JP2011008782A - Method and device of inferring uncertain mismatching ontology regarding specific query - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method and a device of inferring uncertain mismatching ontology regarding a specific query.SOLUTION: The method of inferring ontology includes: (a) a step of inputting an ontology and a query; (b) a step of selecting an element set related to the query from the ontology using a selection function; (c) a step of inspecting whether a query result can be inferred from the selected element set or not; and (d) a step of incrementing the degree of association of the selection function and repeating the steps (b) and (c) if the query result cannot be obtained, and of inputting a path for inferring corresponding to the query result if the query result can be obtained. The method further includes a step of calculating a certainty factor of each query result based on the confidence factor of an ontology element. Accordingly, a user can obtain as correct a query result as possible on the uncertain mismatching ontology without changing the original ontology.

Description

  The present invention relates generally to the field of semantic web and ontology inference, and more particularly to inference methods and apparatus specific to queries that solve uncertain and inconsistent problems with knowledge and ontology.

An ontology is a description of entities in the real world and the relationships between entities. The Semantic Web relies heavily on ontology quality and accuracy. However, knowledge and information in real life is usually inaccurate, resulting in ontology inaccuracy. So-called uncertain ontology (uncertain)
ontology) means that the accuracy of the ontology is stochastic. Ontology uncertainty arises from a variety of causes, including expert subjective uncertainty, original ontology uncertainty, and automatic ontology learning tool uncertainty. The higher the ontology uncertainty, the greater the potential for inaccuracies.

  Ontology uncertainty often results in ontology inconsistencies. So-called inconsistent ontology means that an error or collision is in the ontology. It also prevents some concepts in the ontology from being interpreted correctly. Ontology inconsistencies can arise from a variety of causes, including misrepresentation, ambiguity, conversion from other forms, and integration from multiple sources. In ontology inference, inconsistent ontologies can lead to wrong answers, wrong semantic understanding and knowledge representation.

  Because there are a lot of inconsistent and uncertain ontologies in the Semantic Web, and it is difficult to always ensure the quality of ontologies, you can get wrong results for ontology queries, or you can't get any results It often happens.

  Although incomplete ontology correction is an effective strategy, the cost of the correction is relatively high and difficult to implement. It can also cause other related problems after correction. Therefore, developing new inference methods for inconsistent and uncertain ontologies is sometimes more effective and easier to implement. In other words, it is necessary to develop an inference method for inconsistent and uncertain ontologies, which makes it possible to query the most accurate query without modifying the original ontology. Can be used to obtain results.

Existing ontology inference is mainly based on logical reasoning. The emergence of ontology description languages (eg, Web Ontology Language (OWL)) and its description logic (DL)
Due to its close relationship with logic, modern DL reasoners can detect mismatched ontology very effectively.

  In related technology, some researchers have put a lot of effort into researching ontology uncertainty and inconsistency.

For example, the paper “Guilin Qi, Jeff Z. Pan and Qiu Ji, entitled
“Extending Description Logics with Uncertainty Reasoning in Possibilistic
Logic ”(Proceedings of the 9th European Conference on Symbolic and Quantitative
In “Approaches to Reasoning with Uncertainty-- ECSQARU 2007” ”(Non-Patent Document 1), probability extension is applied to descriptive logic to realize inference for uncertain ontology.
This method uses the decrement function to remove each low confidence element from every ontology element. Therefore, the efficiency is poor. In addition, the method determines only “true” results for the query and does not handle “false” results for the query.

The paper “Z. Huang, F. Van Harmelen and A. Ten”
Teije, entitled "Reasoning with inconsistent ontologies" (Proceedings
of the International Joint Conference on Artificial Intelligence-IJCAI'05,
2005) "(Non-Patent Document 2) proposes a linear extension strategy used to infer mismatched ontology without changing the original ontology.
However, in this method, ontology uncertainty is not taken into account.

In addition, the paper "Silvia Calegari and Elie Sanchez, entitled" A Fuzzy
Ontology-Approach to improve Semantic Information Retrieval "(Proceedings
of URSW, 2007) "(Non-Patent Document 3) defines an ambiguous ontology for semantic information retrieval.
This method takes into account the ontology uncertainty problem, but does not provide a specific solution to solve the problem.

Guilin Qi, Jeff Z. Pan and Qiu Ji, entitled “Extending DescriptionLogics with Uncertainty Reasoning in Possibilistic Logic” (Proceedings of the9th European Conference on Symbolic and Quantitative Approaches to Reasoningwith Uncertainty-- ECSQARU 2007) Z. Huang, F. Van Harmelen and A. TenTeije, entitled "Reasoning with inconsistent ontologies" (Proceedingsof the International Joint Conference on Artificial Intelligence-IJCAI'05,2005) Silvia Calegari and Elie Sanchez, entitled "A Fuzzy Ontology-Approach to improve SemanticInformation Retrieval" (Proceedings of URSW, 2007)

  The method described in Non-Patent Document 1 uses a decrement function to delete low confidence elements from all ontology elements one by one. Therefore, the efficiency is poor. In addition, the method determines only “true” results for the query and does not handle “false” results for the query. In the method described in Non-Patent Document 2, the ontology uncertainty is not taken into consideration. Although the method described in Non-Patent Document 3 takes into account the ontology uncertainty problem, it does not provide a specific solution to solve the problem.

  To summarize the above points, it is difficult to obtain accurate query results for uncertain and inconsistent ontologies with the related inference methods so far, and the processing efficiency is very low. Therefore, there is a need for an inference method that takes into account the problems of ontology uncertainty and inconsistency at the same time and has a relatively high processing efficiency.

