2016 AAAI_Ontology Instance Linking--Towards Interlinked Knowledge Graphs笔记整理
First, the paper organize your thoughts flow
1.1 Related research papers
- NLP entity in question refers to the total
- Centralized data deduplication
- A plurality of different identifiers (entity instance) points to a problem with the entity
- While producing equivalence relation (link) between instances of different entities in different data sets and the relationship between the generated independent art remains
- Generating equivalence relations existing methods of testing and classification
1.2 thesis problem
Papers problem-solving process 1.3
- Semantic Web was first proposed entity co-reference between the problem occurred in the entity instance between free text and entity instances or.
- Then proposed entity Semantic Web refers to the total classification method.
- Link Entity equivalence relationship based on string matching: labeled, such as by calculating a similarity between the example (triplets) of. For instance between high similarity score link.
- The method of string matching logic is bound: first by some basic properties of the owl (owl: theSameAs; owl: InverseFunctionalProperty etc.) configuration refers to a total initial set of examples, and examples of such discrimination whether the initial concentration of co refers. However, due to over-reliance on the relationship owl
- Artificial means were all instances of the package label
- Examples of large-scale co-reference system: the use of supervised learning is divided into blocks so as to solve the problem of large-scale example of link data sets
- Evaluation Index: precision, recall and F1, Fcs;
1.4 Experimental methods used paper
Wherein, t represents a threshold value, N, and M represents the size of matching two examples set
- Test data set: BTC data set, OAEI people and restaurant on a data set
- For example coreferential be a result of a certain modification of the original instance and obtained, all the systems are efficient and accurate methods to identify and connect, but many modifications to be obtained means a total instance, precision, recall rate, Fl It has declined.
- DisNGram (Song and He fl in 2011) method in the comparison process, better performance
- Examples of co-integrated system refers, not only between the detecting means co examples, but an example of one type of research references are examples of how the results will affect other
- Tag determines how missing data processing (extended range tag attributes, i.e., the use of other attributes as tag)
The final evaluation of the results of experiments 1.5
1.6 follow-up paper
Second, the techniques and methods used in the paper
- Owl language
- DisNGram
- Use OAEI, BTC, SWAT datasets
Third, we recommended reading references
[Bizer, Heath, and Berners-Lee 2009]
[DisNGram (Song and Heflin 2011)]