Knowledge entity identification method and knowledge entity identification apparatus

Abstract

A knowledge entity recognition method includes the following steps: receiving a target text to be parsed and metadata, wherein the target text includes a candidate word; using the candidate word to perform comparison in a knowledge base to obtain a plurality of entity names associated with the candidate word from the knowledge base, wherein each of the entity names has corresponding entity description data; comparing the entity description data in the knowledge base with the metadata to obtain a comparison result; and setting the entity names associated with the candidate word in the knowledge base as the output classification of the candidate word in the target text according to the comparison result. The present disclosure also relates to a knowledge entity recognition device.

Description

Technical Field

[0001] This disclosure relates to an electronic device and method thereof, and particularly to a knowledge entity identification device and method. Background Technology

[0002] Traditional knowledge management methods involve experts manually labeling all documents one by one. While current data labeling methods utilize natural language processing (NLP) to analyze syntax and semantics, such corpus analysis cannot enable machines to understand new words; expert labeling is still necessary. Existing training processes using labeled data are too lengthy and inflexible. Furthermore, it's difficult to train new knowledge management systems for different domains based on existing data, and building such systems requires substantial training costs.

[0003] In view of this, knowledge management systems are tools with considerable management capabilities; however, currently, efficient construction methods are lacking, and the accuracy of knowledge management system execution still has room for improvement. Therefore, how to provide efficient system construction methods and highly accurate knowledge management are technical problems that those skilled in the art wish to solve. Summary of the Invention

[0004] According to one embodiment of this disclosure, a knowledge entity recognition method is disclosed, comprising the following steps: receiving target text to be parsed and metadata, wherein the target text includes candidate words; comparing the candidate words with a knowledge base to obtain a plurality of entity names associated with the candidate words from the knowledge base, wherein each entity name has corresponding entity description data; comparing the entity description data and metadata in the knowledge base to obtain a comparison result; and setting the entity names associated with the candidate words in the knowledge base as the output classification of the candidate words in the target text according to the comparison result.

[0005] According to another embodiment, a knowledge entity recognition device is disclosed, comprising a knowledge entity candidate generation module, a knowledge entity verification and enhancement module, and a knowledge entity classification module. The knowledge entity candidate generation module is configured to receive target text to be parsed and metadata, and to compare candidate words from the target text with a knowledge base to obtain a plurality of entity names associated with the candidate words from the knowledge base, wherein each entity name has corresponding entity description data. The knowledge entity verification and enhancement module is coupled to the knowledge entity candidate generation module, and is configured to compare the entity description data and metadata in the knowledge base to obtain a comparison result. The knowledge entity classification module is coupled to the knowledge entity verification and enhancement module, and is configured to set the entity names associated with the candidate words in the knowledge base as the output classification of the candidate words in the target text based on the comparison result. Attached Figure Description

[0006] The following detailed description, when read in conjunction with the accompanying drawings, will facilitate a better understanding of the embodiments described in this disclosure. It should be noted that, for practical purposes, the features in the drawings are not necessarily drawn to scale. In fact, the dimensions of the features may be arbitrarily increased or decreased for clarity of explanation.

[0007] Figure 1 A block diagram of a knowledge entity recognition device according to an embodiment of the present disclosure is shown.

[0008] Figure 2 A flowchart of a knowledge entity identification method according to an embodiment of the present disclosure is shown.

[0009] Figure 3 A flowchart of a knowledge entity identification method according to an embodiment of the present disclosure is shown.

