CN103218362B - A kind of Methodologies for Building Domain Ontology and system - Google Patents

Abstract

The invention discloses a domain ontology construction method, including: listing the names of all terms that need to be described by the target ontology to form a keyword set W ₀ ; sorting all keywords in the keyword set W ₀ to form a keyword sequence S ₀ ; create an ontology set O to be reused, submit all keywords in the continuous subsequences extracted from the keyword sequence _S0 to the ontology retrieval system, and add the highest-ranked ontology in the search results to the ontology set O; All the ontologies in the ontology set O are combined and processed to form a new ontology o; the invention also provides a domain ontology construction system. According to the technical solution of the present invention, a method for constructing ontology-oriented keyword query is provided, which has good definition and operability, and can achieve a higher ontology reuse rate.

Description

Method and system for constructing domain ontology

technical field

The invention relates to the fields of information system modeling and knowledge engineering, in particular to a domain ontology construction method and system based on ontology reuse.

Background technique

Tom Gruber defines an ontology as an explicit specification of a conceptualization built for sharing. Conceptualization refers to the establishment of a model for abstract concepts, concrete objects, object attributes, and relationships between objects in a domain or scope, while ontology explicitly expresses a conceptualization as a specification for sharing by multiple subjects. In an ontology, the above-mentioned concepts, relations, etc. are collectively called terms (Term); an ontology can be regarded as a set composed of term descriptions called axioms (Axiom). Nicola Guarino divided ontology into top-level ontology, domain ontology, task ontology, and application ontology. Among them, the top-level ontology describes general concepts (such as space and time), domain and task ontology describe general domains (such as SLR cameras) and general tasks (such as camera sales), and application ontology describes the specific scope of specific applications. (such as a specific SLR camera sales site). Among them, the top-level ontology is usually relatively stable, and the sharing significance of the application ontology is small. Therefore, the construction of the domain and task ontology is the most active, and its construction method is the most important.

The existing methods of constructing domain ontology can be divided into two categories: manual construction and semi-automatic construction. Manual construction is represented by the Ontology Description Capture Method (IDEF5, Integrated Definition for Ontology Description Capture Method), which divides the process of ontology construction into goals and team establishment, original material collection, material analysis, ontology preliminary construction, ontology refinement and verification, etc. 5 steps, each of which is performed manually by humans. Semi-automatic construction, also known as ontology learning, is a computer program that automatically extracts terms representing concepts and relationships between concepts from the text to form a preliminary ontology, which is then manually refined and verified. However, the quality of the preliminary ontology automatically constructed by computer programs is usually poor, which cannot effectively reduce the dependence on manual work, so manual construction is still the mainstream method.

When manually constructing a domain ontology, one way to improve efficiency is to reuse the existing ontology, that is, to transform an existing ontology in the same or similar domain into a new ontology according to new requirements, thus saving costs compared to redevelopment . However, there is a lack of means to find ontologies suitable for reuse from a large number of existing ontologies. One of the main approaches at present is to browse the ontologies in the online ontology library (such as the ontology library of Defense Advanced Research Projects Agency Agent Markup Language (DAML, Defense Advanced Research Projects Agency Agent Markup Language) of the US Department of Defense) one by one, which is inefficient. Another emerging approach is to perform ontology retrieval, submit query keywords to an ontology retrieval system (such as Swoogle search engine), obtain and only browse ontologies that can match the query keywords, thereby improving efficiency. However, there is no well-defined method to guide the above retrieval process, especially the query construction method. Another way to speed up the manual construction of domain ontology is multi-person collaborative construction. The difficulty of this method lies in the conflict checking and resolution of multi-person construction results.

Although the domain ontology, as a model of the concept level, has been separated from the level of natural language, it is still necessary to name the terms in natural language for human understanding. Therefore, the term name is also an important part of the domain ontology. component. Due to the diversity of natural language, a term may correspond to multiple synonymous natural language vocabulary (such as SLR camera and single-lens reflex camera), therefore, an important part in the construction of domain ontology is to obtain the term name as completely as possible all synonyms for .

The existing methods of obtaining synonyms mainly use the dictionary of synonyms (such as WordNet) constructed by linguistic experts. Although the thesaurus dictionaries have high precision, their coverage is limited, and there are very few thesaurus dictionaries that can be easily processed by computer programs. Among them, there are even fewer thesaurus dictionaries in Chinese. Therefore, it is very difficult to obtain synonyms for Chinese term names in domain ontology construction. Difficult, it can usually only be done based on the experience of the builder (ie domain experts), and it is difficult to guarantee the quality, especially the recall rate (ie completeness) of the acquisition.