  The present invention has been developed to solve the aforementioned problems.

The present invention provides an inference method and apparatus for inaccurate and inconsistent ontology.
First, the present invention uses a probability extension method that obtains a confidence factor (CF) for all elements of the ontology. Thereafter, the incremental selection function selects elements related to the query step by step, and at the same time, determines “true” and “false” for the query. As a result, it is possible to obtain a query result that is most likely to be the correct answer without changing the ontology. This method takes into account the problems of ontology uncertainty and inconsistency at the same time and can achieve high efficiency. In addition, this method allows the user to return all answers (ie, query results) of the query, the corresponding inference path, and the certainty of each answer or result. This allows the user to select the most reliable result more effectively.

According to a first aspect of the present invention, an inference method and apparatus for uncertain ontology is provided. By using this method and apparatus, the user can obtain an inference result that is as correct as possible for the query set by the user without changing the existing ontology.
Specifically, the ontology inference method according to the present invention includes (a) inputting an ontology and a query, (b) selecting an element set related to the query from the ontology using a selection function, and (c) ) Checking whether the query result can be inferred from the selected element set; and (d) if the query result cannot be obtained from the selected element set, the relevance of the selection function is incremented, and the step (b) (C) is repeated, and when a query result is obtained, a step of outputting an inference path corresponding to the query result is included.
In addition, the ontology inference method includes a step of performing probability expansion on the ontology to calculate a confidence factor (CF) value indicating the reliability of the element for each element in the ontology, and a corresponding inference for each query result. Calculating the certainty of the query result using the CF value of the element on the path.

  An inference device for an indefinite ontology includes ontology input means for inputting an ontology, query input means for inputting a query, increment selection means for selecting an element set related to the query from the ontology using a selection function, Query result checking means for checking whether a query result can be inferred from the element set selected by the increment selection means, and each query result when the query result checking means determines that the query result can be obtained from the selected element set An inference path recording means for recording an inferred path corresponding to, and an output means for outputting an inferred path corresponding to the query result. When the query result checking means determines that the query result cannot be obtained from the element set selected by the increment selection means, the increment selection means increments the relevance of the selection function to update the element set, The query result checking means checks whether a query result can be inferred from the updated element set.

According to a second aspect of the present invention, there is provided an inference method for uncertain and inconsistent ontology based on the inference method described above for uncertain ontology. This method simultaneously determines a “true” answer and a “false” answer for a query, and acquires all query results for inconsistent ontologies. Thereafter, the query results are sorted according to the confidence of each result. This allows the user to efficiently select the most reliable result.
Specifically, the ontology inference method according to the present invention includes (a) inputting an ontology and a query, (b) selecting an element set related to the query from the ontology using a selection function, and (c) ) Checking whether the query result can be inferred from the selected element set; and (d) if the query result cannot be obtained from the selected element set, the relevance of the selection function is incremented, and the step (b) If (c) is repeated and a query result is obtained, the inference path corresponding to the query result is recorded, the relevance of the selection function is incremented, and the steps (b) and (c) are repeated; e) The selected element set already contains all of the ontology elements, or the selection function cannot select any more relevant elements for the query If, and a step of outputting all the recorded query results and their corresponding inference pass.
Similarly, the method includes the additional step of performing probability expansion on the ontology to calculate a confidence factor (CF) value that indicates the confidence of the element for each element in the ontology. In addition, this method sorts the query results recorded based on the magnitude of the certainty factor for each query result by calculating the certainty factor of the query result using the CF value of the element on the corresponding inference path. There is a step to do.

  An inference device for an indeterminate and inconsistent ontology includes ontology input means for inputting an ontology, query input means for inputting a query, and increment selecting means for selecting an element set related to the query from the ontology using a selection function. A query result checking unit that checks whether a query result can be inferred from the element set selected by the increment selection unit; and the query result checking unit determines that a query result can be obtained from the selected element set. Query result recording means for recording results, inference path recording means for recording inference paths corresponding to the query results, and output means for outputting inference paths corresponding to the query results. Regardless of whether the query result checking means determines that a query result is obtained, the increment selection means increments the relevance of the selection function to update the element set, and the query result checking means updates Whether or not the query result can be inferred from the selected element set, and the output means continues until the selected element set already contains all elements of the ontology or the selection function cannot select any more relevant elements for the query. Output all recorded query results and their corresponding inferred paths.

  Compared with the related art, the present invention has the following effects.

  1. According to the present invention, it is possible to obtain a query result that is highly likely to be the most correct for an indeterminate and inconsistent ontology. This exempts the user's cost needed to modify the existing ontology.

  2. Since the inference method of the present invention employs an incremental selection function, elements related to a specific query can be selected. Therefore, elements necessary for inference can be acquired very quickly, and query results can be returned quickly. Furthermore, this selection function selects only those elements that are relevant to a particular query and does not select unrelated elements. Thus, interference from other uncertain and inconsistent elements can be avoided as much as possible.

3. The query result obtained using the present invention includes not only the determination about “true” and “false” answers, but also a specific inference path (path) and certainty factor of each answer.
This allows the user to obtain more useful information to facilitate his selection for the most reliable query results.

