CN116049440A - Entity linking method, entity linking device, electronic equipment and computer readable medium - Google Patents

Abstract

The application discloses an entity linking method, an entity linking device, electronic equipment and a computer readable medium, wherein the method comprises the following steps: after determining an entity to be linked and the context information of the entity to be linked from the text to be processed, determining a first link result by utilizing at least one candidate object corresponding to the entity to be linked recorded in a link relation library, and determining a target link object of the entity to be linked according to the first link result and the context information of the entity to be linked, thereby being beneficial to improving the entity link effect.

Description

Entity linking method, entity linking device, electronic equipment and computer readable medium

Technical Field

The present disclosure relates to the field of data processing technologies, and in particular, to an entity linking method, an entity linking device, an electronic device, and a computer readable medium.

Background

An Entity Linking (Entity Linking) task is one of basic tasks in the fields of natural language processing and knowledge graph. Generally, an entity linking task may be, for example, associating an entity in a text with an object in a certain information base, so that other services (such as knowledge question-and-answer, information flow recommendation, text semantic understanding, web page recommendation, and semantic search) can be completed based on the association.

However, how to perform entity linking is a technical problem to be solved.

Disclosure of Invention

In order to solve the technical problems, the application provides an entity linking method, an entity linking device, electronic equipment and a computer readable medium.

In order to achieve the above purpose, the technical scheme provided by the application is as follows:

the application provides an entity linking method, which comprises the following steps:

determining an entity to be linked from a text to be processed and the context information of the entity to be linked;

determining a first link result corresponding to the entity to be linked by utilizing a pre-constructed link relation library; the link relation library comprises a corresponding relation between the entity to be linked and at least one candidate object; the first linking result is determined from the at least one candidate object;

and determining a target link object of the entity to be linked according to the first link result and the context information of the entity to be linked.

In a possible implementation manner, the link relation library further comprises recommendation characteristic data of each candidate object; the recommendation characterizing data of the candidate object is used for representing the possibility of associating the candidate object with the entity to be linked;

The first linking result is determined from the at least one candidate object and recommended characterizing data for the at least one candidate object.

In one possible implementation manner, the determining process of the first link result includes:

comparing the recommended characteristic data of the at least one candidate object to obtain a comparison result;

and if the comparison result indicates that the recommendation characteristic data of the first object in the at least one candidate object is the largest, generating the first link result according to the first object.

In one possible implementation, the recommended characterization data for the candidate object is determined based on historical feedback operations corresponding to the candidate object.

In one possible implementation manner, after the determining the target link object of the entity to be linked according to the first link result and the context information of the entity to be linked, the method further includes:

acquiring an actual feedback operation corresponding to the target link object;

updating recommendation characterization data corresponding to the target link object according to the actual feedback operation corresponding to the target link object;

and updating the link relation library according to the recommendation characteristic data corresponding to the target link object.

In one possible implementation manner, the determining, according to the first link result and the context information of the entity to be linked, the target link object of the entity to be linked includes:

if the context information of the entity to be linked meets the preset information quantity condition, determining a second link result corresponding to the entity to be linked according to the entity to be linked, the context information of the entity to be linked and a pre-constructed entity link model;

if the entity to be linked belongs to an acronym and the full scale information corresponding to the entity to be linked exists in the text to be processed, determining a third link result corresponding to the entity to be linked according to the full scale information;

and determining the target link object of the entity to be linked according to the first link result, the second link result and the third link result.

In one possible implementation manner, the determining process of the second link result includes:

inputting the entity to be linked and the context information of the entity to be linked into the entity link model to obtain an entity link description result output by the entity link model; the entity link description result comprises the confidence level of at least one alternative object corresponding to the entity to be linked; the at least one alternative object includes a second object and a third object; the confidence of the second object is the maximum of the confidence of the at least one candidate object; the confidence level of the third object is a second largest value in the confidence level of the at least one candidate object;

And if the confidence coefficient of the second object is larger than a first threshold value and the difference value between the confidence coefficient of the second object and the confidence coefficient of the third object is larger than a second threshold value, generating the second link result according to the second object.

In a possible implementation manner, the determining, according to the generic information, a third link result corresponding to the entity to be linked includes:

matching the full name information with at least one object to be selected in the entity word lexicon to obtain a matching result;

and if the matching result indicates that a fourth object matched with the full scale information exists in the at least one object to be selected, determining the third link result according to the fourth object.

In one possible embodiment, the method further comprises:

if the entity to be linked belongs to the abbreviation, determining the abbreviation characteristics of the entity to be linked;

generating a template to be used corresponding to the entity to be linked according to a full scale template generation rule corresponding to the abbreviation feature;

performing template retrieval processing on the text to be processed by utilizing the template to be used to obtain a retrieval result;

and determining the full scale information corresponding to the entity to be linked according to the search result.

In one possible embodiment, the method further comprises:

and if the entity to be linked meets the preset abbreviation condition, determining that the entity to be linked belongs to the abbreviation.

In a possible implementation manner, before determining the first link result corresponding to the entity to be linked by using the pre-constructed link relation library, the method further includes:

acquiring a text scene identifier of the text to be processed;

searching a link relation library corresponding to the text scene identifier from a pre-constructed mapping relation; the mapping relation comprises a corresponding relation between the text scene identification and the link relation library.

The application provides an entity linking device, comprising:

the first determining unit is used for determining an entity to be linked and the context information of the entity to be linked from the text to be processed;

the second determining unit is used for determining a first link result corresponding to the entity to be linked by utilizing a pre-constructed link relation library; the link relation library comprises a corresponding relation between the entity to be linked and at least one candidate object; the first linking result is determined from the at least one candidate object;

And the third determining unit is used for determining a target link object of the entity to be linked according to the first link result and the context information of the entity to be linked.

The application provides an electronic device, the device comprising: a processor and a memory;

the memory is used for storing instructions or computer programs;

the processor is configured to execute the instructions or the computer program in the memory, so that the electronic device performs the entity linking method provided in the present application.

The present application provides a computer readable medium having instructions or a computer program stored therein, which when run on a device, cause the device to perform the entity linking method provided herein.

The present application provides a computer program product comprising a computer program embodied on a non-transitory computer readable medium, the computer program comprising program code for performing the entity linking method provided herein.

Compared with the prior art, the application has at least the following advantages:

in the technical scheme provided by the application, after determining the entity to be linked and the context information of the entity to be linked from the text to be processed, determining a first link result corresponding to the entity to be linked by utilizing at least one candidate object corresponding to the entity to be linked recorded in a pre-constructed link relation library, and determining a target link object of the entity to be linked according to the first link result and the context information of the entity to be linked, so that the aim of entity link can be fulfilled.

Drawings

In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings that are required to be used in the embodiments or the description of the prior art will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments described in the present application, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

Fig. 1 is a flowchart of an entity linking method provided in an embodiment of the present application;

FIG. 2 is a schematic diagram of an entity linking process according to an embodiment of the present application;

fig. 3 is a schematic structural diagram of an entity linking device according to an embodiment of the present application;

fig. 4 is a schematic structural diagram of an electronic device according to an embodiment of the present application.

Detailed Description

It has been found that for an entity-linking task, in some application scenarios, the objects linked in the information base by entity words that appear in the same text in different contexts are different (e.g., linked to different paraphrasing content), which can add difficulty to the entity-linking task.

It has also been found that in some implementations of the entity linking task, entity linking disambiguation may be performed by means of a pre-built machine learning model; however, these implementations suffer from at least the following drawbacks:

(1) The processing effect for short text is not good, and the reason is as follows: for the entity word involved in the short text, the short text cannot provide sufficient context information for the entity word due to the limited information carried by the short text, so that the context of the entity word is insufficient; also, the contextual model generally needs to rely on sufficient context to perform physical link disambiguation, such that the processing performance of the machine learning model may be substantially reduced when the context is insufficient or absent, thereby rendering the machine learning model ineffective for short text processing.

(2) The treatment effect for the abbreviations is poor, the reason of which is specifically: because abbreviations may be associated with a variety of entity definitions, the above machine learning model cannot locate its exact definition.

Based on the above-mentioned research, in order to better improve the entity linking effect, the present application provides an entity linking method, which includes: after determining a to-be-linked entity and context information of the to-be-linked entity from the to-be-processed text, determining a first link result corresponding to the to-be-linked entity by utilizing at least one candidate object corresponding to the to-be-linked entity recorded in a pre-constructed link relation library, and determining a target link object of the to-be-linked entity according to the first link result and the context information of the to-be-linked entity, so that the aim of entity link can be achieved.