Another method of obtaining synonyms is to use the group intelligence of the public. This method uses the user query logs of search engines. The basic idea is that if two keywords often appear in user queries, and users often open their corresponding the same web page in the query results, the two keywords are considered synonyms. The main disadvantage of this method is that the accuracy (that is, the correct rate) of obtaining synonyms is very low. The reason is that a webpage may involve multiple different topics, corresponding to multiple keywords that do not have a synonymous relationship. Therefore, even if a user opens the same webpage based on different query keywords, it does not mean that these keywords must be There is a synonymous relationship.

Contents of the invention

In view of this, the main purpose of the present invention is to provide a domain ontology construction method and system, and to provide a construction method for keyword queries oriented to this example retrieval, which has good definition and operability, and can obtain higher ontology reuse rate.

In order to achieve the above object, technical solution of the present invention is achieved in that way:

The present invention provides a domain ontology construction method, including:

List the names of all terms that need to be described by the target ontology to form a keyword set W ₀ ;

Sorting all the keywords in the keyword set W ₀ to form a keyword sequence S ₀ ;

Create an ontology set O to be reused, submit all keywords in the continuous subsequences extracted from the keyword sequence _S0 to the ontology retrieval system, and add the highest ranking ontology in the retrieval results to the ontology set O;

All ontologies in the ontology set O are combined and processed to form a new ontology o.

In the above method, the method further includes: naming the terms described in the new ontology o, and obtaining synonyms according to the names of the terms described in the new ontology o.

In the above method, the list of the names of all terms that need to be described by the target ontology forms the keyword set W0 as _:

For the target field described by the target ontology, use the keywords in the natural language L _S to list the names of all the terms that need to be described by the target ontology to form a keyword set W ₀ .

In the above method, all the keywords in the keyword set W ₀ are sorted to form the keyword sequence S ₀ as:

Build a tree, each node in the tree has a label and a processing tag;

Judging whether the processing marks of all nodes in the tree are "processed", if not, selecting the current node from the nodes of "unprocessed" in all processing marks in the tree, the keyword _set W of the label of the current node is current collection;

Determine whether the current collection contains only one keyword. When the current collection contains more than one keyword, divide the current collection into two subsets, and add the most important subset W _L of the two subsets to the tree as the left child node of the current node In the two subsets, add another subset W _R of the two subsets to the tree as the right child node of the current node, and change the processing mark of the current node to "processed"; otherwise, change the processing mark of the current node to "processed"processing", and then continue to judge whether the processing marks of all nodes in the tree are "processed", until the processing marks of all nodes in the tree are "processed", according to all keywords in the keyword set W ₀ The depth-first traversal order of the nodes forms the keyword sequence S ₀ .

In the above method, the current collection is divided into two subsets as follows:

The keywords in the current collection are used as a description of a field or range, and the keywords in the two subsets are respectively used as descriptions of two different subfields or subranges of the field or range.

In the above method, all the keywords in the continuous subsequences extracted from the keyword sequence _S0 are submitted to the ontology retrieval system, and the highest-ranked ontology in the retrieval results is added to the ontology set O as follows:

Create an ontology set O to be reused, record the keyword sequence S ₀ as S, obtain the longest subsequence _SH among the continuous prefix subsequences satisfying the conditions in S, cut _SH from the front end of S, Obtain the remaining suffix continuous subsequence S _T ;

Determine whether _SH is an empty sequence, if _SH is an empty sequence, delete the first keyword from _ST ; if _SH is not an empty sequence, add the highest ranking ontology in the search result HITS( _SH ) to O;

Determine whether S _T is an empty sequence, if S _T is not an empty sequence, record S _T as S, and then obtain the longest subsequence _{SH among the prefix continuous subsequences of S that meet the conditions, and convert S H from} S The front end is truncated to obtain the remaining suffix continuous subsequence _ST ; otherwise, if _ST is an empty sequence, the process ends.

In the above method, the condition is that all keywords in the subsequences are combined into a query keyword group, and after the query keyword group is submitted to the ontology retrieval system, the retrieval result HITS( _SH ) is not empty.

In the above-mentioned method, all the ontologies in the ontology set O are combined and processed to form a new ontology o as follows:

Combine all the ontologies in the ontology set O to form a new ontology o; and edit the new ontology o according to the requirements of describing the target domain;

The editing process includes at least adding terms and axioms, deleting terms and axioms, and modifying terms and axioms.