[0010] Explanation of reference numerals in the attached figures:

[0011] 100: Knowledge Entity Recognition Device

[0012] 102: Input Data

[0013] 104: Output Data

[0014] 112: Knowledge Entity Candidate Generation Module

[0015] 114: Knowledge Entity Verification and Enhancement Module

[0016] 116: Knowledge Entity Classification Module

[0017] 200, 300: Knowledge Entity Recognition Methods

[0018] 500: Knowledge Base

[0019] 502: Parsing and Storage Module

[0020] 600: External General Knowledge Base

[0021] S210~S240, S310~S340: Steps

[0022] [Biomaterial Storage]

[0023] Domestic storage information (please note in the order of storage institution, date, and number)

[0024] none

[0025] Overseas storage information (please note in the order of storage country, institution, date, and number)

[0026] none Detailed Implementation

[0027] The following disclosure provides many different embodiments to implement the various features of this disclosure. Embodiments of elements and arrangements are described below to simplify this disclosure. Of course, these embodiments are merely exemplary and not intended to be limiting. For example, the terms "first," "second," etc., used in this disclosure to describe elements are only used to distinguish them from the same or similar elements or operations; these terms are not intended to limit the technical elements of this disclosure, nor to limit the order or sequence of operations.

[0028] Please refer to Figure 1 This diagram illustrates a block diagram of a knowledge entity recognition device 100 according to an embodiment of the present disclosure. The knowledge entity recognition device 100 identifies target objects in input data 102 and provides identified output data 104. For example, the knowledge entity recognition device 100 parses input text, sentences, paragraphs, etc., to perform named entity recognition. In one embodiment, the input data 102 received by the knowledge entity recognition device 100 includes target text and domain metadata. The target text is the data to be parsed. The metadata is data used to assist in the parsing of the target text and may be user-defined categories and keywords.

[0029] In one embodiment, the knowledge entity recognition device 100 is coupled to a knowledge base 500. The knowledge base 500 is coupled to an external general knowledge base 600. The external general knowledge base 600 is, for example, a database with different formats and domain content, such as Wikipedia, specialized dictionaries, or domain expert knowledge. The knowledge base 500 may store knowledge data defined internally and / or data built from data from the external general knowledge base 600. For example, the knowledge base 500 is provided with a parsing and storage module 502. The parsing and storage module 502 can read data from the external general knowledge base 600 and convert the external data into a data structure of a specific format, such as formally processing the external data and domain expert knowledge data, so that the data stored in the knowledge base 500 can be provided to the knowledge entity recognition device 100 for use in recognizing target text.

[0030] In one embodiment, the knowledge entity recognition device 100 includes a knowledge entity candidate generation module 112, a knowledge entity verification and enhancement module 114, and a knowledge entity classification module 116. The knowledge entity candidate generation module 112 is electrically coupled to the knowledge entity verification and enhancement module 114. The knowledge entity verification and enhancement module 114 is electrically coupled to the knowledge entity classification module 116. For ease of understanding of this disclosure, please refer to the following description. Figure 1 and Figure 2 . Figure 2A flowchart is shown for a knowledge entity recognition method 200 according to an embodiment of the present disclosure. The knowledge entity recognition method 200 can be derived from... Figure 1 The knowledge entity recognition device 100 performs the operation.

[0031] In step S210, the knowledge entity candidate generation module 112 receives the target text to be parsed and metadata.

[0032] In one embodiment, the target text to be parsed is the text data to be analyzed, including one or more sentences or paragraphs. On the other hand, the metadata includes multiple categories (keys), and each category includes multiple keywords (values). Users can predefine all categories of metadata and the keywords for each category, and input them simultaneously with the target text into the knowledge entity recognition device 100. For the sake of illustration, the target text mentioned below is the sentence "An apple a day keeps the doctor away." and the metadata is as shown in Table 1, used as examples. It should be noted that this disclosure is not limited to these examples.

[0033] Table 1: Metadata

[0034] Category (key) Keywords (value) FRUIT fruit, juicy, tree, ... MEAT animal, hunt, ... DESSERT sugar, sweet, ...

[0035] In one embodiment, the knowledge entity candidate generation module 112 performs natural language processing techniques to extract nouns or noun phrases from the target text. These extracted nouns or noun phrases are used as candidate words for the target text. Following the example of the target text "An apple a day keeps the doctor away.", the candidate words extracted from the target text include "apple," "day," and "doctor." The number of candidate words for the target text varies depending on the content of the target text. In one embodiment, the target text includes one or more candidate words. In this example of the target text, the number of candidate words is three.