In addition, the entity linking scheme provided by the application can be implemented based on the three linking modes shown below, so that the entity linking scheme can effectively overcome the defects shown in the above (1) - (2), and the entity linking effect can be effectively improved.

In addition, the present application does not limit the execution subject of the above entity linking method, and for example, the entity linking method provided in the embodiment of the present application may be applied to a device having a data processing function, such as a terminal device or a server. As another example, the entity linking method provided in the embodiments of the present application may also be implemented by using a data communication procedure between different devices (for example, a terminal device and a server, two terminal devices, or two servers). The terminal device may be a smart phone, a computer, a personal digital assistant (Personal Digital Assitant, PDA), a tablet computer, or the like. The servers may be stand alone servers, clustered servers, or cloud servers.

In order to make the present application solution better understood by those skilled in the art, the following description will clearly and completely describe the technical solution in the embodiments of the present application with reference to the accompanying drawings in the embodiments of the present application, and it is apparent that the described embodiments are only some embodiments of the present application, not all embodiments. All other embodiments, which can be made by one of ordinary skill in the art without undue burden from the present disclosure, are within the scope of the present disclosure.

In order to better understand the technical solutions provided in the present application, the following description will first explain the entity linking method provided in the present application with reference to some drawings. As shown in fig. 1, the entity linking method provided in the embodiment of the present application includes the following S1-S3. Fig. 1 is a flowchart of an entity linking method according to an embodiment of the present application.

S1: and determining the entity to be linked from the text to be processed and the context information of the entity to be linked.

The text to be processed refers to any text data which needs entity identification processing and entity linking processing.

The entity to be linked refers to entity words obtained by identifying the text to be processed; and the application is not limited to the determination of the entity to be linked, for example, it may be implemented by any entity identification method existing or occurring in the future.

The context "context information of the entity to be linked" is used to describe the context that the entity to be linked has in the context to be processed; furthermore, the present application does not limit the determination process of the "context information of the entity to be linked", and for example, it may specifically be: and deleting the entity to be linked from the text to be processed to obtain the context information of the entity to be linked. As another example, it may also be: firstly deleting the entity to be linked from the text to be processed to obtain the residual content; and deleting some preset invalid characters from the residual content to obtain the context information of the entity to be linked. Wherein the preset invalid character can be preset; and the present application is not limited to this preset invalid character, and for example, it may include characters such as punctuation marks, numerals, and the like.

S2: determining a first link result corresponding to an entity to be linked by utilizing a pre-constructed link relation library; the link relation library comprises a corresponding relation between the entity to be linked and at least one candidate object; the first linking result is determined from the at least one candidate object.

The link relation library is used for describing the use condition of a link object presented by a target user aiming at some entity words in a historical time period, so that the link relation library can represent the use state of the link presented by the target user in the historical time period. The target user refers to a user of the display apparatus of the "target link object of the entity to be linked" hereinafter. The history period refers to a period of time before the execution timing of the above S1. The link object refers to an object which needs to establish an association relation with the entity word; and the present application is not limited to the link object, for example, in some application scenarios the link object may refer to a paraphrase content that exists in a lexicon (e.g., the "entity word lexicon" below).

In fact, for the "link relation library" above, the link relation library may include at least correspondence between the mth entity word and some historical link objects corresponding to the mth entity word. The historical link objects refer to link objects provided for the target user aiming at the mth entity word in a historical time period, so that the historical link objects are presented to the target user once as link objects associated with the mth entity word. M is a positive integer, M is less than or equal to M, M is a positive integer, and M represents the number of entities involved in the link relation library.

In fact, in order to better improve the entity linking effect, the present application further provides a possible implementation manner of the above linking relation library, where in this implementation manner, the linking relation library includes not only the corresponding relation between the mth entity word and some historical linking objects corresponding to the mth entity word, but also recommendation characterizing data of each historical linking object. Wherein the recommendation characterizing data of the kth historical link object is used for characterizing the possibility of associating the kth historical link object with the mth entity word (namely, the possibility of recommending the kth historical link object as the link object associated with the mth entity word to the target user), and the greater the value of the recommendation characterizing data of the kth historical link object, the greater the possibility of associating the kth historical link object with the mth entity word. K is a positive integer, K is less than or equal to K, K is a positive integer, and K represents the number of objects in a plurality of history link objects corresponding to the mth entity word.

In addition, regarding the "recommendation characterizing data of the kth historical link object" above, the "recommendation characterizing data of the kth historical link object" may be determined according to the historical feedback operation corresponding to the kth historical link object. The "historical feedback operation corresponding to the kth historical link object" refers to a feedback operation triggered by the target user for the kth historical link object when the kth historical link object is recommended to the target user as the link object associated with the mth entity word in a historical time period, so that the feedback operation can indicate whether the kth historical link object is a correct object expected by the target user.

It should be noted that the present application is not limited to the embodiment of the "feedback operation" in the upper stage, and for example, it may be implemented in two feedback modes shown in (one) - (two) below.

The dominant feedback mode is specifically as follows: the target user explicitly expresses by some operations whether the "kth historical link object" above is the correct object that it expects. The present application is not limited to these operations, and may be, for example, a praise operation or a click operation. Wherein the praise operation is to express explicitly that the "kth history link object" is the correct object it expects (i.e., explicit correct); but the click operation is used to express explicitly that the "kth history link object" is not the correct object it expects (i.e., an explicit error).

And (II) invisible feedback mode, which is specifically as follows: after the target user receives the "kth historical link object" fed back for the "mth entity word" above, if the target user actively views the "kth historical link object" or opens the relevant document of the "kth historical link object", then the stealth is considered to be correct; however, if the target user switches to view other link objects and their related content, then it is a stealth error.

It should also be noted that the present application is not limited to the determination process of the "recommendation characterizing data of the kth historical link object" above, and for example, it may be implemented using the following formula.

Recommending characterization data _k =cumulative number of links _k +feedback correct times _k X weight _{Correct and correct} -number of feedback errors _k X weight _Errors

In the formula, the above "recommends characterization data _k "recommendation characterizing data representing the kth historical linking object" above; the above "cumulative number of links _k "means the number of times, in the history period, when the kth history link object is recommended to the target user as the link object associated with the mth entity word above; the above "feedback of correct times _k "means the number of times the target user gives correct feedback for the kth historical linking object during the historical period; the above "feedback error times _k "means the number of times the target user gives error feedback for the kth historical link object during the historical period; the' accumulated number of links _k "=the" feedback correct number of times _k "+the number of feedback errors _k "; the above "weights _{Correct and correct} "means to be directed to the" in advance "Feeding back the correct times _k "set weighting weight value; the above "weights _Errors "refers to the number of feedback errors for this" in advance _k "set weighting weight value".

Based on the related content of the "link relation library", a plurality of historical link objects corresponding to a large number of entity words and recommendation characterization data of the historical link objects under the entity words can be recorded in the link relation library, so that the link relation library can show the use feedback results of the target user for the historical link objects in a historical time period, the link relation library can better show the link use states of the target user for the entity words, and further, the entity link results conforming to the link use conditions of the target user can be provided based on the link relation library.

In fact, in different text scenes (for example, scenes of document writing, instant messaging, etc.), the link usage conditions presented by the target user for the same entity word may be different, so in order to better improve the entity link effect, the present application further provides the determination process of the above "link relation library", which may specifically include the following steps 11-12.

Step 11: and acquiring a text scene identifier of the text to be processed.

The text scene identifier is used for representing a text scene to which the text to be processed belongs; furthermore, the present application is not limited to the determination process of the text scene identifier, for example, if the text to be processed is a character string extracted from a document, the text scene identifier of "document" may be determined as the text scene identifier of the text to be processed. For another example, if the text to be processed is a character string extracted from a conversation page of the instant messaging software, the text scene identifier of "instant messaging" may be determined as the text scene identifier of the text to be processed.

It should be noted that the execution time of the above step 11 is not limited in this application, and for example, the step 11 may be executed after the above "text to be processed" is acquired.

Step 12: searching a link relation library corresponding to the text scene identifier from a pre-constructed mapping relation; the mapping relation comprises a corresponding relation between the text scene identifier and the link relation library.