In the above method, naming the terms described in the new ontology o is: naming each term described in the new ontology o with a _vocabulary in LS.

In the above method, the acquisition of synonyms according to the names of the terms described in the new ontology o is as follows:

For the name t of each term described in the new ontology o, create three keyword sets SYN, TRANS, TS;

Submit _t to the translation system from L _S to another natural language LT, and add all keywords in the translation results to the set TRANS;

According to each keyword trans in the set TRANS, all synonyms of trans obtained from the synonym dictionary of _LT are added to the set TS;

Add all keywords in the set TS to the set TRANS, and submit trans' to the translation system from L _T to L _S according to each keyword trans' in the set TRANS, and add all keywords in the translation results to set SYN;

All keywords that are not suitable as synonyms of t are deleted from the set SYN, and all remaining keywords in SYN are used as synonyms of t obtained.

The present invention also provides a domain ontology construction system, including: a listing unit, a sorting unit, an adding unit, and an operation processing unit; wherein,

A listing unit is used to list the names of all terms that need to be described by the target ontology to form a keyword set W ₀ ;

A sorting unit, configured to sort all keywords in the keyword set W ₀ to form a keyword sequence S ₀ ;

The adding unit is used to create the ontology set O to be reused, submit all the keywords in the continuous subsequences extracted from the keyword sequence S ₀ to the ontology retrieval system, and add the highest-ranked ontology in the retrieval results to the ontology set O;

The union operation processing unit is used to perform collective union operation processing on all ontologies in the ontology set O to form a new ontology o.

In the above system, the system also includes:

Naming unit, used to name the terms described in the new ontology o;

The obtaining unit is used for obtaining synonyms according to the names of the terms described in the new ontology o.

The domain ontology construction method and system provided by the present invention list the names of all terms that need to be described by the target ontology to form a keyword set W ₀ ; sort all the keywords in the keyword set W ₀ to form a keyword sequence S ₀ ;Create an ontology set O to be reused, submit all keywords in the continuous subsequences extracted from the keyword sequence S ₀ to the ontology retrieval system, and add the highest-ranked ontology in the retrieval results to the ontology set O; All the ontologies in the set O are combined and processed to form a new ontology o. Therefore, a method for constructing keyword queries oriented to retrieval in this example is provided, so that fewer ontologies can be searched and more important keywords can be covered. effect, has good definition and operability, and can achieve a higher ontology reuse rate; in addition, based on the above method, the present invention can also name the terms described in the new ontology o, and according to the new ontology o The name of the described term is used to obtain synonyms, so a method for obtaining synonyms is provided. Through the dictionary of synonyms in natural language, it can achieve a wide range of applications and achieve high precision and recall.

Description of drawings

Fig. 1 is a schematic flow diagram of the method for realizing domain ontology construction in the present invention;

FIG. 2 is a schematic flow chart of a specific method for realizing step 102 in the present invention;

FIG. 3 is a schematic flowchart of Embodiment 1 of a specific method for realizing step 102 of the present invention;

Fig. 4 is the example figure of binary tree data structure among the present invention;

FIG. 5 is a schematic flow chart of a specific method for realizing step 103 in the present invention;

FIG. 6 is a schematic flowchart of Embodiment 1 of a specific method for implementing step 103 in the present invention;

FIG. 7 is a schematic flow chart of a specific method for implementing step 106 in the present invention;

Fig. 8 is a schematic diagram of the structure of the domain ontology construction system implemented in the present invention.

detailed description

The basic idea of the present invention is: list the names of all terms that need to be described by the target ontology to form a keyword set W ₀ ; sort all the keywords in the keyword set W ₀ to form a keyword sequence S ₀ ; Use the ontology set O, submit all the keywords in the continuous subsequences extracted from the keyword sequence S ₀ to the ontology retrieval system, and add the highest ranked ontology in the retrieval results to the ontology set O; All ontologies are aggregated and processed to form a new ontology o.

The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

The present invention provides a method for constructing a domain ontology. FIG. 1 is a schematic flowchart of the method for constructing a domain ontology in the present invention. As shown in FIG. 1 , the method includes the following steps:

Step 101, list the names of all terms that need to be described by the target ontology to form a keyword set W ₀ ;

Specifically, for the field (called the target field) described by the ontology to be constructed (called the target ontology), such as the SLR camera field, use the keywords in the natural language L _S to list all the terms that need to be described by the target ontology Name, forming a keyword set W ₀ , for example, L _S = Chinese, W ₀ = {"lens", "pixel", "aperture", "focal length", "sensor"}.