[0036] In step S220, the knowledge entity candidate generation module 112 uses the candidate words of the target text to compare them in the knowledge base 500 to obtain multiple entity names associated with the candidate words from the knowledge base 500.

[0037] In one embodiment, the knowledge entity verification and enhancement module 114 compares these candidate words one by one with the knowledge base 500. The knowledge base 500 records a plurality of entity data. The data structure of each entity data includes, but is not limited to, number, entity name, entity description, entity type, etc., as shown in Table 2.

[0038] Table 2: Knowledge Base

[0039]

[0040] Following the example above, when the candidate word is "apple", the knowledge entity verification and enhancement module 114 compares the candidate word "apple" with the knowledge base 500 in Table 2, and obtains a plurality of entity names associated with "apple", such as "Apple Inc." (number 0), "Apple" (number 1), "Pineapple" (number 2), and "Apple, Oklahoma" (number 3). In one embodiment, these obtained entity names numbered 0 to 3 can be recorded in the candidate list of the candidate word "apple". On the other hand, since the entity name "Orange" (number N) is not the same as / similar to the candidate word "apple", the entity name "Orange" (number N) will not be recorded in the candidate list of the candidate word "apple".

[0041] In one embodiment, the information retrieval method for searching and comparing candidate words in the knowledge base 500 may be the term frequency-inverse document frequency (TF-IDF) method or other data exploration / term frequency statistics methods, and this disclosure is not limited thereto.

[0042] In step S230, the knowledge entity verification and enhancement module 114 compares the entity description data and metadata in the knowledge base 500 to obtain the comparison result.

[0043] In one embodiment, words are searched in the entity description data to obtain more content description information as enhanced information for candidate words, which is then used by the subsequent knowledge entity classification module 116.

[0044] As illustrated in the example above, the candidate list for the candidate word "apple" records four data entries: "Apple Inc.", "Apple", "Pineapple", and "Apple, Oklahoma". Further, based on the metadata received in step S210, the entity description data corresponding to each entity name in the candidate list is retrieved one by one. Taking the metadata category "FRUIT" and its keywords "fruit, juicy, tree" as an example (as shown in Table 1 above), the knowledge entity verification and enhancement module 114 compares the keyword "fruit" with the entity description data corresponding to the entity name "Apple": "An apple is an edible fruit produced by an apple tree (Malus domestica). Apple trees are cultivated worldwide and are the most widely grown species in the genus Malus." It determines whether any word matches "fruit," and counts one match for each matching word obtained. In this embodiment, the category "FRUIT" has three keywords; therefore, the same retrieval and matching are performed on each of these three keywords, finally obtaining the total number of matches for this category. For example, the keyword "fruit" in the category "FRUIT" matches 1 time in the entity description data corresponding to the entity name "Apple"; the keyword "juicy" in the category "FRUIT" matches 0 times in the entity description data corresponding to the entity name "Apple"; and the keyword "tree" in the category "FRUIT" matches 2 times in the entity description data corresponding to the entity name "Apple". Therefore, the total number of matches for the category "FRUIT" relative to the entity name "Apple" is 3.

[0045] Similarly, the keywords "animal,hunt" in the category "MEAT" match 0 times in the entity description data corresponding to the entity name "Apple". The keywords "sugar,sweet" in the category "DESSERT" also match 0 times in the entity description data corresponding to the entity name "Apple". Therefore, among the three categories of metadata input in step S210, the category "FRUIT" has the highest total match count. Thus, the category "FRUIT" in the metadata is the comparison result of the target text. At the same time, the entity name "Apple", which has the strongest correlation with the category "FRUIT", is set as the most relevant entity name.

[0046] In one embodiment, the comparison between metadata and each entity description data can be a cosine similarity method. Entity description data in the knowledge base 500 is searched using metadata, and the entity name closest to the metadata is selected using the cosine similarity method.

[0047] In step S240, the knowledge entity classification module 116 sets the entity name associated with the candidate word in the knowledge base 500 as the output classification of the candidate word in the target text based on the comparison result.