The mapping relation is used for recording a link relation library corresponding to the text scene identifications.

In addition, the present application does not limit the above mapping relationship, and for example, it may include a correspondence relationship between the y-th text scene identifier and the y-th link relationship library. Wherein the y text scene identifier is used for uniquely identifying the y text scene; the y-th link relation library is determined based on some entity words in the y-th text scene, link objects provided for the entity words to the target user, historical feedback operations of the link objects and the like, so that the y-th link relation library can represent the link use condition presented by the target user in the y-th text scene. Y is a positive integer, Y is less than or equal to Y, Y is a positive integer, and Y represents the number of text scene identifications involved in the mapping relation.

Based on the related content in the steps 11 to 12, if entity identification and entity linking processing are required for the text to be processed, determining a text scene identifier of the text to be processed after the text to be processed is acquired, so that the text scene identifier can indicate which text scene the text to be processed is taken from; judging whether the text scene identifier is the same as the y text scene identifier in the mapping relation, if so, directly determining a y-th link relation library corresponding to the y-th text scene identifier as a link relation library corresponding to the text to be processed, so that the link result of any entity word in the text to be processed can be determined in an auxiliary manner by the corresponding link relation library; if the two types are different, discarding the two types. Wherein Y is a positive integer, and Y is less than or equal to Y.

The above "candidate object" refers to a link object that exists in the above link relation library and has a correspondence relation with the above entity to be linked.

In addition, the present application does not limit the determination process of the "at least one candidate object" above, for example, when the above link relation library includes correspondence relations between the mth entity word and some historical link objects corresponding to the mth entity word, the determination process may specifically be: after the entity to be linked is obtained, comparing the entity to be linked with the m-th entity word in the link relation library, if the comparison result shows that the entity to be linked is the same as the m-th entity word, determining all the historical link objects corresponding to the m-th entity word as candidate objects corresponding to the entity to be linked, so that the link result of the entity to be linked can be determined from the candidate objects. Wherein M is a positive integer, and M is less than or equal to M.

The above "first link result" refers to a link result determined for the above entity to be linked using the above link relation library. For example, the "first link result" may refer to "link result determined based on the history behavior" shown in fig. 2.

In addition, the present application does not limit the determination process of the "first link result" above, for example, when the link relation library includes the correspondence between the entity to be linked and at least one candidate object, the determination process may specifically be: one candidate object is randomly selected from the candidate objects, and the first link result is determined.

For another example, when the above link relation library includes a correspondence between the entity to be linked and at least one candidate object, and recommendation characterization data of each candidate object, the above determination process of the "first link result" may specifically be: and determining a first link result corresponding to the entity to be linked according to the at least one candidate object and the recommended characterization data of the at least one candidate object. Wherein the "candidate recommendation characterizing data" is used to represent a likelihood of associating the candidate object with the entity to be linked; and the "recommended characterizing data of the candidate object" may be determined based on the historical feedback operations corresponding to the candidate object. In addition, the relevant content of the "recommendation characterizing data of the candidate object" is similar to the "recommendation characterizing data of the kth historical link object" above, and for brevity, will not be described herein.

To facilitate an understanding of the foregoing, a description is provided below in connection with examples.

As an example, when the above link relation library includes the correspondence between the above entity to be linked and at least one candidate object, and the recommendation characterizing data of each of the candidate objects, the above determination process of the "first link result" may specifically include the following steps 21 to 22.

Step 21: comparing the recommended characterizing data of at least one candidate object to obtain a comparison result, so that the comparison result can represent the relative size between the recommended characterizing data of the candidate objects.

It should be noted that the embodiment of the above step 21 is not limited in this application, and for example, it may specifically be a size comparison result between the recommended characterizing data including any two objects of the above "at least one candidate object". As another example, in some application scenarios, the step 21 may specifically be: and sequencing the recommended characteristic data of at least one candidate object according to the sequence from large to small to obtain a comparison result, so that the comparison result can represent the relative size between the recommended characteristic data of the candidate objects.

Step 22: and if the comparison result indicates that the recommendation characteristic data of the first object in the at least one candidate object is maximum, generating a first link result according to the first object.

Wherein the first object refers to the candidate with the largest recommendation characterizing data in the above "at least one candidate"; and the present application is not limited to this first object, which may refer to paraphrasing content, for example.

In addition, the embodiment of the step "generating the first link result according to the first object" in the step 22 is not limited in this application, and may specifically be, for example: and directly determining the first object as the first link result, so that the first link result can represent that the object which is determined based on the link relation library and can be recommended to the target user as the link object associated with the entity to be linked is the first object.

Based on the related content of S2 above, after identifying the entity to be linked from a text, a plurality of candidate objects having a corresponding relationship with the entity to be linked may be searched from a pre-constructed link relationship library; then, ranking the candidate objects by using the recommendation characteristic data of the candidate objects recorded in the link relation library to obtain ranking results, so that the ranking results can represent ranking conditions of the candidate objects according to ranking of the recommendation characteristic data from large to small; finally, the candidate object with the forefront ranking sequence is determined as a first link result (for example, the link result determined based on the historical behavior shown in fig. 2) corresponding to the entity to be linked, so that the first link result can represent the object which is determined based on the above link relation library and can be recommended to the target user as the link object associated with the entity to be linked, and the purpose of entity link processing by means of the historical behavior of the target user can be achieved.

S3: and determining a target link object of the entity to be linked according to the first link result and the context information of the entity to be linked.

To facilitate understanding of S3, the following description is made in connection with an example.

As an example, the above S3 may specifically include the following steps one to three.

Step one: if the context information of the entity to be linked meets the preset information quantity condition, determining a second link result corresponding to the entity to be linked according to the entity to be linked, the context information of the entity to be linked and a pre-constructed entity link model.

The preset information amount condition may be preset, for example, it may specifically be: the number of effective characters in the context information of the entity to be linked is larger than a preset number threshold. The valid characters refer to the characters which exist in the context information of the entity to be linked and carry semantics; the present application is not limited to this determination process of valid characters, and may specifically be, for example: after the "context information of the entity to be linked" is obtained, some preset invalid characters (e.g., punctuation marks, numbers, etc.) are deleted from the "context information of the entity to be linked", and valid characters in the "context information of the entity to be linked" are obtained.

The above "entity link model" refers to a machine learning model with entity link disambiguation function, which is built in advance; and the application is not limited to this physical link model, and for example, it may be implemented using any machine learning model with physical link disambiguation function that exists in the present or future.

The above "second link result" refers to a link result determined for an entity to be linked using the above entity link model. For example, the "second link result" refers to the "link result determined based on the model" shown in fig. 2.

In addition, the present application is not limited to the above determination of the "second link result", and for example, it may be implemented using any method that can perform entity link processing based on a model, existing or occurring in the future.

Indeed, in order to better enhance the physical linking effect, the present application also provides a determination process of the above "second linking result", which may specifically include the following steps 31-32.

Step 31: if the context information of the entity to be linked meets the preset information quantity condition, inputting the entity to be linked and the context information of the entity to be linked into an entity link model to obtain an entity link description result output by the entity link model. The entity link description result comprises the confidence level of at least one alternative object corresponding to the entity to be linked; the at least one alternative object includes a second object and a third object; the confidence of the second object is the maximum of the confidence of the at least one candidate object; the confidence of the third object is the second largest value in the confidence of the at least one candidate object.

The entity link description result refers to an entity link result given by the entity link model aiming at an entity to be linked; moreover, embodiments of the present application do not limit the entity link description results, which may include, for example: at least one candidate object and a confidence level of each candidate object.

The t-th alternative object refers to a link object determined by the entity link model aiming at the entity to be linked; and the present application is not limited to this t-th alternative object, which may be paraphrased content, for example. Wherein T is a positive integer, T is less than or equal to T, T is a positive integer, and T represents the number of objects in the above 'at least one candidate object'.

The confidence of the t-th candidate object is used for representing the possibility of using the t-th candidate object as the link object associated with the entity to be linked; and the higher the "confidence of the t-th candidate object", the greater the likelihood that the t-th candidate object is indicated as the link object associated with the entity to be linked. Wherein T is a positive integer, and T is less than or equal to T.

The above "second object" refers to the candidate object having the greatest confidence in the above "at least one candidate object".

The above "third object" refers to the candidate object having the second greatest confidence in the above "at least one candidate object". That is, the confidence of the third object is lower than the confidence of the second object above, and the confidence of the third object is not lower than the confidence of the other entities than the second object in the "at least one candidate object" above.