Step 102, sort all the keywords in the keyword set W ₀ to form a keyword sequence S ₀ .

Step 103, create an ontology set O to be reused, extract continuous subsequences from the keyword sequence _S0 , submit all keywords in the subsequences to the ontology retrieval system, and add the highest ranking ontology in the retrieval results to the ontology Set O.

Step 104, performing a set union operation on all the ontology in the ontology set O to form a new ontology o;

Specifically, all ontologies in O (namely o ₁ and o ₂ , each ontology regarded as a set of axioms) are combined and processed to form a new ontology o; and according to the description of the target field (such as SLR Domain) to edit the new ontology o, including adding terms and axioms, deleting terms and axioms, modifying terms and axioms, etc.

Step 105, naming the terms described in the new ontology o;

Specifically, each term described in the new ontology o is named with a _vocabulary in LS, for example, a term in the new ontology o is named "lens".

Step 106, obtain synonyms according to the names of the terms described in the new ontology o.

Fig. 2 is a schematic flow chart of a specific method for realizing step 102 of the present invention. As shown in Fig. 2, the method includes the following steps:

Step 201, building a tree, each node in the tree has a label and a processing tag;

Specifically, a binary tree data structure (called a tree) is established, and each node in the tree is attached with a label and a processing tag; initially, the tree contains only one node, and its attached label is W ₀ , and the attached The processing flag is "unprocessed".

Step 202, judge whether the processing marks of all nodes in the tree are "processed", if the processing marks of all nodes in the tree are "processed", execute step 207; otherwise, execute step 203.

Step 203, randomly select a node from all the nodes whose processing flag is "unprocessed" in the tree, this node is the current node, and the keyword set _W0 of the label of the current node is the current set.

Step 204, judge whether the current collection contains only one keyword, if the current collection contains only one keyword, execute step 206; otherwise, execute step 205.

Step 205, divide the current collection into two subsets, add the most important subset W _L of the two subsets to the tree as the left child node of the current node, and use the other subset W _R of the two subsets as the left child node of the current node The right child node is added to the tree;

Specifically, the current collection is divided into two subsets, and the principle of division is: use the keywords in the current collection as a description of a field or scope, and use the keywords in the two subsets as two descriptions of the field or scope, respectively. a description of the different subfields or subranges;

Evaluate the importance of the above two subsets for describing the target domain, the most important subset W _L of the two subsets is added to the tree as the left child node of the current node, and the label of the most important subset W _L is W _L , The treatment label of the most important subset W _L is "unprocessed"; another subset W _R of the two subsets is added to the tree as the right child node of the current node, the label of this subset is W _R , the subset The processing flag of _WR is "unprocessed".

Step 206, change the processing flag of the current node to "processed", and then execute step 202.

Step 207, for each keyword w in the keyword set W ₀ , it can correspond to a node in the tree that satisfies the condition, and the condition is: the keyword set as the label of the node contains and only contains w; based on W The corresponding relationship between the keywords in ₀ and the labels of the nodes in the tree forms a keyword sequence S _{0 according to the depth-first traversal order of the nodes corresponding to all the keywords in W 0 .}

Fig. 3 is a schematic flow chart of Embodiment 1 of a specific method for implementing step 102 of the present invention. As shown in Fig. 3, the method includes the following steps:

Step 301, building a tree, each node in the tree has a label and a processing tag;

Specifically, a binary tree data structure (called a tree) is established, and each node in the tree is attached with a label and a processing tag; initially, the tree only contains one node, such as node A in Figure 4, with The label is W ₀ , W ₀ ={"lens", "pixel", "aperture", "focal length", "sensor"}, and the accompanying processing flag is "unprocessed".

Step 302, selecting a current node from all nodes whose processing flag is "unprocessed" in the tree, and the keyword set of the label of the current node is the current set;

Specifically, since there is a node whose processing mark is "unprocessed" in the tree, a node is randomly selected from all nodes whose processing mark is "unprocessed" in the tree, and this node is called the current node, and the label of the current node is The set of keywords is called the current set; for example, node A in Figure 4, the current set is {"lens", "pixel", "aperture", "focal length", "sensor"}.