[0048] As illustrated in the example above, the comparison result for the candidate word category in the target text is "FRUIT". Further, the knowledge entity classification module 116 compares this comparison result "FRUIT" with the entity categories ("Fruits; Malus; Plants") corresponding to the most relevant entity name ("Apple") in the knowledge base 500. Since the word "Fruits" matching the comparison result "FRUIT" can be found within the entity categories, it can be verified that the aforementioned found comparison result "FRUIT" is the output category for the candidate word in the target text.

[0049] In one embodiment, the data augmentation results of the candidate words obtained in step S230 and the categories and keywords predefined by the user in the metadata can be input into a text classification model. Figure 1 (Not shown) is used for classification, which can be used to determine that candidate words are knowledge entities of the target text, and classify the knowledge entities into the corresponding categories to obtain the final knowledge entities and their categories.

[0050] Please refer to the above as well. Figure 1 and Figure 3 . Figure 3 The flowchart illustrates a knowledge entity recognition method 300 according to an embodiment of the present disclosure. The knowledge entity recognition method 300 can be derived from... Figure 1 The knowledge entity recognition device 100 performs the operation.

[0051] In step S310, the knowledge entity verification and enhancement module 114 performs a comparison of the candidate words of the target text in the knowledge base 500, and obtains multiple entity names sorted according to their similarity.

[0052] In one embodiment, the target text can be processed by natural language processing techniques performed by the knowledge entity candidate generation module 112 to extract candidate words. Following the example of the target text "An apple a day keeps the doctor away.", the candidate word "apple" is compared with all entity names in the knowledge base 500 in Table 2. The entity name most similar to the candidate word "apple" in the knowledge base 500 has the highest ranking. Based on the similarity of each entity name, they are sorted from high to low to obtain the sorted entity names. The sorted entity names are shown in Table 3, with entity name number 1 being the first priority, entity name number 0 being the second priority, and so on. After the similarity comparison, four sorted entity names are selected from the knowledge base 500, representing data that are identical or similar to the candidate word.

[0053] Table 3

[0054]

[0055] In step S320, the knowledge entity verification and enhancement module 114 compares the keywords of each category of the metadata with the words in the entity description data corresponding to the sorted entity names to obtain comparison results. In some embodiments, the knowledge entity verification and enhancement module 114 compares the keywords of each category of the metadata with the words in the entity description data corresponding to the sorted entity names to calculate the number of word matches that are the same or similar to the keywords of each category in the entity description data, so that each category has a corresponding number of word matches.

[0056] In one embodiment, the metadata includes a plurality of categories, each category including a plurality of keywords. For example, the metadata includes a first category "FRUIT" and a second category "MEAT". The first category "FRUIT" includes the keywords "fruit", "juicy", and "tree". The second category "MEAT" includes the keywords "animal" and "hunt". The third category "DESSERT" includes the keywords "sugar" and "sweet".

[0057] In one embodiment, the keyword "fruit" is compared with the sorted first entity description data "An apple is anedible fruit produced by an apple tree (Malus domestica). Apple trees are cultivated worldwide and are the most widely grown species in the genus Malus.", resulting in one matching word. Similarly, the keywords "juicy" and "tree" are compared with the first entity description data, resulting in 0 and 2 matching words, respectively. In other words, the total number of matching words associated with the first entity name for the first category "FRUIT" is 3. Similarly, the total number of matching words associated with the first entity name for the second category "MEAT" is 0. The total number of matching words between the keywords of each category and the entity description data for the first entity name "Apple" is shown in Table 4.

[0058] Table 4:

[0059] Metadata The first entity describes the number of word matches in the data. Category 1 "FRUIT" 3 The second category is "MEAT". 0 The third category is "DESSERT" 0

[0060] In step S330, the knowledge entity verification and enhancement module 114 sets the category with the largest number of word matches as the output category of the candidate word in the target text.

[0061] Following the example above, the first category has the largest number of word matches (i.e., 3), therefore, the first category "FRUIT" will be set as the output category for this candidate word in the target text.