Based on the above information about step 31, for the entity to be linked, after the entity to be linked and the context information of the entity to be linked are obtained, the entity to be linked and the context information thereof may be input into a pre-constructed entity link model, so that the entity link model performs entity link disambiguation processing on the entity to be linked, to obtain and output an entity link description result, so that the entity link description result may represent the link disambiguation arrangement sequence corresponding to the entity to be linked and the confidence level of at least one candidate object corresponding to the entity to be linked, so that the link object associated with the entity to be linked may be determined based on the two output contents. The disambiguation arrangement sequence is used for describing arrangement conditions of all the candidate objects according to confidence from large to small.

The "second object" refers to an object with the forefront arrangement position in the upper section "disambiguation arrangement order"; the above "third object" refers to an object located at the second arrangement position in the upper-stage "disambiguation arrangement order".

Step 32: if the confidence coefficient of the second object is larger than the first threshold value, and the difference value between the confidence coefficient of the second object and the confidence coefficient of the third object is larger than the second threshold value, generating a second link result according to the second object.

The first threshold may be preset; and the second threshold value may be set in advance.

In addition, the present application is not limited to the embodiment of the step "generating the second link result according to the second object" in the step 32 above, and may specifically be, for example: and determining the second object as a second link result, so that the second link result can represent the object which is determined based on the entity link model and can be recommended to the target user as the link object associated with the entity to be linked, and the aim of carrying out entity link processing by means of the machine learning model can be fulfilled.

Based on the above related content in steps 31 to 32, for the entity to be linked, the entity link disambiguation process may be performed on the entity to be linked by using a pre-built entity link model, so as to obtain the link disambiguation arrangement sequence corresponding to the entity to be linked and the confidence level of at least one candidate object corresponding to the entity to be linked; and determining a second object with highest confidence and a third object with second highest confidence based on the link disambiguation sequence, so that when the confidence of the second object is determined to be larger than a first threshold and the difference between the confidence of the second object and the confidence of the third object is determined to be larger than a second threshold, the link object which is related to the entity to be linked by the second object can be determined to be reliable and difficult to cause ambiguity, and the second object can be directly determined to be a second link result, so that the second link result can represent the object which is determined based on the entity link model and can be recommended to a target user as the link object which is related to the entity to be linked, and the aim of entity link processing by means of the machine learning model can be realized.

Indeed, to better enhance the physical linking effect, the present application also provides another determination of the above "second linking result", which may specifically include the following steps 41-44.

Step 41: after obtaining the context information of the entity to be linked, judging whether the context information meets the preset information quantity condition, if so, executing the following steps 42-44; if not, determining a first preset result (for example, a character string of "no result") as a second link result corresponding to the entity to be linked.

The first preset result is used for showing that the link result of one entity word cannot be accurately determined by means of the entity link model; and the present application is not limited to the implementation of the first preset result, and for example, it may be implemented using a character string of "no result".

Based on the above information about step 41, it is known that, for the entity to be linked, after obtaining the context information of the entity to be linked, if it is determined that the context information does not meet the preset information amount condition, it may be determined that the context information cannot provide sufficient context for the entity to be linked, so that it may be determined that the entity link model gives an accurate link result for the entity to be linked due to the fact that the entity link model cannot obtain sufficient context, and thus, a preset first preset result may be determined as a second link result corresponding to the entity to be linked, so that the second link result may indicate that the link result of the entity to be linked cannot be accurately determined by means of the entity link model.

Step 42: and when the context information of the entity to be linked meets the preset information quantity condition, inputting the entity to be linked and the context information of the entity to be linked into an entity link model to obtain an entity link description result output by the entity link model. The entity link description result comprises the confidence level of at least one alternative object corresponding to the entity to be linked; the at least one alternative object includes a second object and a third object; the confidence of the second object is the maximum of the confidence of the at least one candidate object; the confidence of the third object is the second largest value in the confidence of the at least one candidate object.

It should be noted that, for brevity, reference is made to step 31 above for the relevant content of step 42, and further description is omitted here.

Step 43: if the confidence coefficient of the second object is larger than the first threshold value, and the difference value between the confidence coefficient of the second object and the confidence coefficient of the third object is larger than the second threshold value, generating a second link result according to the second object.

It should be noted that, for brevity, please refer to the above step 32 for the relevant content of step 43, and further description is omitted here.

Step 44: if the confidence coefficient of the second object is not greater than the first threshold value, or the difference between the confidence coefficient of the second object and the confidence coefficient of the third object is not greater than the second threshold value, determining the first preset result as a second link result corresponding to the entity to be linked.

In the application, if the confidence coefficient of the second object is not greater than the first threshold, it may be determined that the second object is not reliable as the link object associated with the entity to be linked, so that the first preset result may be directly determined as the second link result corresponding to the entity to be linked, so that the second link result may indicate that the link result of the entity to be linked cannot be accurately determined by means of the entity link model; in addition, if it is determined that the difference between the confidence level of the second object and the confidence level of the third object is not greater than the second threshold, it may be determined that ambiguity is likely to occur when determining the link object associated with the entity to be linked based on the above entity link description result, so the first preset result may be directly determined as the second link result corresponding to the entity to be linked, so that the second link result may indicate that the link result of the entity to be linked cannot be accurately determined by means of the above entity link model.

Based on the above related content in steps 41 to 44, it can be known that, if the entity to be linked cannot accurately give the link object associated with the entity to be linked by using the pre-constructed entity link model, the pre-set first preset result can be directly determined as the second link result corresponding to the entity to be linked, so that the second link result can indicate that the link result of the entity to be linked cannot be accurately determined by using the entity link model.

Based on the above description of S3, after obtaining the entity to be linked and the context information of the entity to be linked, a second linking result (for example, "linking result determined based on the model" shown in fig. 2) corresponding to the entity to be linked may be determined by means of a pre-constructed entity linking model, so that the second linking result may represent the linking result given by the entity linking model for the entity to be linked, and thus the purpose of performing entity linking processing by means of a machine learning model may be achieved.

Step two: if the entity to be linked belongs to an abbreviation and the full scale information corresponding to the entity to be linked exists in the text to be processed, determining a third link result corresponding to the entity to be linked according to the full scale information.

Wherein abbreviations may be used to represent a phrase (e.g., each of the phrases shown in table 1 below).

Table 1 some phrases and their corresponding abbreviations

In fact, since the abbreviations generally conform to a certain format, in order to better enhance the recognition effect of the abbreviations, the present application also provides a recognition process of the abbreviations, which may specifically include the following steps 51-53.

Step 51: after obtaining the entity to be linked, it may be determined whether the entity to be linked meets a preset abbreviation condition, if yes, the following step 52 is executed; if not, the following step 53 is performed.

Wherein, the preset abbreviation condition is used for representing the format characteristics shared by all abbreviations; moreover, the present application is not limited to the preset abbreviation condition, for example, in some application scenarios, the preset abbreviation condition may specifically be: the abbreviations should consist of english, numerals, specific symbols only, and the first characters are not symbols. As another example, in other application scenarios, the preset abbreviation condition may also be: the abbreviation shall consist of english, numbers, specific symbols only, consecutive numbers not exceeding 1 bit, and the first character not being a symbol. It can be seen that, in one possible implementation, the preset abbreviation condition may be preset according to a specific application scenario.

Step 52: if the entity to be linked meets the preset abbreviation condition, the entity to be linked can be determined to belong to the abbreviation.

Step 53: if the entity to be linked does not meet the preset abbreviation condition, the entity to be linked can be determined not to belong to the abbreviation.

Based on the above related content in steps 51 to 53, after the entity to be linked is obtained, it may be determined whether the entity to be linked meets a preset abbreviation condition, if so, it may be determined that the entity to be linked belongs to an abbreviation, so that whether the full-name information corresponding to the entity to be linked exists may be queried from the above text to be processed.

The phrase "the entity to be linked corresponds to" refers to the phrase represented by the entity to be linked. For example, the entity to be linked is the abbreviation "URL" above, and the full term information "corresponding to the entity to be linked is the phrase" uniform resource locator "above.

It has been found that there is a correspondence shown in the following (1) to (3) between an abbreviation and its corresponding full name information.

(1) When an abbreviation has the feature of "pure english", there is a correspondence relationship shown below (1.1) - (1.2) between the abbreviation and its corresponding full name information.