Step 303, if the current collection contains more than one keyword, divide the current collection into two subsets; for example, the subset {"pixel", "sensor"} and the subset {"lens", "aperture", "focal length"}.

Step 304, adding the most important subset W _L of the two subsets to the tree as the left child node of the current node, and adding another subset W _R of the two subsets to the tree as the right child node of the current node;

Specifically, evaluate the importance of the above two subsets for describing the target field (such as the SLR camera field), the most important subset W _L of the two subsets, for example W _L = {"pixel", "sensor"}, as the current The left child node (node B as shown in Figure 4) of a node (node A as shown in Figure 4) is added in the tree, and the label of this most important subset W _L is W _L , W _L = {" pixel ", "sensor"}, the processing mark of the most important subset W _L is "unprocessed"; another subset W _R of the two subsets, for example W _R = {"lens", "aperture", "focal length "}, as the right child node (node C as shown in Figure 4) of the current node (node A as shown in Figure 4) is added to the tree, the label of this subset is W _R , W _R = {" lens ", "aperture", "focal length"}, the processing flag of the subset W _R is "unprocessed".

Step 305, change the processing mark of current node (node A as shown in Figure 4) into "processed"; Added to the tree, during this addition, the processing flags of nodes B, C have been changed to "processed".

Step 306, if there is still a node whose processing mark is "unprocessed" in the tree, a node is randomly selected from all nodes whose processing mark is "unprocessed" in the tree, and this node is called the current node, and the label of the current node is The keyword set of is called the current set; for example, node D in Figure 4, the current set is {"sensor"}.

Step 307, if the current collection only contains a keyword, then change the processing mark of the current node (such as node D in Fig. 4) to "processed", and so on, the processing of points E, F, H, I Flag changed to "Processed".

Step 308, when the processing marks of all nodes in the tree are "processed", according to the depth-first traversal order of the nodes corresponding to all keywords in the keyword set W ₀ , a keyword sequence S ₀ is formed;

Specifically, if the processing marks of all nodes in the tree are "processed", then based on the correspondence between the keywords in W ₀ and the labels of the nodes in the tree, for example, "lens" corresponds to the label of node F, and "pixel" corresponds to The label of node E, the label of node I corresponding to "aperture", the label of node H corresponding to "focal length", and the label of node D corresponding to "sensor", according to the depth-first traversal order of nodes corresponding to all keywords in W ₀ ( For example, the nodes D, E, F, H, and I) shown in Figure 4 sort all the keywords in W ₀ to form a keyword sequence S ₀ , that is, S ₀ =<"sensor","pixel" , "Lens", "Focal Length", "Aperture">.

Fig. 5 is a schematic flow chart of a specific method for implementing step 103 of the present invention. As shown in Fig. 5, the method includes the following steps:

Step 501, create an ontology set O to be reused, initially O is an empty set.

Step 502, denote the keyword sequence S ₀ as S.

Step 503, obtain the longest subsequence _SH among the continuous prefix subsequences of S that meet the condition, the condition is that after all the keywords in the subsequence are combined into a query keyword group, the query keyword group is submitted to the ontology After searching the system, the search result is not empty (the search result corresponding to _SH is denoted as HITS( _SH )); and _SH is truncated from the front end of S to obtain the remaining suffix continuous subsequence _ST .

Step 504, judge whether _SH is an empty sequence, if _SH is an empty sequence (that is, there is no prefix continuous subsequence satisfying the condition in _S in step 503), then delete the top keyword from ST; otherwise, Add the highest-ranked ontology in HITS(S _H ) to O.

Step 505, judge whether _ST is an empty sequence, if _ST is not an empty sequence (that is, _SH in step 503 is a subsequence of _S ), record ST as S, and then execute step 503; otherwise, the process ends.

Fig. 6 is a schematic flowchart of Embodiment 1 of a specific method for implementing step 103 of the present invention. As shown in Fig. 6, the method includes the following steps:

Step 601, create an ontology set O to be reused, initially O is an empty set.

Step 602, record the keyword sequence S ₀ as S, that is, S=<"sensor", "pixel", "lens", ) "focal length", "aperture">.