The above and other features of the present invention will be more clearly understood by reading the following detailed description with reference to the following drawings, in which:
It is a block diagram which shows the internal structure of the uncertain ontology reasoning apparatus 100 by the 1st Embodiment of this invention. It is a flowchart explaining the operation | movement process of the indeterminate ontology inference apparatus 100 shown in FIG. It is a block diagram which shows the internal structure of the indefinite and mismatched ontology inference apparatus 300 by the 2nd Embodiment of this invention. FIG. 4 is a flowchart illustrating an operation process of the indeterminate and inconsistent ontology inference device 300 shown in FIG. 3. FIG. 2 is a schematic block diagram of a computer system used to implement the present invention.

First, the present invention provides an inference method and apparatus for inaccurate ontology.
By using this method and apparatus, the user can obtain inference results as accurate as possible for the query set by the user without changing the existing ontology.
Further, the inference result includes the inference path and reliability of each answer.
Thus, the user can obtain more useful information to facilitate his selection for the most reliable query results (ie, answers).

Furthermore, based on the inaccurate ontology inference method and apparatus described above, another improved embodiment of the present invention aims to handle both inaccurate and inconsistent ontology.
In particular, the query result detection means of the present apparatus can determine both “true” and “false” answers for a query in order to obtain all query results for inconsistent ontology. The query results after sorting based on the confidence of each result are provided to the user to more conveniently select the most reliable result.

  For ease of explanation, the terms used in this specification will be briefly described below.

  Uncertain Ontology: Uncertain Ontology means that the accuracy of the ontology is not fixed but stochastic.

  Inconsistent ontology: Inconsistent ontology means that an error or collision exists in the ontology and as a result some concepts in the ontology cannot be interpreted correctly. Although ontology inconsistencies are undesirable for the user, they are often unavoidable.

  Unsatisfiable concept: An unsatisfiable concept means that the concept has no reasonable interpretation in the ontology. Each concept has an interpretation function. But for unsatisfactory concepts, the interpretation function is empty.

Confidence factor: The confidence factor (CF) defined in the present invention is used to indicate the uncertainty of the ontology and the elements (eg, concepts, axioms, examples, relationships) included therein. CF means a score for indicating a confidence level of accuracy of an element in the ontology. The higher the CF, the higher the probability that the element is accurate. The CF value can be obtained according to expert advice when constructing the ontology, or it can be calculated using a predefined calculation method.
For example, CF is represented by a numerical value in a range between “0” and “1”. That is, it can be expressed as CF: N → [0, 1]. Where N denotes the set of all possible elements in the ontology.

  Directly Relevant: Assume that there are two elements Φ and Ψ in a given ontology. If the same name exists in both Φ and Ψ (for example, example name, concept name, relation name, etc.), Φ and Ψ are considered to be directly related.

  K-relevant: Assume that there are two elements Φ and Φ ′ in a given ontology. If there is a set of elements satisfying Ψ0,..., Ψk∈O, Φ and Ψ0 are directly related, Ψ0 and Ψ1 are directly related, ..., Ψk and Φ ′ are directly related, Φ and Φ ′ are It is assumed that the relation is K-related or the relation between Φ and Φ ′ is K.

  Query: A query φ of ontology Σ can be expressed as a description of some concepts, examples, relationships, and axioms in the ontology. There are generally two types of answers for queries. If φ can be inferred from Σ, the answer is “true”, which means that the query statement can be established, and expressed as Σ | = φ. If it is inferred that φ does not hold from Σ, the answer is “false” and is expressed as Σ | = ¬φ.

Inference route (path) (Reasoning
Route (Path)): The inferred route is represented as a set R. This set R records all ontology elements necessary to infer the query φ. That is, if any element of the collection R is deleted, an answer to the query φ cannot be acquired. R | = φ and R⊂R, R ′ | ≠ φ.

Answer confidence (Certainty
Degree of Answer (CDA)): The reliability of an answer to a query is expressed as a score indicating the certainty or reliability of the answer in the ontology. The higher the CDA, the higher the probability that the answer will be accurate.
For example, CDA is represented by a numerical value in a range between “0” and “1”. That is, CDA: A → [0, 1]. Here, A indicates a set of all answers to the query.

  Hereinafter, inference methods and apparatuses for indeterminate and inconsistent ontology according to the first and second embodiments of the present invention will be described in detail with reference to the drawings.

(First embodiment)
FIG. 1 is a block diagram for showing an internal structure of an uncertain ontology inference apparatus 100 according to the first embodiment of the present invention. FIG. 2 is a flowchart for explaining the operation process of the indeterminate ontology inference apparatus 100 shown in FIG.

  As shown in FIG. 1, according to the present embodiment, the indeterminate ontology inference apparatus 100 includes an ontology input means 101, a probability extension means 102, an increment selection means 103, a query result checking means 104, an inference path. A recording unit 105, a certainty factor calculation unit 106, a query input unit 107, and an output unit 108 are provided. In addition, as shown in FIG. 1, there are several such as an ontology storage device 109, a probability ontology storage device 110, an inference path corresponding to a query result, and a query storage measure 111 and a storage device 112 for storing the reliability of each result. Associated storage device.

The operation of the uncertain ontology inference apparatus 100 will be described below with reference to FIG.
The process starts from step 201. In step 201, the ontology input unit 101 and the query input unit 107 input the ontology B and the query of interest φ used for inference from the ontology storage device 109 and the query storage unit 111, respectively.
That is, it is necessary to determine whether the result of the query φ can be obtained from B: B | = φ?
Here, ontology B includes concepts, examples, axioms, relations, etc., and is defined as B = (T, A), T = {φ _i , i = 1, 2,..., N}, A = { c _j , j = 1, 2,..., m}, T represents an axiom collection, A represents an assertion collection, φ represents an axiom, and c represents a concept.