(1.1) if a special lowercase form does not exist in the middle of an abbreviation, the number of fully-called words corresponding to the abbreviation corresponds to the number of letters of the abbreviation, and any connectors (e.g., space, underline, dash-, and the presence of a mixture of connectors, etc.) may be used between the fully-called words. For example, when the abbreviation is the abbreviation "URL" above, the full scale information corresponding to the abbreviation conforms to the templates shown in table 2 below.

Abbreviations	URL
		Full scale template 1	Uxxx Rxxx Lxxx
Full scale template 2	Uxxx-Rxxx-Lxxx
		Full scale template 3	Uxxx_Rxxx_Lxxx
Full scale template 4	Uxxx-Rxxx Lxxx

TABLE 2 correspondence between abbreviations and full name templates

Note that "xxx" referred to in table 2 above may be used to represent a character string of any length.

(1.2) if a special lowercase form exists in the middle of an abbreviation, it can be distinguished from both uppercase, and the lowercase is considered to be a new word form formed by partial nouns and abbreviations, and for the uppercase it can be processed in the manner as shown in (1.1) above. For example, when the abbreviation is the abbreviation "WebRTC" above, the full scale information corresponding to the abbreviation conforms to the templates shown in table 3 below.

Abbreviations	WebRTC
		Full scale template 1	Web Rxxx Txxx Txxx
Full scale template 2	Web Rxxx-Txxx-Txxx
		Full scale template 3	Web Rxxx_Txxx_Txxx
Full scale template 4	Web Rxxx-Txxx Txxx

TABLE 3 correspondence between another abbreviation and full name template

Note that "xxx" referred to in table 3 above may be used to represent a character string of any length.

(2) When the abbreviation has the characteristic of 'English + symbol', the number of the whole words corresponding to English parts in the abbreviation is consistent with the number of the letters of the abbreviation, the abbreviation content symbols are reserved in the whole names, and the English part link symbols are distinguished from the abbreviation symbols. For example, when the abbreviation is the abbreviation "TFT-LCD" above, the full scale information corresponding to the abbreviation corresponds to the templates shown in table 4 below.

Abbreviations	TFT-LCD
		Full scale template 1	Txxx Fxxx Txxx-Lxxx Cxxx Dxxx
Full scale template 2	Txxx_Fxxx_Txxx-Lxxx_Cxxx_Dxxx

TABLE 4 correspondence between yet another abbreviation and full term template

Note that "xxx" referred to in table 4 above may be used to represent a character string of any length.

(3) When an abbreviation has the feature of "english+number", there is a correspondence relationship shown below (3.1) to (3.2) between the abbreviation and its corresponding full name information.

(3.1) if the abbreviation has the feature of "single number + letter", the numerical meaning of the abbreviation is combined to form a full name for the english beginning word of the repeated number of digits, also using connector concatenation. For example, when the abbreviation is the abbreviation "4S" above or the abbreviation "W3C" above, the full scale information corresponding to the abbreviation corresponds to the templates shown in table 5 below.

TABLE 5 correspondence between a further abbreviation and a full term template

Note that "xxx" referred to in table 5 above may be used to represent a character string of any length.

(3.2) if the abbreviation has the feature of "consecutive numbers+letters", the numerical meaning of the abbreviation is a plurality of english letter strings, and is not associated with the other parts of the abbreviation, only the string length is indicated. For example, when the abbreviation is the abbreviation "i18n" above, the full scale information corresponding to the abbreviation corresponds to the templates shown in table 6 below.

TABLE 6 correspondence between a further abbreviation and a full term template

Note that "xxx" referred to in table 6 above is used to denote a character string composed of 18 letters.

Based on the above description of (1) to (3), it is known that different types of abbreviations have different abbreviation features, and that different types of abbreviations correspond to different full scale structures (i.e., full scale templates). Based on this, the present application also provides a determination process of the above "generic information corresponding to the entity to be linked", which may specifically include the following steps 61-64.

Step 61: if the entity to be linked belongs to an abbreviation, determining an abbreviation feature of the entity to be linked, so that the abbreviation feature can represent an abbreviation state (for example, any one of the abbreviation states shown in (1) to (3) above) presented by the entity to be linked.

It should be noted that the determination process of the "abbreviated feature of the entity to be linked" above is not limited in this application, and may be implemented by means of a machine learning model having an abbreviated feature extraction function, which is constructed in advance, for example. As another example, it may be implemented by means of pre-set abbreviated feature decision rules (e.g., regular expressions, etc.).

Step 62: and generating a template to be used corresponding to the entity to be linked according to the full-scale template generation rule corresponding to the abbreviation characteristics.

The full scale template generation rule refers to a rule which is preset for the abbreviation feature of the entity to be linked and is used for generating the full scale template; and the present application does not limit the rule for generating a full-scale template, for example, when the "abbreviation feature of the entity to be linked" is the abbreviation feature shown in the above (1.1) (i.e., the two features of "pure english" and "no special lowercase form in the middle of the abbreviation"), the rule for generating a full-scale template may be the rule shown in the above (1.1) (i.e., the rule that the number of full-scale words corresponding to the "abbreviation corresponds to the number of letters of the abbreviation, and any connector" may be used between the full-scale words).

The above "templates to be used" refers to the generic templates generated for the above entities to be linked; also, the present application is not limited to this "template to be used", and for example, when the above "entity to be linked" is the abbreviation of the above "URL", the "template to be used" may include four commonly called templates shown in table 2 above.

Step 63: and carrying out template retrieval processing on the text to be processed by utilizing the template to be used to obtain a retrieval result.

The template retrieval process is used for determining a character string conforming to a certain template from text data; the present application is not limited to the embodiment of the template search process, and may be implemented by, for example, string matching, regular matching, dictionary matching, or the like.

The search result is used for indicating whether the character strings conforming to the templates to be used exist in the text to be processed.

Step 64: and determining the full scale information corresponding to the entity to be linked according to the retrieval result.

It should be noted that the present application is not limited to the embodiment of step 64, and for example, it may include the following steps 641 to 643.

Step 641: if the above search result indicates that the text to be processed does not have a character string conforming to the template to be used, it may be determined that the text to be processed does not have the full name information corresponding to the entity to be linked, so that a second preset result (for example, a character string of "no result") set in advance may be determined as a third link result corresponding to the entity to be linked, so that the third link result may indicate that the link object associated with the entity to be linked cannot be determined by means of the strong related information.

The second preset result is used for showing that the link result of one entity word cannot be determined by means of the strong related information; and the present application is not limited to the implementation of the second preset result, and for example, it may be implemented using a character string of "no result".

Step 642: if the above search result indicates that only one character string conforming to the template to be used exists in the text to be processed, the character string is used as full scale information corresponding to the entity to be linked, and a third link result corresponding to the entity to be linked is determined according to the full scale information.

Step 643: if the search result indicates that a plurality of character strings conforming to the templates to be used exist in the text to be processed, full scale information corresponding to the entities to be linked is screened out from the character strings according to a preset selection rule, and a third link result corresponding to the entities to be linked is determined according to the full scale information.

The preset selection rule may be preset according to an application scenario, and for convenience of understanding, the following description will be made with reference to steps 71-73.

As an example, when the above search result indicates that there are E character strings conforming to the template to be used in the text to be processed, the above determination process of "the full name information corresponding to the entity to be linked" may include the following steps 71 to 73.

Step 71: and judging whether preset indication words exist before and after the position of the e-th character string in the text to be processed, and obtaining a judging result corresponding to the e-th character string. Wherein E is a positive integer, E is less than or equal to E, and E is a positive integer.

The preset indicator may be preset, for example, it may be: "full," "abbreviation," "representation," or "meaning" and the like.

Step 72: and calculating the distance between the position of the e-th character string in the text to be processed and the position of the entity to be linked in the text to be processed, and obtaining the distance corresponding to the e-th character string. Wherein E is a positive integer, E is less than or equal to E, and E is a positive integer.

Step 73: and selecting an optimal character string from the E character strings according to the judging result and the distance corresponding to the E character strings, and taking the optimal character string as full name information corresponding to the entity to be linked.

It should be noted that, the filtering rule of the "optimal character string" above is: character strings with preset indication words and smaller distances are preferentially selected; and the screening rules may be preset.