Step 603, obtain the longest subsequence _SH among the continuous prefix subsequences satisfying the condition in S, cut off _SH from the front end of S, and obtain the remaining continuous suffix subsequence S _T ;

Specifically, obtain the longest subsequence _SH among prefix continuous subsequences in S that meet the following conditions, wherein the condition is that after all keywords in the subsequence are combined into a query keyword group, the query After the keyword group is submitted to the ontology retrieval system (such as Swoogle), the retrieval result is not empty (the retrieval result corresponding to S _H is denoted as HITS( _SH ));

For example, after "sensor pixel lens focal length aperture", "sensor pixel lens focal length", and "sensor pixel lens" are respectively submitted to Swoogle, the retrieval results are all empty, and after "sensor pixel" is submitted to Swoogle, the retrieval results are not empty, then S _H =<"sensor", "pixel">; S _H is cut off from the front end of S, then the remaining suffix continuous subsequence is recorded as S _T , S _T =<"lens", "focal length", "aperture">.

Step 604, since _SH is not an empty sequence, add ontology o ₁ with the highest rank in the retrieval result HITS( _SH ) to ontology set O.

Step 605, since _ST is not an empty sequence, the remaining suffix continuous subsequence _ST is recorded as S.

Step 606, obtain the longest subsequence _SH among the continuous prefix subsequences satisfying the condition in S, cut off _SH from the front end of S, and obtain the remaining continuous suffix subsequence S _T ;

Specifically, obtain the longest subsequence _SH among prefix continuous subsequences in S that meet the following conditions, wherein the condition is that after all keywords in the subsequence are combined into a query keyword group, the query After the keyword group is submitted to the ontology retrieval system (such as Swoogle), the retrieval result is not empty (the retrieval result corresponding to S _H is denoted as HITS( _SH ));

For example, after "lens focal length aperture", "lens focal length" and "lens" are submitted to Swoogle respectively, and the search results are all empty, then _SH is an empty sequence. If _SH is cut off from the front end of S, the remaining suffixes are continuous The sequence is denoted as S _T , S _T =<"lens", "focal length", "aperture">.

Step 607, since S _H is an empty sequence, delete the first keyword from S _T , such as "lens", and then obtain S _T =<"focal length", "aperture">.

Step 608, since _ST is not an empty sequence, record _ST as S.

Step 609, obtain the longest subsequence _SH among the continuous prefix subsequences satisfying the condition in S, cut off _SH from the front end of S, and obtain the remaining continuous suffix subsequence S _T ;

Specifically, obtain the longest subsequence _SH among prefix continuous subsequences in S that meet the following conditions, wherein the condition is that after all keywords in the subsequence are combined into a query keyword group, the query After the keyword group is submitted to the ontology retrieval system (such as Swoogle), the retrieval result is not empty (the retrieval result corresponding to S _H is denoted as HITS( _SH ));

For example, after "focal length aperture" is submitted to Swoogle, the retrieval result is not empty, so _SH = <"focal length", "aperture">, if _SH is cut off from the front end of S, then the remaining suffix continuous subsequence is recorded as S _T , S _T is an empty sequence.

Step 610, since _SH is not an empty sequence, add ontology o ₂ with the highest rank in HITS( _SH ) to ontology set O.

Step 611, because _ST is an empty sequence, the final ontology combination O={o ₁ , o ₂ }.

Fig. 7 is a schematic flow chart of a specific method for implementing step 106 of the present invention. As shown in Fig. 7, the method includes the following steps:

Step 701, for the name t of each term described in the new ontology o, for example, t="lens", create three keyword sets, which are respectively recorded as SYN, TRANS, and TS. Initially, SYN, TRANS, and TS are all empty set.

Step 702, submit t (such as "lens") to the translation system from _LS to another natural language _LT , for example, _LT =English, the translation system is Google Translate, and all keywords in the translation results, such as "shot", "camera lens", "camera shot" are added to the set TRANS, that is, TRANS={"shot", "camera lens", "camera shot"}.

Step 703, according to each keyword trans in the set TRANS, such as trans="cameralens", all synonyms of trans (such as "camera lens") obtained from the synonym dictionary (such as WordNet) of _LT , such as "optical lens ", add all the synonyms obtained to the set TS, and so on, for example, the synonyms of "shot" include "guess", "snap", and "camera shot" has no synonyms, then the set TS={"guess", "snap"","opticallens"}.

Step 704, add all keywords in the set TS to the set TRANS, then the set TRANS={"shot", "camera lens", "camera shot", "guess", "snap", "opticallens"}.