[0062] It is worth noting that steps S320 and S330 also calculate the total number of word matches for the sorted second entity name using the first, second, and third categories of metadata; calculate the total number of word matches for the sorted third entity name using the first, second, and third categories of metadata; and calculate the total number of word matches for the sorted fourth entity name using the first, second, and third categories of metadata. In other words, all categories of metadata are matched against each sorted entity name to obtain the total number of word matches for each entity name across all categories. For the sake of brevity, the matching steps will not be repeated here.

[0063] In step S340, the knowledge entity verification and enhancement module 114 compares the output classification with the entity categories corresponding to the sorted entity names in the knowledge base 500 to verify whether the output classification of the candidate word in the target text is correct.

[0064] Following the example above, the category with the largest number of word matches is the first category "FRUIT". Therefore, the output category for this candidate word in the target text is set to "FRUIT". In step S340, to verify whether this output category is correct, the output category "FRUIT" is further compared with the first entity category. As shown in Table 3, the first entity categories include "Fruits", "Malus", and "Plants". Since the first entity category "Fruits" matches the output category "FRUIT", it can be verified that this output category is correct.

[0065] In one embodiment, the knowledge entity recognition 100 may be implemented as, but is not limited to, a portable electronic device, mobile phone, tablet computer, personal digital assistant (PDA), wearable device, or laptop computer.

[0066] In one embodiment, the knowledge entity recognition 100 includes at least a processor ( Figure 1 (not shown), storage media ( Figure 1 (not shown) and input / output interfaces ( Figure 1 (Not shown). The processor is used to operate and control the knowledge entity candidate generation module 112, the knowledge entity verification and enhancement module 114, and the knowledge entity classification module 116. The storage medium is used to store a plurality of program instructions and temporary data during instruction execution. An input / output interface is coupled to the processor to receive input data 102 and send output data 104.

[0067] The processor may be implemented as, but is not limited to, a central processing unit (CPU), a system-on-a-chip (SoC), an application processor, an audio processor, a digital signal processor (DSP), or a processing chip or controller with specific functions.

[0068] The storage medium can be implemented as, but is not limited to, random access memory (RAM) or non-volatile memory (e.g., flash memory, read-only memory (ROM), hard disk drive (HDD), solid state drive (SSD), or optical storage, etc.).

[0069] In one embodiment, the text classification model can be an artificial intelligence model that can be built from multiple sub-algorithms, including Artificial Neural Network (ANN) and supervised learning in machine learning, wherein supervised learning includes algorithms such as Support Vector Machine (SVM), regression analysis, and statistical classification.

[0070] In one embodiment, this disclosure proposes a non-transitory computer-readable recording medium capable of storing multiple program codes. The program codes are loaded into a medium such as... Figure 1 After the processor of the knowledge entity recognition device 100 is processed, the processor executes the program code and performs actions such as... Figure 2 and Figure 3 The steps.

[0071] Compared with existing technologies, the knowledge entity recognition method and apparatus disclosed herein can analyze more knowledge entities with the same number of knowledge entities to be analyzed, achieving a high recall rate. Furthermore, given the same number of analyzed knowledge entities, this disclosure can obtain a larger number of correct knowledge entities, achieving high precision.

[0072] In summary, this disclosure improves the accuracy of classifying knowledge entities in target text by inputting metadata along with the target text to be annotated, searching for entity names in the knowledge base, and then further retrieving entity description data for verification. Furthermore, the knowledge entity recognition device and method of this disclosure can be applied to the annotation of a large number of documents. When the documents to be annotated change to different fields, only the corresponding knowledge base needs to be switched to achieve the field switching. In terms of expansion, it can be updated simply by adding new terms to the knowledge base. Moreover, this method can reduce the cost of manual annotation and the burden on experts, saving a significant amount of manual annotation work, and has diverse subsequent applications (automatic annotation of categories and keywords is possible simply by inputting an article).

[0073] The foregoing outlines features of several embodiments to enable those skilled in the art to better understand the implementation of this disclosure. Those skilled in the art will understand that the foregoing can be readily used as the basis for designs or modifications to achieve other variations without departing from the concept and scope of this disclosure, in order to implement the same objectives and / or realize the same advantages of the embodiments described herein. The foregoing should be understood as examples of this disclosure, and its scope of protection should be determined by the claims.