Based on the above related content in steps 61 to 64, for the entity to be linked, after determining that the entity to be linked belongs to an abbreviation, the full scale template corresponding to the entity to be linked may be determined according to the abbreviation feature of the entity to be linked; then, according to the full-scale template, carrying out template retrieval processing on the text to be processed to obtain a retrieval result; and finally, determining the full scale information corresponding to the entity to be linked according to the search result so that a third link result corresponding to the entity to be linked can be determined based on the full scale information.

The above "third link result" refers to a link result determined for an entity to be linked by means of strong association information (e.g., full name information). For example, the "third link result" may be "link result determined based on strong correlation information" shown in fig. 2.

In addition, the present application does not limit the determination process of the "third link result" above, and for example, when the entity linking method provided in the present application is used to associate an entity to be linked with a certain object (for example, paraphrased content, etc.) in the entity word stock, the determination process of the "third link result" may specifically include the following steps 81-83.

Step 81: and matching the full name information corresponding to the entity to be linked with at least one object to be selected in the entity word stock to obtain a matching result.

The entity word library refers to a database to which a link object associated with the entity to be linked belongs; and the application is not limited to the entity word library, and for example, the application can be implemented by using any database which exists in the prior art or exists in the future and needs to be physically linked. It should be noted that all objects referred to in this application come from the entity word stock.

The above "object to be selected" refers to any one of the objects that exist in the above "entity word lexicon" and can be associated by the entity word; and the present application is not limited to this "candidate object", for example, it may refer to paraphrase content.

The above "matching result" is used to indicate whether there is a to-be-selected object matching the above "full scale information corresponding to the to-be-linked entity" in the above "entity word stock"; and the application does not limit the specific meaning of the match, for example, when all objects referred to in the application refer to paraphrase content, the specific meaning of the match may be: if the "full name information corresponding to the entity to be linked" or the "full name information corresponding to the entity to be linked" appears in the q-th object to be selected as the alias of the q-th object to be selected, it can be determined that the q-th object to be selected is matched with the "full name information corresponding to the entity to be linked"; otherwise, the two are not matched. Wherein Q is a positive integer, Q is less than or equal to Q, Q is a positive integer, and Q represents the number of objects to be selected in the entity word stock.

Step 82: if the above matching result indicates that a fourth object matched with the full scale information exists in at least one object to be selected, determining a third linking result corresponding to the entity to be linked according to the fourth object.

The fourth object refers to a candidate object which exists in the entity word stock and is matched with the full name information corresponding to the entity to be linked.

In addition, the present application is not limited to the embodiment of the step "determining the third link result according to the fourth object" in the step 82 above, and may specifically be, for example: and determining the fourth object as a third link result corresponding to the entity to be linked.

Step 83: if the above matching result indicates that the above at least one object to be selected does not have the object to be selected matched with the full scale information, determining the above second preset result as a third link result corresponding to the above entity to be linked, so that the third link result can indicate that the link object associated with the entity to be linked cannot be determined by means of the strong related information.

Based on the related content of the above steps 81 to 83, if the entity to be linked belongs to an abbreviation and the corresponding full scale information of the entity to be linked exists in the above text to be processed, the third linking result corresponding to the entity to be linked can be determined according to the full scale information.

Indeed, in order to better enhance the physical linking effect, the present application also provides another possible implementation of the above "third linking result", which may specifically include the following steps 91-93.

Step 91: if the entity to be linked does not belong to the abbreviation, determining the second preset result as a third link result corresponding to the entity to be linked.

Step 92: if the entity to be linked belongs to the abbreviation, but the full scale information corresponding to the entity to be linked does not exist in the text to be processed, determining the second preset result as a third link result corresponding to the entity to be linked.

Step 93: if the entity to be linked belongs to an abbreviation and the full scale information corresponding to the entity to be linked exists in the text to be processed, determining a third link result corresponding to the entity to be linked according to the full scale information.

Based on the related content of the above steps 91 to 93, it can be known that, for the entity to be linked, if it is determined that the entity to be linked does not belong to an abbreviation, or it is determined that the entity to be linked belongs to an abbreviation, but the full scale information corresponding to the entity to be linked does not exist in the above text to be processed, it can be determined that the strong related information of the entity to be linked does not exist in the text to be processed, so that it can be determined that the link object associated with the entity to be linked cannot be determined by means of the strong related information, so that the above second preset result can be determined as a third link result corresponding to the entity to be linked, so that the third link result can indicate that the link object associated with the entity to be linked cannot be determined by means of the strong related information; however, if it is determined that the entity to be linked belongs to an abbreviation and there is full name information corresponding to the entity to be linked in the text to be processed, it may be determined that there is strong correlation information of the entity to be linked in the text to be processed, so that a third link result (for example, a "link result determined based on the strong correlation information" shown in fig. 2) corresponding to the entity to be linked may be determined by means of the strong correlation information (i.e., the full name information), so that the third link result may represent an object determined based on the strong correlation information and recommended to the target user as a link object associated with the entity to be linked above, and thus the object link processing by means of the strong correlation information existing in the text to be processed may be achieved.

Step three: and determining a target link object of the entity to be linked according to the first link result, the second link result and the third link result.

The target link object refers to a link object finally determined for the entity to be linked, so that the target link object can be provided for a target user as a link object associated with the entity to be linked.

In addition, the present application is not limited to the above determination of the "target link object", and for example, it may specifically include the following steps 101 to 103.

Step 101: if the above "third link result" is the above fourth object, it may be determined that there is a link object determined based on the strong correlation information, and thus the fourth object may be determined as the target link object of the above entity to be linked.

Step 102: if the above "third link result" is the second preset result and the above "second link result" is the above second object, it may be determined that there is no link object determined based on the strong correlation information, but there is a link object determined based on the model, so the second object may be determined as the target link object of the above entity to be linked.

Step 103: if the above "third link result" is the second preset result and the above "second link result" is the first preset result, it may be determined that there is neither a link object determined based on the strong correlation information nor a link object determined based on the model, so the above first link result (i.e., the above "first object") may be determined as the target link object of the above entity to be linked.

Based on the related content of the steps 101 to 103, it is known that, for the entity to be linked, after obtaining the plurality of link results determined for the entity to be linked, the link object determined based on the strong related information is first selected as the link object associated with the entity to be linked; selecting the link object determined based on the model as the link object associated with the entity to be linked; the link object determined based on the historical behavior is selected as the link object associated with the entity to be linked.

Based on the above description of the first to third steps, in one possible implementation manner, the entity linking method may include: after determining a to-be-linked entity and context information of the to-be-linked entity from a to-be-processed text, determining a first link result corresponding to the to-be-linked entity by utilizing at least one candidate object corresponding to the to-be-linked entity recorded in a pre-built link relation library; if the context information of the entity to be linked meets the preset information quantity condition, determining a second link result corresponding to the entity to be linked according to the entity to be linked, the context information of the entity to be linked and a pre-constructed entity link model; if the entity to be linked belongs to an abbreviation and the full scale information corresponding to the entity to be linked exists in the text to be processed, determining a third link result corresponding to the entity to be linked according to the full scale information; and finally, determining the target link object of the entity to be linked according to the first link result, the second link result and the third link result, so that the aim of entity link can be fulfilled. The target link object is determined based on three link modes, so that the target link object can better represent which object the entity to be linked is linked to, and the entity link effect is better improved.

Based on the above descriptions of S1 to S3, for the entity linking method provided in the embodiment of the present application, the method may include: after determining a to-be-linked entity and context information of the to-be-linked entity from the to-be-processed text, determining a first link result corresponding to the to-be-linked entity by utilizing at least one candidate object corresponding to the to-be-linked entity recorded in a pre-constructed link relation library, and determining a target link object of the to-be-linked entity according to the first link result and the context information of the to-be-linked entity, so that the aim of entity link can be achieved.

Indeed, to better improve the accuracy of the above "link object determined based on historical behavior", the present application also provides a possible implementation of the above entity linking method, in which the entity linking method may include not only the above S1-S3, but also the following steps 111-113. Wherein the execution time of step 111 is later than the execution time of the above S3.

Step 111: and acquiring the actual feedback operation corresponding to the target link object.

The actual feedback operations refer to some feedback operations triggered by the target user for the target link object after the target link object is provided to the target user as the link object associated with the entity to be linked. For example, the "actual feedback operation" may refer to the "link feedback operation" shown in fig. 2.