In step 202, the probability extension means 102 performs probability extension on the ontology B acquired by the ontology input means 101 to generate a probability extension ontology B *. This probabilistic extension ontology B * is stored in the probability ontology storage device 110.
The so-called probability extension means that each element in the ontology is given a probability value for representing the degree of confidence of the element, and the probability value indicates a confidence factor (CF).
That is, B = (T, A) → B * = (T *, A *). Here, B * indicates a probability ontology and includes concepts, examples, axioms, relationships, and the like. T * indicates a probabilistic collection of axioms, T * = {(φi, αi), i = 1, 2,. . , N}. A * indicates a collection of probabilistic statements, A * = {, j = 1, 2 (cj, αj),. . m}. α represents a confidence factor (CF).
As a method for calculating the reliability factor, a method well known in the art can be used. For example, the method described in Non-Patent Document 1 described above can be used.
Since the calculation method of the reliability factor is not a characteristic point of the present invention, the detailed description thereof is omitted.

  Next, in the cyclic processing of steps 203 to 210, the increment selection unit 103, the query result checking unit 104, the inference path recording unit 105, and the certainty factor calculation unit 106 are the probabilistic ontology B input from the probability ontology storage device 110. * According to the query φ given by the user and the query result, the inference path and the certainty factor of each answer are calculated.

First, the increment selection unit 103 incrementally selects elements having a degree of association k with the query φ using a selection function.
The selection function s expands the query sentence “Σ | = φ” by using the syntax similarity.
The function s (Σ, φ, k) has three parameters, and the first Σ represents a set of all elements to be selected. φ indicates the original query. k represents the relevance of the initial value 1, and “1” is added every time a selection is added.

  In step 203, first, it is defined that the initial selection function is the query statement φ. That is, k = 1, s (Σ, φ, 0) = φ is set.

When k = 1, the selection function selects, as a working set, elements of ontology Σ that are directly related to the query φ (step 204), and the query result checking means 104 determines that this set obtains a query result. Inspect this set to determine if it can.
More specifically, in step 205, the query result checking means 104 first determines whether the set Σ ′ can obtain a “true” answer for the query φ. That is, it is determined whether φ is true.
If so (“YES” in step 205), in step 207, the inference path recording means 105 and the confidence calculation means 106 are directly used to calculate and record the logical path of the answer and its confidence.
The inference path contains all the elements (eg, axioms, examples, etc.) in the ontology necessary to infer the answer. In addition, inference results cannot be obtained without any element.
As a method for calculating the certainty factor, it is possible to employ a method of multiplying all the elements by CF along the inference path.
Conversely, if a “true” answer is not obtained in step 205, the query result checking means 104 determines whether or not a “false” answer can be inferred from the set Σ ′ for the query φ (step 206). That is, if it is possible to estimate conflicting conclusions about the query φ, it is determined that φ is false.
If φ is false in step 206, the inference path recording unit 105 and the certainty factor calculating unit 106 calculate and record the answer and the inference path of the certainty factor (step 207). The calculation method is as described above.

In step 208, it is determined whether the query φ has an answer.
If so (either “true” answer or “false” answer), the output means 108 is generated by the query result checking means 104, the inference path recording means 105, and the certainty calculation means 106 in step 211, respectively. Output the query result, the inference path of each result and its corresponding certainty.
If not ("NO" in step 208), that is, the current set of elements is not sufficient to make a determination, processing proceeds to the next step 209.

In step 209, whether the set Σ ′ is equal to ontology B * (this means that the current set selects all the elements in the ontology, so no elements can be newly added), or Σ ′ Determine whether it is equal to ontology Σ (this means that the current processing round of the selection function has not selected and added a new element, that is, there are no more elements associated with Σ ′).
If the determination condition is satisfied (“YES” in step 209), the output is “no result” in step 212. That is, the query φ cannot be determined based on ontology B.
Otherwise (“NO” in step 209), the process proceeds to step 210.
In step 210, the increment selection unit 103 increments k by “1”. That is, “1” is added to the relevance of the selection function, and a set Σ ′ is assigned to Σ.
The process then returns to step 204 to continue the cycle until an associated query result can be found or a “no result” determination is made.

  In order to explain the principle of the first embodiment more easily, a specific example is shown below.

For example, for the existing ontology B = {A⊆B, A⊂C, C⊂B, D⊂E, E⊂B, A (a)}, the probability ontology acquired as a result after passing through the probability extension means 102 B * is as follows.
B * = {(A ⊆ B, 0.7), (A ⊂ C, 0.8), (C ⊂ B, 0.6), (D ⊂ E, 0.8), (E ⊂ B, 0.4), (A (a), 0.5)}

  Further, assume that the input query φ is “A“ B? ”.

Based on FIG. 2, the selection function is initialized as follows.
k = 1, Σ = s (B *, φ, 0) = φ = {A⊂B}
The element set is then calculated as follows:
Σ '= s (B *, φ, 1) = {(A⊆B, 0.7), (A⊂C, 0.8), (C⊂B, 0.6), (E, B, 0.4), (A (a ), 0.5)}
That is, select all ontology elements that are directly related to concepts A and B.

  Whether the query φ can be inferred from the set Σ ′, that is, the “true” answer Σ ′ | = φ? Determine if exists.

  Inference finds two inference paths that can yield a “true” answer. Next, two passes are recorded and their confidence is calculated.

