D-Net
/
openaire-graph-docs
6
1
Fork 5
Code Issues 5 Pull Requests 2 Packages Projects Releases 69 Wiki Activity

Descripe dedupe profile v4

This commit is contained in:
Giambattista Bloisi 2024-02-08 12:20:05 +01:00
parent d8e23c2277
commit 24bdb4e8fd
1 changed files with 14 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

23
docs/graph-production-workflow/deduplication/research-products.md
View File

@ -12,14 +12,17 @@ The next sections describe how each stage of the deduplication workflow is faced
### Candidate identification (clustering) ### Candidate identification (clustering)
To match the requirements of limiting the number of comparisons, OpenAIRE clustering for research products works with two functions: To match the requirements of limiting the number of comparisons, OpenAIRE clustering for research products works with two functions:
* *DOI-based function*: the function generates the DOI when this is provided as part of the record properties; * *PID extraction function*: the function generates the PIDs when at least one is provided as part of the ``pid`` record properties;
* *Title-based function*: the function generates a key that depends on (i) number of significant words in the title (normalized, stemming, etc.), (ii) module 10 of the number of characters of such words, and (iii) a string obtained as an alternation of the function prefix(3) and suffix(3) (and vice versa) on the first 3 words (2 words if the title only has 2). For example, the title ``Search for the Standard Model Higgs Boson`` becomes ``search standard model higgs boson`` with two keys key ``5-3-seaardmod`` and ``5-3-rchstadel``. * *Author and Title extraction function*: the function generates a key that depends on (i) the number of authors of the product, with a cap of 21 authors (ii) number of significant words in the title (normalized, stemming, etc.), divided by 10, and (iii) a string obtained as an alternation of the function prefix(3) and suffix(3) (and vice versa) on the first 3 words (2 words if the title only has 2).
<br>For example, a product composed by 197 authors and titled ``Search for the Standard Model Higgs Boson`` becomes ``search standard model higgs boson`` with two keys key ``21-0-seaardmod`` and ``21-0-rchstadel``.
To give an idea, this configuration generates around 77Mi blocks, which we limited to 200 records each (only 15K blocks are affected by the cut), and entails 260Bi matches. <!-- UPDATE THOSE NUMBERS, CUT IS NO LONGER DONE, To give an idea, this configuration generates around 77Mi blocks, which we limited to 200 records each (only 15K blocks are affected by the cut), and entails 260Bi matches. -->
### Duplicates identification (pair-wise comparisons) ### Duplicates identification (pair-wise comparisons)
Comparisons in a block are performed using a *sliding window* set to 50 records. The records are sorted lexicographically on a normalized version of their titles. The 1st record is compared against all the 50 following ones using the decision tree, then the second, etc. for an NlogN complexity. Comparisons in a block are performed using a *sliding window* set to 50 records. The records are sorted lexicographically on the normalized version of their titles. The 1st record is compared against all the 50 following ones using the decision tree, then the second, etc.
Local information about matching records is kept and possibly used to prune unneeded comparisons, for example once it is known that A equals to both B and C, B will not be compared against C because the A,B,C group will be anyway discovered by the global transitive closure step later.
<br>
A different decision tree is adopted depending on the type of the entity being processed. A different decision tree is adopted depending on the type of the entity being processed.
Similarity relations drawn in this stage will be consequently used to perform the duplicates grouping. Similarity relations drawn in this stage will be consequently used to perform the duplicates grouping.
@ -31,7 +34,7 @@ The comparison goes through different stages:
2. *instance type check*: comparison of the instance types (indicating the subtype of the record, i.e. presentation, conference object, etc.). If the instance types are not compatible then the records does not match. Otherwise, the comparison proceeds to the next stage 2. *instance type check*: comparison of the instance types (indicating the subtype of the record, i.e. presentation, conference object, etc.). If the instance types are not compatible then the records does not match. Otherwise, the comparison proceeds to the next stage
3. *untrusted pids check*: comparison of all the available pids (in the `pid` and the `alternateid` fields of the record). In every case, no similarity relation is drawn in this stage. If at least one pid is equivalent, the next stage will be a *soft check*, otherwise the next stage is a *strong check*. 3. *untrusted pids check*: comparison of all the available pids (in the `pid` and the `alternateid` fields of the record). In every case, no similarity relation is drawn in this stage. If at least one pid is equivalent, the next stage will be a *soft check*, otherwise the next stage is a *strong check*.