In addition, the relevant contents of the above "actual feedback operation" are similar to those of the above "feedback operation", and for brevity, they are not described here again.

Furthermore, the present application is not limited to the above "actual feedback operation" acquisition procedure, and for example, it may be implemented in any operation acquisition manner existing or occurring in the future.

Step 112: and updating the recommendation characteristic data corresponding to the target link object according to the actual feedback operation corresponding to the target link object.

It should be noted that the present application is not limited to the embodiment of the above step 112, and for example, it may be implemented using the above formula.

Step 113: and updating the link relation library according to the recommendation characteristic data corresponding to the target link object.

It should be noted that the embodiment of step 113 is not limited in this application, and for example, it may specifically include steps 1131 to 1132 below.

Step 1131: if the corresponding relation between the entity to be linked and the target link object does not exist in the link relation library, the corresponding relation and recommendation characteristic data corresponding to the target link object can be added to the link relation library.

Step 1132: if the corresponding relation between the entity to be linked and the target link object exists in the link relation library, the recommendation characterizing data corresponding to the target link object is directly utilized to update the recommendation characterizing data corresponding to the target link object recorded in the link relation library.

Based on the related content of the above steps 111 to 113, after the target link object of the entity to be linked is obtained, the target link object may be provided to the target user as the link object associated with the entity to be linked, so that after receiving the actual feedback operation triggered by the target user for the target link object, the above link relation library is updated based on the actual feedback operation, so that the link relation library can represent the link usage habit of the target user as real time as possible, which is beneficial to improving the accuracy of the above "link object determined based on historical behavior", thereby being beneficial to improving the entity link determination effect.

Based on the entity linking method provided in the embodiment of the present application, the embodiment of the present application further provides an entity linking device, which is explained and illustrated in the following with reference to fig. 3. Fig. 3 is a schematic structural diagram of an entity linking device according to an embodiment of the present application. It should be noted that, for the technical details of the entity linking device provided in the embodiment of the present application, please refer to the related content of the entity linking method.

As shown in fig. 3, an entity linking apparatus 300 provided in an embodiment of the present application includes:

a first determining unit 301, configured to determine an entity to be linked and context information of the entity to be linked from a text to be processed;

a second determining unit 302, configured to determine a first link result corresponding to the entity to be linked by using a pre-constructed link relation library; the link relation library comprises a corresponding relation between the entity to be linked and at least one candidate object; the first linking result is determined from the at least one candidate object;

and a third determining unit 303, configured to determine a target link object of the entity to be linked according to the first link result and the context information of the entity to be linked.

In a possible implementation manner, the link relation library further comprises recommendation characteristic data of each candidate object; the recommendation characterizing data of the candidate object is used for representing the possibility of associating the candidate object with the entity to be linked;

the first linking result is determined from the at least one candidate object and recommended characterizing data for the at least one candidate object.

In a possible implementation manner, the second determining unit 302 is specifically configured to: comparing the recommended characteristic data of the at least one candidate object to obtain a comparison result; and if the comparison result indicates that the recommendation characteristic data of the first object in the at least one candidate object is the largest, generating the first link result according to the first object.

In one possible implementation, the recommended characterization data for the candidate object is determined based on historical feedback operations corresponding to the candidate object.

In one possible implementation, the entity linking apparatus 300 further includes:

the feedback acquisition unit is used for acquiring actual feedback operation corresponding to the target link object;

the first updating unit is used for updating the recommendation characteristic data corresponding to the target link object according to the actual feedback operation corresponding to the target link object;

and the second updating unit is used for updating the link relation library according to the recommended representation data corresponding to the target link object.

In a possible implementation manner, the third determining unit 303 includes:

a first determining subunit, configured to determine, if the context information of the entity to be linked meets a preset information amount condition, a second link result corresponding to the entity to be linked according to the entity to be linked, the context information of the entity to be linked, and a pre-constructed entity link model;

A second determining subunit, configured to determine, if the entity to be linked belongs to an abbreviation and there is full scale information corresponding to the entity to be linked in the text to be processed, a third linking result corresponding to the entity to be linked according to the full scale information;

and the third determining subunit is used for determining the target link object of the entity to be linked according to the first link result, the second link result and the third link result.

In a possible embodiment, the first determining subunit is specifically configured to: inputting the entity to be linked and the context information of the entity to be linked into the entity link model to obtain an entity link description result output by the entity link model; the entity link description result comprises the confidence level of at least one alternative object corresponding to the entity to be linked; the at least one alternative object includes a second object and a third object; the confidence of the second object is the maximum of the confidence of the at least one candidate object; the confidence level of the third object is a second largest value in the confidence level of the at least one candidate object; and if the confidence coefficient of the second object is larger than a first threshold value and the difference value between the confidence coefficient of the second object and the confidence coefficient of the third object is larger than a second threshold value, generating the second link result according to the second object.

In a possible embodiment, the second determining subunit is specifically configured to: matching the full name information with at least one object to be selected in the entity word lexicon to obtain a matching result; and if the matching result indicates that a fourth object matched with the full scale information exists in the at least one object to be selected, determining the third link result according to the fourth object.

In one possible implementation, the entity linking apparatus 300 further includes:

a fourth determining unit, configured to determine abbreviation characteristics of the entity to be linked if the entity to be linked belongs to an abbreviation;

the template generation unit is used for generating a template to be used corresponding to the entity to be linked according to a full-scale template generation rule corresponding to the abbreviation feature;

the template retrieval unit is used for carrying out template retrieval processing on the text to be processed by utilizing the template to be used to obtain a retrieval result;

and a fifth determining unit, configured to determine, according to the search result, full name information corresponding to the entity to be linked.

In one possible implementation, the entity linking apparatus 300 further includes:

and a sixth determining unit, configured to determine that the entity to be linked belongs to an abbreviation if the entity to be linked meets a preset abbreviation condition.

In one possible implementation, the entity linking apparatus 300 further includes:

the scene acquisition unit is used for acquiring the text scene identification of the text to be processed;

the information searching unit is used for searching a link relation library corresponding to the text scene identifier from a pre-constructed mapping relation; the mapping relation comprises a corresponding relation between the text scene identification and the link relation library.

Based on the above-mentioned related content of the entity linking device 300, it is known that, for the entity linking device 300 provided in this embodiment of the present application, after determining the entity to be linked and the context information of the entity to be linked from the text to be processed, the first linking result corresponding to the entity to be linked is determined by using at least one candidate object corresponding to the entity to be linked recorded in the pre-built linking relation library, and the target linking object of the entity to be linked is determined according to the first linking result and the context information of the entity to be linked, which is beneficial to better improving the entity linking effect.

In addition, the embodiment of the application also provides electronic equipment, which comprises a processor and a memory: the memory is used for storing instructions or computer programs; the processor is configured to execute the instructions or the computer program in the memory, so that the electronic device performs any implementation of the entity linking method provided in the embodiments of the present application.

Referring to fig. 4, a schematic diagram of an electronic device 400 suitable for use in implementing embodiments of the present disclosure is shown. The terminal devices in the embodiments of the present disclosure may include, but are not limited to, mobile terminals such as mobile phones, notebook computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and the like, and stationary terminals such as digital TVs, desktop computers, and the like. The electronic device shown in fig. 4 is merely an example and should not be construed to limit the functionality and scope of use of the disclosed embodiments.

As shown in fig. 4, the electronic device 400 may include a processing means (e.g., a central processing unit, a graphics processor, etc.) 401, which may perform various suitable actions and processes according to a program stored in a Read Only Memory (ROM) 402 or a program loaded from a storage means 408 into a Random Access Memory (RAM) 403. In the RAM403, various programs and data necessary for the operation of the electronic device 400 are also stored. The processing device 401, the ROM 402, and the RAM403 are connected to each other by a bus 404. An input/output (I/O) interface 405 is also connected to bus 404.

In general, the following devices may be connected to the I/O interface 405: input devices 406 including, for example, a touch screen, touchpad, keyboard, mouse, camera, microphone, accelerometer, gyroscope, etc.; an output device 407 including, for example, a Liquid Crystal Display (LCD), a speaker, a vibrator, and the like; storage 408 including, for example, magnetic tape, hard disk, etc.; and a communication device 409. The communication means 409 may allow the electronic device 400 to communicate with other devices wirelessly or by wire to exchange data. While fig. 4 shows an electronic device 400 having various means, it is to be understood that not all of the illustrated means are required to be implemented or provided. More or fewer devices may be implemented or provided instead.