Then, it is determined whether there is an answer.
Since the result is “YES”, the query result is output as follows.
Query result A = {{answer
1: true, pass: {A⊆B, 0.7}, 0.7}, {answer 2: true, pass: {A⊂C, 0.8}, (C⊂B, 0.6)}, 0.48}}
Here, since the inference path of answer 1 (answer 1) includes only one axiom, the certainty is equal to the axiom confidence factor 0.7. Answer 2 (answer
For 2), the reasoning path reports two axioms. Therefore, the certainty factor is calculated as 0.8 × 0.6 = 0.48 as the product of the reliability factors of the two axioms.

  Because this calculation process obtains results more effectively and quickly, and the query results reflect a specific inference path and corresponding confidence, users can obtain more useful information, It becomes possible to select a more reliable answer according to the certainty factor.

(Second Embodiment)
FIG. 3 is a block diagram showing the internal structure of an indeterminate and inconsistent ontology inference device 300 according to the second embodiment of the present invention. FIG. 4 is a flowchart for explaining the operation process of the indeterminate and inconsistent ontology inference device 300 shown in FIG.
Compared to the first embodiment, the difference from the second embodiment is that an inference method and apparatus for inconsistent ontology are provided.

  Real ontologies are not only indeterminate but also largely inconsistent, and in many cases, ontology uncertainties lead to inconsistencies, so even if the ontology is indeterminate and inconsistent, it can be as much as possible for the input query. There is a need to provide inference methods that can provide appropriate answers.

According to the second embodiment, the method of the present invention does not require modification of the original ontology. In other words, in cases where ontology uncertainty and inconsistency remain, all “true” and “false” answers can be provided, and a specific inference path for each answer and its certainty can be obtained. it can.
The user can select the answer with the highest certainty, or select the answer that is considered most appropriate based on the set of query results.

In the second embodiment, the selection function used in the first embodiment described above is similarly used to search ontology elements related to the query.
Because the ontology is inconsistent, both “true” and “false” answers may exist simultaneously. For this reason, a system for recording all answers and calculating their certainty is required. Finally, it is possible to select the most reliable answer by sorting all answers according to their confidence score.

  As shown in FIG. 3, as in the first embodiment, the indeterminate and inconsistent ontology inference device 300 includes an ontology input means 101, a probability extension means 102, an increment selection means 103, and a query result checking means 104. An inference path recording unit 105, a certainty factor calculation unit 106, a query input unit 107, and an output unit 108. The difference from the first embodiment is that the indeterminate and inconsistent ontology inference apparatus 300 further includes a query result recording unit 301 and a query result sorting unit 302.

  The operation processing of the indeterminate and inconsistent ontology inference device 300 shown in FIG. 3 will be described below with reference to FIG.

In step 401, as in the first embodiment, the ontology input unit 101 and the query input unit 107 respectively store the ontology B used for inference and the query of interest φ from the ontology storage device 109 and the query storage unit 111, respectively. input.
In step 402, the probability extension means 102 performs probability extension on the ontology B acquired by the ontology input means 101 to generate a probability extension ontology B *. This probabilistic extension ontology B * is stored in the probability ontology storage device 110.

In step 403, the increment selection means 103 initializes k = 1. k is an increment count for selecting the relevance of the selection function.
In this initialization, the function s is selected as the query φ, and the set is assigned to Σ. That is, Σ = s (B *, φ, 0) = φ.
Thereafter, in step 404, a set of elements having a degree of relevance k with the query φ is calculated by the selection function s and expressed as Σ ′ (Σ ′ = s (B *, φ, k)).

Next, in step 405, the query result checking means 104 determines whether or not the set Σ ′ can obtain a “true” answer for the query φ. That is, it is determined whether φ is true.
If so (“YES” in step 405), the inference path recording unit 105 and the certainty factor calculating unit 106 calculate the logical path of the answer and the corresponding certainty factor, and the query result recording unit 301 outputs the answer as a result. Record in set A (step 406). This is expressed as {answer ID: true, pass {…}, confidence}.
If a “true” answer is not obtained, the process proceeds to the next step 407.

In step 407, the query result checking means 104 continues to determine whether or not the set Σ ′ obtains a “false” answer for the query φ.
That is, if it is possible to infer a conclusion opposite to the query φ, it is determined that φ is false.
Similarly, when it is determined that φ is false, the inference path recording unit 105 and the certainty factor calculating unit 106 respectively calculate the inference path of the answer and the certainty factor thereof, and the query result recording unit 301 stores the result set A. The answer is recorded as {answer ID: false, pass {...}, Certainty level} ... (step 408).
Otherwise, the process proceeds to the next step 409.

In step 409, whether the set Σ ′ is equal to ontology B * (this means that the current set selects all the elements in the ontology, so no new elements can be added), or Σ ′ is the ontology Determine if it is equal to Σ (this means that the current processing round of the selection function has not selected and added a new element, that is, there are no more relevant elements for Σ ′).
If the determination condition is satisfied (“YES” in step 409), the process proceeds to step 411. Otherwise, go to step 410.

In step 411, it is determined whether or not there is an answer, that is, whether or not there is an answer to the query φ recorded by the query result recording unit 301.
If so (including a “true” answer and a “false” answer), the process proceeds to step 413. Otherwise, in step 412, the output is “no result”.

In step 410, when the determination condition of step 409 cannot be satisfied, the increment selection unit 103 increments k by “1”. This means that “1” is added to the relevance and a set Σ ′ is assigned to Σ.
Thereafter, the process returns to step 204 to continue the circulation.