Step 705, according to each keyword trans' in the set TRANS, such as trans'="opticallens", submit trans' (such as "optical lens") to the L _T (ie English) to L _S (ie Chinese) A translation system (such as Google Translate), adds all keywords in the translation results, such as "optical lens", to the collection SYN, and so on, for example, the translation results of "shot" include "shooting", "lens", "dose ", the translation results of "camera lens" include "lens", the translation results of "camera shot" include "lens", the translation results of "guess" include "guess", "guess", and the translation results of "snap" include "unit ", " shooting indiscriminately ", then set SYN={"shooting", "camera", "dose", "guessing", "guessing", "unit", "shooting indiscriminately", "optical lens"}.

Step 706. Optionally, in order to improve the accuracy of the synonym acquisition result, all keywords (including t itself, such as "shot") that are not suitable as synonyms for t (such as "shot") can also be deleted from the set SYN, For example, "shooting", "lens", "dose", "guessing", "guessing", "unit", and "random shooting" are not suitable keywords for synonyms of t, then all remaining keywords in the SYN, such as " "Optical lens", as a synonym for the acquired t (such as "lens"); among them, keywords that are not suitable as synonyms for t refer to keywords that cannot be substituted for each other in the current field, and keywords that can be substituted for each other in the current field The word is suitable as a keyword for synonyms of t.

In order to realize the above method, the present invention also provides a domain ontology construction system. FIG. 8 is a schematic structural diagram of the domain ontology construction system in the present invention. As shown in FIG. 8, the system includes: a listing unit 81, a sorting unit 82, an adding unit 83, and operate the processing unit 84; wherein,

Listing unit 81, used to list the names of all terms that need to be described by the target ontology to form a keyword set W ₀ ;

A sorting unit 82, configured to sort all keywords in the keyword set W ₀ to form a keyword sequence S ₀ ;

Adding unit 83, used to create ontology set O to be reused, submit all keywords in continuous subsequences extracted from keyword sequence _S0 to ontology retrieval system, add ontology with the highest ranking in the retrieval results to ontology set O;

The union operation processing unit 84 is configured to perform union operation processing on all ontologies in the ontology set O to form a new ontology o.

The system also includes:

Naming unit 85, used for naming the terms described in the new ontology o;

The obtaining unit 86 is configured to obtain synonyms according to the names of the terms described in the new ontology o.

The above description is only a preferred embodiment of the present invention, and is not used to limit the protection scope of the present invention. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present invention shall be included in the within the protection scope of the present invention.

Claims (11)

1. A domain ontology construction method is characterized by comprising the following steps:

enumerating the names of all terms that need to be described by the target ontology, forming a set of keywords W₀；

For the keyword set W₀All the keywords in (1) are sequenced to form a keyword sequence S₀；

Creating an ontology set O to be multiplexed, from the keyword sequence S₀Submitting all key words in the extracted continuous subsequence to an ontology retrieval system, and retrieving the key words in the retrieval resultAdding the highest-ranking ontology to an ontology set O;

performing collective and operational processing on all ontologies in the ontology set O to form a new ontology O;

wherein the sequence of slave keywords S₀Submitting all the keywords in the extracted continuous subsequence to an ontology retrieval system, and adding the ontology with the highest ranking in the retrieval result to an ontology set O:

creating an ontology set O to be multiplexed, and sequencing a keyword sequence S₀Recording as S, obtaining the longest subsequence S in the prefix continuous subsequences satisfying the condition in S_HWill S_HTruncating from the front end of S to obtain a residual suffix continuous subsequence S_T；

Judgment S_HWhether it is a null sequence, if S_HFor null sequences, from S_TDeleting the top keyword; if S is_HNot null sequence, the search result HITS (S)_H) Adding the highest ranked ontology to O;

judgment S_TWhether it is a null sequence, if S_TNot null sequence, will S_TRecording as S, and acquiring the longest subsequence S in prefix continuous subsequences satisfying the condition of S_HWill S_HTruncating from the front end of S to obtain a residual suffix continuous subsequence S_T(ii) a Otherwise, if S_TAnd the sequence is empty, and the process is ended.

2. The method of claim 1, further comprising: naming the terms described in the new ontology o and performing synonym acquisition according to the names of the terms described in the new ontology o.