Claims

1. A knowledge entity recognition method, comprising: Receive a target text to be parsed and metadata, wherein the target text includes a candidate word; The candidate word is compared with a knowledge base to obtain a plurality of entity names associated with the candidate word, each entity name having a corresponding entity description data; Compare the entity description data in the knowledge base with the metadata to obtain a comparison result; as well as Based on the comparison result, the entity name associated with the candidate word in the knowledge base is set as an output category of the candidate word in the target text, wherein the metadata includes multiple categories, each category includes multiple keywords, and the knowledge entity recognition method includes: The keywords of each category of the metadata are compared with the words in the entity description data corresponding to the entity name to obtain the comparison result.

2. The knowledge entity recognition method as described in claim 1 further includes: The candidate words of the target text are compared in the knowledge base, and the ranked entity name is obtained based on the similarity of the comparison. The keywords of each category of the metadata are compared with the words in the entity description data corresponding to the sorted entity name to calculate the number of word matches that are the same or similar to the keywords of each category in the entity description data, so that each category has a corresponding number of word matches. as well as The category with the largest number of matching words is set as the output category for the candidate word in the target text.

3. The knowledge entity recognition method as described in claim 2 further includes: The output classification is compared with the entity category corresponding to the entity name in the knowledge base to verify whether the output classification of the candidate word in the target text is correct.

4. The knowledge entity recognition method as described in claim 1 further includes: The candidate words of the target text are compared with those in the knowledge base, and the sorted entity name is obtained based on the similarity. Each keyword in these categories is compared with the words in the sorted entity name's entity description data to obtain a match count for each, wherein the sum of these match counts for each category is the match count for the corresponding category's word. as well as The category corresponding to the word with the largest number of matches is used as the output category.

5. A knowledge entity recognition device, comprising: A knowledge entity candidate generation module is configured to receive a target text to be parsed and metadata, and to compare a candidate word of the target text with a knowledge base to obtain a plurality of entity names associated with the candidate word from the knowledge base, wherein each entity name has a corresponding entity description data; A knowledge entity verification and enhancement module is coupled to the knowledge entity candidate generation module, wherein the knowledge entity verification and enhancement module is configured to compare the entity description data and the metadata in the knowledge base to obtain a comparison result; as well as A knowledge entity classification module is coupled to the knowledge entity verification and enhancement module. The knowledge entity classification module is configured to set the entity name associated with the candidate word in the knowledge base as an output category of the candidate word in the target text based on the comparison result. The metadata includes a plurality of categories, and each category includes a plurality of keywords. The knowledge entity candidate generation module is further configured to perform a comparison in the knowledge base using the candidate word of the target text to obtain the entity name, and to compare the keywords of each category of the metadata with the words in the entity description data corresponding to the entity name to obtain the comparison result.

6. The knowledge entity recognition apparatus as claimed in claim 5, wherein the knowledge entity verification and enhancement module performs a comparison of the candidate words of the target text in the knowledge base, obtains the sorted entity name based on the similarity of the comparison, and compares the keywords of each category of the metadata with the words in the entity description data corresponding to the sorted entity name, so as to calculate the number of word matches of the same or similar keywords of each category in the entity description data, such that each category has a corresponding number of word matches, and sets the category with the largest number of word matches as the output category of the candidate words in the target text.

7. The knowledge entity recognition device as claimed in claim 6, wherein the knowledge entity classification module is further configured to compare the output classification with an entity category corresponding to the entity name in the knowledge base, so as to verify whether the output classification of the candidate word in the target text is correct.

8. The knowledge entity recognition device as described in claim 5, wherein the knowledge entity verification and enhancement module is further configured to: The candidate words of the target text are compared with those in the knowledge base, and the sorted entity name is obtained based on the similarity. Based on each keyword in these categories, the words in the sorted entity name's entity description data are compared to obtain a match count for each, wherein the sum of these match counts for each category is the match count for that word in the corresponding category; and The category corresponding to the word with the largest number of matches is used as the output category.

Images