In step 413, the query result sorting unit 302 sorts all the query results (including “true” and “false” answers) recorded by the query result recording unit 301. The default sorting method is sorting from large to small for the certainty of each result.
Thereafter, in step 414, the output means 108 outputs the sorted query results, the inference path for each result, and the corresponding certainty factor.

  In order to explain the principle of the second embodiment more easily, a specific example is shown below.

について、確率拡張手段１０２を経由することによって、B*は以下のようになる。

For example, ontology

By passing through the probability extension means 102, B * becomes as follows.

  Assume that the input query φ is “A⊂B?”.

  According to FIG. 4, the initialization is k = 1, Σ = s (B *, φ, 0) = φ = {A⊂B}.

すなわち、概念ＡおよびＢに対して直接関連する要素をすべて選択する。 The set of elements is then calculated as follows:

That is, all elements directly related to concepts A and B are selected.

  Next, it is determined whether or not the query φ can be inferred as “true” by the set Σ ′, that is, whether Σ ′ | = φ? Exists.

Inference finds two inference paths that can yield a “true” answer. Next, two passes are recorded and the corresponding certainty is calculated as follows:
Query result A = {{answer 1: true, path: {(A⊆B, 0.7)},
0.7}, {answer 2: true, pass: {(A⊂C, 0.8), (C⊂B, 0.6)},
0.48}}

Whether the set Σ ′ can infer that the query φ is “false”, that is, Σ ′ | = ¬φ? Determine if exists.
In this example, it is determined that acquisition is not possible.

This determination continues to determine whether Σ ′ = B * or Σ ′ = Σ holds.
In this example, it is found that neither of the two conditions holds.
Therefore, k = k + 1, that is, k = 2, and a set Σ ′ is assigned to Σ.

At this time, the selection function again selects a set of elements as follows.

  Next, it is determined whether or not the query φ can be inferred as “true” by the set Σ ′, that is, whether Σ ′ | = φ? Exists.

This inference finds one inference path that leads to a “true” answer. The path is then added to the answer set.
Answer set A = {{Answer 1:
True, Pass: {(A⊆B, 0.7)}, 0.7}, {Answer 2: True, Pass: {(A⊂C, 0.8), (C⊂B, 0.6)}, 0.48}, {Answer 3: True, path: {(A⊂D∩Q, 0.9), (D⊂E, 0.8), (E⊂B, 0.7)}, 0.5}}

  Whether the set Σ ′ can infer that the query φ is “false”, that is, Σ ′ | = ¬φ? Determine if exists.

This inference finds one inference path that leads to a “false” answer. This path is also added to the answer set and is expressed as follows.
{Answer 4: fake, pass:
{(B⊆G, 0.3), (G⊂H∩K, 0.5), (H⊂A, 0.6)},
0.09}

This determination continues to determine whether Σ ′ = B * or Σ ′ = Σ holds.
It is found that the former condition is satisfied.
That is, the current working set already contains all ontology elements.
Next, it is determined whether there is an answer.
Since the answer set “has” the answers, the query results need to be sorted.

The final output is acquired as follows by the sorting process by the certainty score of the result.
Query result A = {{answer 1: true, path: {(A⊆B, 0.7)},
0.7}, {answer 3: true, pass: {(A⊂D∩Q, 0.9), (D⊂E, 0.8), (E⊂B, 0.7)},
0.5}, {answer 2: true, pass: {(A⊂C, 0.8), (C⊂B, 0.6)},
0.48}, {Answer 4: False, Pass: {(B⊆G, 0.3), (G⊂H∩K, 0.5), (H⊂A, 0.6)},
0.09}}

  The ontology inference devices according to the first and second embodiments of the present invention and their operating principles have been described in detail above with reference to FIGS.

  FIG. 5 is a schematic block diagram of a computer system used to implement the present invention.

As shown in the figure, the computer system 500 includes a CPU 501, a user interface 502, a peripheral device 503, a memory 505, an external storage device 506, and an internal bus 504 that connects the above components to each other.
The memory 505 further includes a semantic web application 5051, an ontology inference module 5052, an ontology editing application 5053, an ontology learning application 5054, other applications 5055 and an operating system (OS) 5056.
The present invention mainly relates to the ontology inference module 5052. The ontology inference module 5052 corresponds to, for example, the indeterminate ontology inference device 100 shown in FIG. 1 or the indeterminate and inconsistent ontology inference device 300 shown in FIG. .
Each application in memory 505 operates in parallel to provide a plurality of different functions.
The external storage device 506 includes various storage devices related to the present invention, such as ontology storage devices, probability ontology storage devices, query storage devices, and the like.

  As can be seen from the above description, the present invention achieves the following effects.

  According to the present invention, it is possible to obtain a query result that is highly likely to be the most correct for an indeterminate and inconsistent ontology. This exempts the user's cost needed to modify the existing ontology.

  Since the inference method of the present invention employs an incremental selection function, elements related to a specific query can be selected. Therefore, elements necessary for inference can be acquired very quickly, and query results can be returned quickly. Furthermore, this selection function selects only those elements that are relevant to a particular query and does not select unrelated elements. Thus, interference from other uncertain and inconsistent elements can be avoided as much as possible.

  The query result obtained by using the present invention includes not only the determination about “true” and “false” answers, but also the specific inference path (path) and certainty factor of each answer. This allows the user to obtain more useful information to facilitate his selection for the most reliable query results.

  While specific embodiments of the present invention have been described above with reference to the accompanying drawings, the present invention is not limited to the specific configurations and processes shown in the accompanying drawings. In the embodiments described above, some specific steps are shown and described as examples. However, the methods and processes of the present invention are not limited to these specific steps. Those skilled in the art will readily understand that these steps can be altered, modified, supplemented, or the order between steps can be changed without departing from the spirit and essential function of the present invention. I will.