3. The method of claim 1, wherein the listing of the names of all terms that need to be described by the target ontology forms a keyword set W₀Comprises the following steps:

for the target domain described by the target ontology, the natural language L is used_SThe keyword list in (1) needs to be described by the target ontologyThe names of all the terms form a keyword set W₀。

4. The method of claim 1, wherein the set of keywords W₀All the keywords in (1) are sequenced to form a keyword sequence S₀Comprises the following steps:

establishing a tree, wherein each node in the tree is provided with a label and a processing mark;

judging whether the processing marks of all nodes in the tree are processed, if not, selecting a current node from all nodes with unprocessed processing marks in the tree, wherein the keyword set W of the label of the current node₀Is a current set;

judging whether the current set only contains one keyword, when the current set contains more than one keyword, dividing the current set into two subsets, and dividing the most important subset W in the two subsets_LAdding the left child node as the current node to the tree, and adding the other of the two subsets W_RAdding the right child node as the current node into the tree, and changing the processing mark of the current node into 'processed'; otherwise, changing the processing mark of the current node into processed, and then continuously judging whether the processing marks of all nodes in the tree are processed until the processing marks of all nodes in the tree are processed, wherein the processing marks of all nodes in the tree are processed according to the keyword set W₀The depth-first traversal order of the nodes corresponding to all the keywords in the sequence S of the keywords is formed₀。

5. The method of claim 4, wherein the dividing the current set into two subsets is:

the keywords in the current set are used as descriptions of a domain or range, and the keywords in the two subsets are respectively used as descriptions of two different sub-domains or sub-ranges of the domain or range.

6. The method of claim 1, wherein the step of removing the metal oxide layer comprises removing the metal oxide layer from the metal oxide layerThe condition is that all the key words in the subsequence are combined into a query key word group, and after the query key word group is submitted to the ontology retrieval system, the result HITS is retrieved (S)_H) Not empty.

7. The method according to claim 1, wherein the union operation process of all ontologies in the ontology set O forms a new ontology O as:

performing collective and operational processing on all ontologies in the ontology set O to form a new ontology O; editing the new body o according to the requirement of describing the target field;

the editing process includes at least adding terms and axioms, deleting terms and axioms, and modifying terms and axioms.

8. The method according to claim 2, characterized in that the term described in the new ontology o is named: one L for each term described in the new ontology o_SThe words in (1) are named.

9. The method of claim 2, wherein the synonym derivation from the name of the term described in the new ontology o is:

creating three keyword sets SYN, TRANS, TS for the name t of each term described in the new ontology o;

submitting t to slave L_STo another natural language L_TThe translation system of (2) adding all keywords in the translation result to the set TRANS;

from L according to each keyword TRANS in the set TRANS_TAdding all the obtained synonyms of trans into a set TS;

adding all keywords in the TS set to the TS set, and submitting TRANS 'to the TS host according to each keyword TRANS' in the TS set_TTo L_SA translation system ofAll keywords in the result are added to the set SYN;

all keywords that are not suitable as synonyms for t are deleted from the set SYN, and all keywords remaining in the SYN are used as synonyms for the obtained t.

10. A domain ontology construction system, the system comprising: the device comprises a listing unit, a sorting unit, an adding unit and an operation processing unit; wherein,

a listing unit for listing the names of all the terms required to be described by the target ontology to form a keyword set W₀；

A sorting unit for sorting the keyword set W₀All the keywords in (1) are sequenced to form a keyword sequence S₀；

An adding unit for creating an ontology set O to be multiplexed, from the keyword sequence S₀Submitting all the key words in the extracted continuous subsequence to an ontology retrieval system, and adding an ontology with the highest ranking in the retrieval result to an ontology set O;

the operation processing unit is used for carrying out collective and operation processing on all ontologies in the ontology set O to form a new ontology O;

the adding unit is specifically configured to create an ontology set O to be multiplexed, and sequence S of keywords₀Recording as S, obtaining the longest subsequence S in the prefix continuous subsequences satisfying the condition in S_HWill S_HTruncating from the front end of S to obtain a residual suffix continuous subsequence S_T(ii) a And judging S_HWhether it is a null sequence, if S_HFor null sequences, from S_TDeleting the top keyword; if S is_HNot null sequence, the search result HITS (S)_H) Adding the highest ranked ontology to O; and judging S_TWhether it is a null sequence, if S_TNot null sequence, will S_TRecording as S, and acquiring the longest subsequence S in prefix continuous subsequences satisfying the condition of S_HWill S_HTruncating from the front end of S to obtain a residual suffix continuous subsequence S_T(ii) a Otherwise, e.g.Fruit S_TAnd the sequence is empty, and the process is ended.

11. The system of claim 10, further comprising:

a naming unit for naming a term described in the new ontology o;

and the acquisition unit is used for carrying out synonym acquisition according to the name of the term described in the new ontology o.