In particular, according to embodiments of the present disclosure, the processes described above with reference to flowcharts may be implemented as computer software programs. For example, embodiments of the present disclosure include a computer program product comprising a computer program embodied on a non-transitory computer readable medium, the computer program comprising program code for performing the method shown in the flow chart. In such an embodiment, the computer program may be downloaded and installed from a network via communications device 409, or from storage 408, or from ROM 402. The above-described functions defined in the methods of the embodiments of the present disclosure are performed when the computer program is executed by the processing device 401.

The electronic device provided by the embodiment of the present disclosure belongs to the same inventive concept as the method provided by the above embodiment, and technical details not described in detail in the present embodiment can be seen in the above embodiment, and the present embodiment has the same beneficial effects as the above embodiment.

The embodiments of the present application also provide a computer readable medium, in which instructions or a computer program is stored, which when executed on a device, cause the device to perform any implementation of the entity linking method provided by the embodiments of the present application.

It should be noted that the computer readable medium described in the present disclosure may be a computer readable signal medium or a computer readable storage medium, or any combination of the two. The computer readable storage medium can be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or a combination of any of the foregoing. More specific examples of the computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this disclosure, a computer-readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. In the present disclosure, however, the computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, with the computer-readable program code embodied therein. Such a propagated data signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination of the foregoing. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to: electrical wires, fiber optic cables, RF (radio frequency), and the like, or any suitable combination of the foregoing.

In some implementations, the clients, servers may communicate using any currently known or future developed network protocol, such as HTTP (Hyper Text Transfer Protocol ), and may be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), the internet (e.g., the internet), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future developed networks.

The computer readable medium may be contained in the electronic device; or may exist alone without being incorporated into the electronic device.

The computer readable medium carries one or more programs which, when executed by the electronic device, cause the electronic device to perform the method described above.

Computer program code for carrying out operations of the present disclosure may be written in one or more programming languages, including, but not limited to, an object oriented programming language such as Java, smalltalk, C ++ and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computer (for example, through the Internet using an Internet service provider).

The flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.

The units involved in the embodiments of the present disclosure may be implemented by means of software, or may be implemented by means of hardware. Where the names of the units/modules do not constitute a limitation of the units themselves in some cases.

The functions described above herein may be performed, at least in part, by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: a Field Programmable Gate Array (FPGA), an Application Specific Integrated Circuit (ASIC), an Application Specific Standard Product (ASSP), a system on a chip (SOC), a Complex Programmable Logic Device (CPLD), and the like.

In the context of this disclosure, a machine-readable medium may be a tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. The machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

It should be noted that, in the present description, each embodiment is described in a progressive manner, and each embodiment is mainly described in a different manner from other embodiments, and identical and similar parts between the embodiments are all enough to refer to each other. For the system or device disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple, and the relevant points refer to the description of the method section.

It should be understood that in this application, "at least one" means one or more, and "a plurality" means two or more. "and/or" for describing the association relationship of the association object, the representation may have three relationships, for example, "a and/or B" may represent: only a, only B and both a and B are present, wherein a, B may be singular or plural. The character "/" generally indicates that the context-dependent object is an "or" relationship. "at least one of" or the like means any combination of these items, including any combination of single item(s) or plural items(s). For example, at least one (one) of a, b or c may represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", wherein a, b, c may be single or plural.

It is further noted that relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.

The steps of a method or algorithm described in connection with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. The software modules may be disposed in Random Access Memory (RAM), memory, read Only Memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the application. Thus, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims (14)

1. A method of entity linking, the method comprising:

determining an entity to be linked from a text to be processed and the context information of the entity to be linked;

determining a first link result corresponding to the entity to be linked by utilizing a pre-constructed link relation library; the link relation library comprises a corresponding relation between the entity to be linked and at least one candidate object; the first linking result is determined from the at least one candidate object;

and determining a target link object of the entity to be linked according to the first link result and the context information of the entity to be linked.

2. The method of claim 1, wherein the link relation library further comprises recommendation characterizing data for each of the candidate objects; the recommendation characterizing data of the candidate object is used for representing the possibility of associating the candidate object with the entity to be linked;

the first linking result is determined from the at least one candidate object and recommended characterizing data for the at least one candidate object.

3. The method of claim 2, wherein the determining of the first link result comprises:

comparing the recommended characteristic data of the at least one candidate object to obtain a comparison result;

and if the comparison result indicates that the recommendation characteristic data of the first object in the at least one candidate object is the largest, generating the first link result according to the first object.

4. The method of claim 2, wherein the candidate recommendation characterizing data is determined based on historical feedback operations corresponding to the candidate.

5. The method according to claim 1, wherein after determining the target link object of the entity to be linked according to the first link result and the context information of the entity to be linked, the method further comprises:

Acquiring an actual feedback operation corresponding to the target link object;

updating recommendation characterization data corresponding to the target link object according to the actual feedback operation corresponding to the target link object;

and updating the link relation library according to the recommendation characteristic data corresponding to the target link object.

6. The method according to claim 1, wherein determining the target link object of the entity to be linked according to the first link result and the context information of the entity to be linked includes:

if the context information of the entity to be linked meets the preset information quantity condition, determining a second link result corresponding to the entity to be linked according to the entity to be linked, the context information of the entity to be linked and a pre-constructed entity link model;

if the entity to be linked belongs to an acronym and the full scale information corresponding to the entity to be linked exists in the text to be processed, determining a third link result corresponding to the entity to be linked according to the full scale information;

and determining the target link object of the entity to be linked according to the first link result, the second link result and the third link result.

7. The method of claim 6, wherein the determining of the second link result comprises:

inputting the entity to be linked and the context information of the entity to be linked into the entity link model to obtain an entity link description result output by the entity link model; the entity link description result comprises the confidence level of at least one alternative object corresponding to the entity to be linked; the at least one alternative object includes a second object and a third object; the confidence of the second object is the maximum of the confidence of the at least one candidate object; the confidence level of the third object is a second largest value in the confidence level of the at least one candidate object;

and if the confidence coefficient of the second object is larger than a first threshold value and the difference value between the confidence coefficient of the second object and the confidence coefficient of the third object is larger than a second threshold value, generating the second link result according to the second object.

8. The method of claim 6, wherein the determining, according to the full scale information, a third link result corresponding to the entity to be linked includes:

Matching the full name information with at least one object to be selected in the entity word lexicon to obtain a matching result;

and if the matching result indicates that a fourth object matched with the full scale information exists in the at least one object to be selected, determining the third link result according to the fourth object.

9. The method of claim 6, wherein the method further comprises:

if the entity to be linked belongs to the abbreviation, determining the abbreviation characteristics of the entity to be linked;

generating a template to be used corresponding to the entity to be linked according to a full scale template generation rule corresponding to the abbreviation feature;

performing template retrieval processing on the text to be processed by utilizing the template to be used to obtain a retrieval result;

and determining the full scale information corresponding to the entity to be linked according to the search result.

10. The method of claim 6, wherein the method further comprises:

and if the entity to be linked meets the preset abbreviation condition, determining that the entity to be linked belongs to the abbreviation.

11. The method according to claim 1, wherein before determining the first link result corresponding to the entity to be linked by using the pre-constructed link relation library, the method further comprises:

Acquiring a text scene identifier of the text to be processed;

searching a link relation library corresponding to the text scene identifier from a pre-constructed mapping relation; the mapping relation comprises a corresponding relation between the text scene identification and the link relation library.

12. An entity linking apparatus, comprising:

the first determining unit is used for determining an entity to be linked and the context information of the entity to be linked from the text to be processed;

the second determining unit is used for determining a first link result corresponding to the entity to be linked by utilizing a pre-constructed link relation library; the link relation library comprises a corresponding relation between the entity to be linked and at least one candidate object; the first linking result is determined from the at least one candidate object;

and the third determining unit is used for determining a target link object of the entity to be linked according to the first link result and the context information of the entity to be linked.

13. An electronic device, the device comprising: a processor and a memory;

the memory is used for storing instructions or computer programs;

The processor for executing the instructions or computer program in the memory to cause the electronic device to perform the method of any of claims 1-11.

14. A computer readable medium, characterized in that it has stored therein instructions or a computer program which, when run on a device, causes the device to perform the method of any of claims 1-11.

Images