; and the general word frequency of the target word; the preset condition for the judgment is: whether there is a ratio of the at least one feature value to the target word common word frequency in the feature value Belong to the preset range.  7. The method of claim 1 wherein:  The feature information is: a feature value that the target word is input by the user in the input method application; and a general word frequency of the target word;  The preset condition for determining is: whether the ratio of the feature value to the target word common word frequency belongs to a preset range.  8. The method of claim 1 wherein:  The feature information is: user-sorted position information of the target word in each candidate word encoded for the same input; and original sorted position information of the target word in each candidate word encoded for the same input; wherein, the user The sorting information is related to the feature value separately input by the user in the input method application; the original sorting information is related to the general word frequency of the target word; the preset condition for determining is: the user sorting position information Whether the difference from the original sort position information belongs to a preset range.  9. The method according to any one of claims 1-8, further comprising: before the feature information obtaining step, the step of: optimizing the step of selecting the target word.  10. The method according to any one of claims 1-8, wherein the restriction information comprises: a weight of the restriction word that is separately output in each preset scenario.  11. A method according to any of claims 1-8, characterized in that  The restriction information further includes: a linguistic collocation parameter of the restriction word in a preset corpus; the linguistic collocation parameter is used to limit the ordering of the word when the intelligent group word is output.  12. The method of any of claims 1-8, further comprising: generating a term or vocabulary, the vocabulary or vocabulary including the restricted words and their associated restriction information;  Alternatively, a vocabulary is generated, the vocabulary including the qualifiers and their associated restriction information, as well as generic words.  13. A method for obtaining restricted word information, characterized by comprising: Obtain a target word; Obtaining the linguistic collocation parameters of the target word in the default corpus;  Determining whether the linguistic collocation parameter meets a preset condition, if yes, recording restriction information of the target word, the restriction information includes a corresponding linguistic collocation parameter; the restriction information is used to limit the word intelligent group word output Sort of time.  14. The method of claim 13 wherein:  The linguistic collocation parameter is a general parameter;  Alternatively, the linguistic collocation parameters include sub-parameters for each preset scene.  15. A method of updating a thesaurus, characterized by comprising:  Obtain a target word;  Obtaining corresponding feature information of the target word;  Determining whether the feature information or its corresponding calculation result meets a preset condition, if yes, determining that the target word is a restriction word and recording related restriction information, wherein the restriction information is used to limit the ordering when the word is separately output, and / or, used to limit the sorting of the word intelligent group word output;  Adding the restriction words and their related restriction information to the existing vocabulary of the input method.  16. The method of claim 15 wherein:  The adding is: determining whether the restricted word already exists in the original thesaurus, and if so, recording only the relevant restriction information into the existing thesaurus of the input method;  Or the adding is: directly recording the restriction word and its related restriction information into an existing vocabulary of the input method, and if the vocabulary is repeated, overwriting the original vocabulary;  Alternatively, the adding is: storing the restricted words and their related restriction information as a restricted vocabulary, and the restricted vocabulary and the input lexicon are used to collaboratively perform candidate sorting.  17. The method of claim 15, wherein the restriction word has restriction information in each of the preset scenarios.  18. An apparatus for obtaining restricted word information, comprising:  a target word obtaining unit, configured to acquire a target word;  a feature information acquiring unit, configured to acquire feature information corresponding to the target word; a restriction information acquiring unit, configured to determine whether the feature information or a corresponding calculation result thereof meets a preset condition, and if yes, determine that the target word is a restriction word and record related restriction information, where the restriction information is used to limit the word Sorting when outputting separately, and/or, is used to limit the word intelligence Sorting when group words are output.  19. A method of optimizing output, comprising:  Receiving user input information, and converting the input information;  Get the output option;  Determining whether an output candidate meets a preset condition of the application restriction information;  If yes, the restriction information corresponding to the output candidate is extracted, and each candidate is sorted according to the restriction information.  20. The method of claim 19, wherein:  The preset condition of the application restriction information is: whether the output candidate is a separately output word;  Alternatively, the preset condition of the application restriction information is: whether the output candidate belongs to a smart group word situation.  21. The method according to claim 19, wherein the restriction information is obtained by the following steps:  Obtain a target word;  Obtaining corresponding feature information of the target word;  Determining whether the feature information or its corresponding calculation result meets a preset condition, and if so, recording relevant restriction information for the target word.  22. The method according to claim 20, wherein when it is required to determine whether the output candidate is a separately output word, the following steps are performed:  Determining whether an output candidate includes only one element and the length is greater than 1 output character; the element is a word stored in the preset vocabulary;  If so, it is determined that the output candidate is a separately output word.  An input method system, comprising: an input interface unit and a display unit, wherein the input method system further comprises:  a vocabulary, the vocabulary includes restriction information for the vocabulary; the restriction information is used to limit the ordering when the word is outputted separately, and/or, and is used to limit the ordering when the word intelligent group word is output; a candidate obtaining unit, configured to obtain an output candidate according to the input information of the user; and a determining unit, configured to determine whether an output candidate meets a preset condition of the application restriction information; The candidate sorting unit is configured to extract the restriction information corresponding to the output candidate when the preset condition is met, and sort the candidates according to the restriction information.  24. The system of claim 23, wherein:  The preset condition of the application restriction information is: whether the output candidate is a separately output word;  Alternatively, the preset condition of the application restriction information is: whether the output candidate belongs to a smart group word situation.  The input method system according to claim 23, wherein the determining unit further comprises:  a subunit for determining whether an output candidate includes only one element; wherein the element is a word stored in a preset vocabulary;  a subunit for determining whether the length of the output candidate is greater than one output character; and, for determining that the output candidate is a subunit of a separately outputted word when the output candidate meets the above two determination conditions.  The input method system according to claim 24, wherein the input interface unit, the display unit, and the vocabulary of the input method system are located in the same computing device;  Alternatively, the input interface unit and the display unit of the input method system are located in the first computing device, and the vocabulary is located in the second computing device, and the input method system obtains corresponding information from the second computing device according to the information input by the user. , displaying the corresponding word on the first computing device.