4. *soft check*: comparison of the record titles with the Levenshtein distance. If the distance measure is above 0.9 then the similarity relation is drawn. 4. *soft check*: comparison of the record titles with the Levenshtein distance. If the distance measure is above 0.9 then the similarity relation is drawn.
5. *strong check*: comparison composed by three substages involving the (i) comparison of the author list sizes and the version of the record to determine if they are coherent, (ii) comparison of the record titles with the Levenshtein distance to determine if it is higher than 0.99, (iii) "smart" comparison of the author lists to check if common authors are more than 60%. 5. *strong check*: comparison composed by three substages involving the (i) comparison of the author list sizes and the version of the record to determine if they are coherent, (ii) comparison of the record titles with the Levenshtein distance to determine if it is higher than 0.95, (iii) "smart" comparison of the author lists to check if common authors are more than 60% in case of titles whose length is greater than 30 chars or more than 90% otherwise.
<p align="center"> <p align="center">
<img loading="lazy" alt="Publications Decision Tree" src={require('../../assets/img/decisiontree-publication.png').default} width="100%" className="img_node_modules-@docusaurus-theme-classic-lib-theme-MDXComponents-Img-styles-module"/> <img loading="lazy" alt="Publications Decision Tree" src={require('../../assets/img/decisiontree-publication.png').default} width="100%" className="img_node_modules-@docusaurus-theme-classic-lib-theme-MDXComponents-Img-styles-module"/>
@ -42,9 +45,11 @@ The comparison goes through different stages:
#### Software #### Software
For each pair of software in a cluster the following strategy (depicted in the figure below) is applied. For each pair of software in a cluster the following strategy (depicted in the figure below) is applied.
The comparison goes through different stages: The comparison goes through different stages:
1. *pids check*: comparison of the pids in the records. No similarity relation is drawn in this stage, it is only used to establish the final threshold to be used to compare record titles. If there is at least one common pid, then the next stage is a *soft check*. Otherwise, the next stage is a *strong check* 1. *DOI pids and URLs check*: comparison of the pids of type DOI and URLs in the records. If at least 1 doi is equivalent or 1 URL is equivalent, then records match and the similarity relation is drawn.
2. *soft check*: comparison of the record titles with Levenshtein distance. If the measure is above 0.9, then the similarity relation is drawn 2. *title check*: comparison of the record titles with Levenshtein distance, excluding versioning information.
3. *strong check*: comparison of the record titles with Levenshtein distance. If the measure is above 0.99, then the similarity relation is drawn If the distance is below 0.95 then the records does not match. Otherwise, the comparison proceeds to the next stage
3. *untrusted doi check*: comparison of all the available DOIs (in the `pid` and the `alternateid` fields of the record). If at least 1 doi is equivalent, records match and the similarity relation is drawn.
4. *authors check*: "smart" comparison of the author lists to check if the two products share all authors.
<p align="center"> <p align="center">
<img loading="lazy" alt="Software Decision Tree" src={require('../../assets/img/decisiontree-software.png').default} width="85%" className="img_node_modules-@docusaurus-theme-classic-lib-theme-MDXComponents-Img-styles-module"/> <img loading="lazy" alt="Software Decision Tree" src={require('../../assets/img/decisiontree-software.png').default} width="85%" className="img_node_modules-@docusaurus-theme-classic-lib-theme-MDXComponents-Img-styles-module"/>
@ -66,4 +71,4 @@ The decision tree is almost identical to the publication decision tree, with the
The general concept is that the field coming from the record with higher "trust" value is used as reference for the field of the representative record. The general concept is that the field coming from the record with higher "trust" value is used as reference for the field of the representative record.
The IDs of the representative records are obtained by appending the prefix ``dedup_`` to the MD5 of the first ID (given their lexicographical ordering). If the group of merged records contains a trusted ID (i.e. the DOI), also the ``doi`` keyword is added to the prefix. The IDs of the representative records are obtained by prepending the prefix ``dedup_`` to the MD5 of the first ID (given their lexicographical ordering). If the group of merged records contains a trusted ID type (i.e. the DOI), also the type keyword (i.e. ``doi``) is added to the prefix.