quhfus/doser

Disambiguation of Semantic Resources - Full framework

doser's Issues

Rename "Organisation" to "Organization"

As discussed today, please rename the mentioned entity type when extracting entities/categories ;)

Index Generation: Candidates for Persons

Extend person candidates:
e.g. http://dbpedia.org/page/Bill_Gates
Candidates: Bill Gates and Gates Bill

Integrate IT-IDF as importance of mention (entity) in document

The TF-IDF weight might be good indicator for importance of a surface form in the document. Hence, the integration of TF-IDF in the PageRank might lead to better results.

Disambiguation is not working when Doc2Vec embeddings are not loaded

Disambiguation service should be able to disambiguate entities if no doc2vec embeddings are loaded. Thus far, Word2Vec service sends an error message instead of similarity value 0.

pom dependency problem for org.aksw + gerbil.nif.transfer

when building with maven, there is an error:
Message:
[ERROR] Failed to execute goal on project doser-gerbilrest:
Could not resolve dependencies for project doser.sub:doser-gerbilrest:war:0.6:
Could not find artifact org.aksw:gerbil.nif.transfer:jar:0.0.1 in spotlight-releases-repository (https://github.com/dbpedia-spotlight/maven-repo/raw/master/releases)

/doser-gerbilrest/pom.xml
the following dependency cannot be resolved:

org.aksw
gerbil.nif.transfer
0.0.1

Impact of Sense Prior

The Sense Prior's impact on popular entities is too big.
Example:
"Three of the greatest guitarists started their career in a single band Clapton, Beck and Page".
Surface Forms: Clapton, Beck and Page.

The Sense Prior of http://dbpedia.org/resource/Beck is quite large due to its many occurrences with surface form Beck.
The Graph transition probability to http://dbpedia.org/resource/Jeff_Beck is highest.
Disambiguation Result: http://dbpedia.org/resource/Beck

However, it is wrong in the case.