  Elements of the present invention may be implemented in hardware, software, firmware or a combination thereof and utilized in its system, subsystem, component or subcomponent. When implemented in software, elements of the present invention are programs or code segments for performing the necessary tasks. The program or code segment can be stored on a computer readable medium or transmitted by a data signal contained in a carrier wave on a transmission cable or communication link. Computer-readable media includes all media that can store or transfer information. Specific examples of computer-readable media are electronic circuits, semiconductor storage devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROM optical disks, hard disks, optical fiber media, radio frequency (RF) links. Etc. The code segment can also be downloaded via a computer network such as the Internet or an intranet.

  Although the present invention has been described above with reference to specific embodiments, the present invention is not limited to the specific embodiments and specific configurations shown in the drawings. For example, some of the components shown can be combined together as one component. Alternatively, one component can be divided into several subcomponents, and other known components can be added. The operation process is not limited to that shown in the embodiment. Those skilled in the art will appreciate that the present invention can be implemented in other specific forms without departing from the spirit and essential function of the invention. Therefore, it should be considered that the above embodiment is illustrative and not restrictive in all respects. The scope of the invention is indicated by the appended claims rather than by the foregoing description. Accordingly, all modifications that come within the meaning and range of equivalency of the claims are embraced within the scope of the invention.

  Further, a part or all of the above-described embodiment can be described as in the following supplementary notes, but is not limited thereto.

(Appendix 1)
(A) inputting an ontology and a query;
(B) selecting an element set associated with the query from an ontology using a selection function;
(C) checking whether a query result can be inferred from the selected set of elements;
(D) If a query result cannot be obtained from the selected element set, the relevance of the selection function is incremented, and the steps (b) and (c) are repeated, and if a query result is obtained, it corresponds to the query result. An ontology inference method comprising: a step of outputting a path for inference.

(Appendix 2)
Performing probability expansion on the ontology to calculate a confidence factor (CF) value for each element in the ontology;
The ontology inference method according to appendix 1, further comprising a step of calculating a certainty factor of the query result using a CF value of an element on the corresponding inference path for each query result.

(Appendix 3)
The ontology inference method according to claim 2, wherein the step of calculating the certainty factor of the Eli result includes a step of calculating a product of CF values of elements on the corresponding inference path.

(Appendix 4)
The ontology inference method according to appendix 1, wherein the query result is a “true” answer or a “false” answer.

(Appendix 5)
The ontology inference method according to claim 1, wherein the element is a concept, an axiom, an example, or a relationship.

(Appendix 6)
(A) inputting an ontology and a query;
(B) selecting a set of elements associated with the query from an ontology using a selection function;
(C) checking whether a query result can be inferred from the selected set of elements;
(D) If a query result cannot be obtained from the selected element set, the relevance of the selection function is incremented, and the steps (b) and (c) are repeated, and if a query result is obtained, it corresponds to the query result. Recording an inferred path, incrementing the relevance of the selection function, and repeating steps (b) and (c);
(E) Output all recorded query results and their corresponding inference paths if the selected set of elements already contains all of the ontology elements or if the selection function cannot select any more relevant elements for the query An ontology inference method comprising the steps of:

(Appendix 7)
Performing probability expansion on the ontology to calculate a confidence factor (CF) value for each element in the ontology;
For each query result, calculating the confidence of the query result using the CF value of the element on the corresponding inference path;
The ontology inference method according to claim 6, further comprising the step of sorting the query results recorded based on the magnitude of the certainty factor.

(Appendix 8)
The ontology inference method according to appendix 6, wherein the query result includes a “true” answer or a “false” answer.

(Appendix 9)
Ontology input means for inputting ontology;
A query input means for inputting a query;
An increment selection means for selecting an element set related to the query from an ontology using a selection function;
Query result checking means for checking whether a query result can be inferred from the element set selected by the increment selection means;
If the query result checking means determines that a query result can be obtained from the selected element set, an inference path recording means for recording an inference path corresponding to each query result;
An output means for outputting a query result and a corresponding inferred path,
If the query result checking unit determines that the query result cannot be obtained from the element set selected by the increment selection unit, the increment selection unit increments the relevance of the selection function to update the element set. The ontology inference device, wherein the query result checking means checks whether or not a query result can be inferred from the updated element set.

(Appendix 10)
A probability extension means for performing probability extension on the ontology to calculate a confidence factor (CF) value for each element in the ontology;
For each query result, there is further provided a certainty factor calculating means for calculating the certainty factor of the query result using the CF value of the element on the corresponding inference path,
The ontology inference device according to appendix 9, wherein the output means outputs a certainty factor of the calculated query result.

(Appendix 11)
Ontology input means for inputting ontology;
A query input means for inputting a query;
An increment selection means for selecting an element set related to the query from an ontology using a selection function;
Query result checking means for checking whether a query result can be inferred from the element set selected by the increment selection means;
A query result recording unit that records a query result when the query result checking unit determines that a query result is obtained from the selected element set;
An inference path recording means for recording an inference path corresponding to each query result;
An output means for outputting a query result and a corresponding inferred path,
Regardless of whether the query result checking means determines that a query result is obtained, the increment selection means increments the relevance of the selection function to update the element set, and the query result checking means includes: Check whether the query result can be inferred from the updated element set,
The output means infers all recorded query results and their corresponding until the selected set of elements already contains all the elements of the ontology or until the selection function can no longer select relevant elements for the query An ontology inference device characterized by outputting a path.

(Appendix 12)
A probability extension means for performing probability extension on the ontology to calculate a confidence factor (CF) value for each element in the ontology;
For each query result, a certainty factor calculation means for calculating the certainty factor of the query result using the CF value of the element on the corresponding inference path;
Comprising query result sorting means for sorting query results recorded based on the degree of certainty,
The ontology inference device according to appendix 11, wherein the output means outputs the sorted query results, the corresponding inference paths, and the certainty of the query results.

DESCRIPTION OF SYMBOLS 100: Uncertain ontology inference apparatus 101: Ontology input means 102: Probability expansion means 103: Increment selection means 104: Query result inspection means 105: Inference path recording means 106: Certainty factor calculation means 107: Query input means 108: Output means 109 : Ontology storage device 110: Probability ontology storage device 111: Query storage measure 112: Storage device 300: Uncertain ontology inference device 301: Query result recording means 302: Query result sorting means 501: CPU
502: User interface 503: Peripheral device 504: Internal bus 505: Memory 506: External storage device 5051: Semantic web application 5052: Ontology inference module 5053: Ontology editing application 5054: Ontology learning application 5055: Other application 5056: Operating・ System (OS)
5061: Query result / inference path / reliability storage device 5062: Probability ontology storage device 5063: Query storage device 5064: Ontology storage device 5065: Other storage device

Claims (10)

(A) inputting an ontology and a query;
(B) selecting an element set associated with the query from an ontology using a selection function;
(C) checking whether a query result can be inferred from the selected set of elements;
(D) If a query result cannot be obtained from the selected element set, the relevance of the selection function is incremented, and the steps (b) and (c) are repeated, and if a query result is obtained, it corresponds to the query result. An ontology inference method comprising: a step of outputting a path for inference. Performing probability expansion on the ontology to calculate a confidence factor (CF) value for each element in the ontology;
The ontology inference method according to claim 1, further comprising: for each query result, calculating a certainty factor of the query result using a CF value of an element on the corresponding inference path.   The ontology inference method according to claim 2, wherein the step of calculating the certainty of the Eli result includes a step of calculating a product of CF values of elements on the corresponding inference path.   The ontology inference method according to claim 1, wherein the element is a concept, an axiom, an example, or a relationship. (A) inputting an ontology and a query;
(B) selecting an element set associated with the query from an ontology using a selection function;
(C) checking whether a query result can be inferred from the selected set of elements;
(D) If a query result cannot be obtained from the selected element set, the relevance of the selection function is incremented, and the steps (b) and (c) are repeated, and if a query result is obtained, it corresponds to the query result. Recording an inferred path, incrementing the relevance of the selection function, and repeating steps (b) and (c);
(E) if the selected set of elements already contains all the elements of the ontology, or if the selection function cannot select any more relevant elements for the query, output all recorded query results and their corresponding inference paths An ontology inference method comprising the steps of: Performing probability expansion on the ontology to calculate a confidence factor (CF) value for each element in the ontology;
For each query result, calculating the confidence of the query result using the CF value of the element on the corresponding inference path;
6. The ontology inference method according to claim 5, further comprising the step of sorting the query results recorded based on the magnitude of the certainty factor. Ontology input means for inputting ontology;
A query input means for inputting a query;
An increment selection means for selecting an element set related to the query from an ontology using a selection function;
Query result checking means for checking whether a query result can be inferred from the element set selected by the increment selection means;
If the query result checking means determines that a query result can be obtained from the selected element set, an inference path recording means for recording an inference path corresponding to each query result;
An output means for outputting a query result and a corresponding inferred path,
If the query result checking unit determines that the query result cannot be obtained from the element set selected by the increment selection unit, the increment selection unit increments the relevance of the selection function to update the element set. The ontology inference device, wherein the query result checking means checks whether or not a query result can be inferred from the updated element set. A probability extension means for performing probability extension on the ontology to calculate a confidence factor (CF) value for each element in the ontology;
For each query result, there is further provided a certainty factor calculating means for calculating the certainty factor of the query result using the CF value of the element on the corresponding inference path,
The ontology reasoning apparatus according to claim 7, wherein the output means outputs a certainty factor of the calculated query result. Ontology input means for inputting ontology;
A query input means for inputting a query;
An increment selection means for selecting an element set related to the query from an ontology using a selection function;
Query result checking means for checking whether a query result can be inferred from the element set selected by the increment selection means;
A query result recording unit that records a query result when the query result checking unit determines that a query result is obtained from the selected element set;
An inference path recording means for recording an inference path corresponding to each query result;
An output means for outputting a query result and a corresponding inferred path,
Regardless of whether the query result checking means determines that a query result is obtained, the increment selection means increments the relevance of the selection function to update the element set, and the query result checking means includes: Check whether the query result can be inferred from the updated element set,
The output means infers all recorded query results and their corresponding until the selected set of elements already contains all the elements of the ontology or until the selection function can no longer select relevant elements for the query An ontology inference device characterized by outputting a path. A probability extension means for performing probability extension on the ontology to calculate a confidence factor (CF) value for each element in the ontology;
For each query result, a certainty factor calculation means for calculating the certainty factor of the query result using the CF value of the element on the corresponding inference path;
Comprising query result sorting means for sorting query results recorded based on the degree of certainty,
The ontology inference device according to claim 9, wherein the output means outputs the sorted query result, the corresponding inference path, and the certainty factor of the query result.

Images