Khresmoi for Scientists

This page provides links to information of interest to scientists, both computer scientists and scientists from other disciplines, who may wish to use the results to assist in their work.

General Information

Prototypes

Khresmoi for Everyone: A search engine aimed at easing medical information search for the broadest group of end users: http://everyone.khresmoi.eu

Khresmoi Professional: A search engine providing professional search features for medical professionals.

The desktop prototypes of Khresmoi for Everyone and Khresmoi Professional are described in Deliverable 8.5.2. The mobile prototypes are described in Deliverable 8.4.

Khresmoi Radiology: A search engine with advanced image search capabilities for radiologists. Watch the videos showing the prototypes for 3D image search and 2D image search. It is described Deliverable 9.4.2.

Software

Khresmoi is producing software for text analysis, semantic search, medical image analysis and search, search interfaces, machine translation and knowledge bases. See the software page for links to this software, and links to Khresmoi Deliverables giving technical details of its use in Khresmoi.

Data sets

See the data sets page. A list of language and evaluation resources produced in the Khresmoi project is in Deliverable 6.2.

Medical Image Analysis and Retrieval Algorithms

  • Khresmoi is developing image analysis algorithms for use on 2D and 3D radiology data. See more information on
  • Feature extraction, indexing and retrieval for both 2D and 3D images are described in Deliverable 2.6.
  • Details of the evaluation of the developed algorithms are available in Deliverable 2.3 and Deliverable 2.7.
  • The use of information from the Khresmoi Knowledge Base in image classification and retrieval is described in Deliverable 5.4.
  • Videos showing the prototypes for 3D image search and 2D image search.
  • Demo of MedSearch, which allows searching in the text and images of medical publications.

Information Extraction and Search

  • The complete approach to semantic indexing and annotation for retrieval is described in Deliverable 1.7.
  • Techniques are being developed for information extraction (Deliverable 1.2), also considering the multilingual aspects (Deliverable 4.1.2).
  • An annotation pipeline with cycles of manual correction and subsequent learning from the corrections is being developed (Deliverable 1.4.2), which will result in an annotated corpus (Deliverable 1.4.1) that will be made available for use by others.
  • A proposed approach for linking named entities from information extraction to resources in the Khresmoi Knowledge Base is described in Deliverable 1.5.
  • The automatic classification of web sites based on content as well as of users based on search log analysis is described in Deliverable 1.6.
  • Automatic summarization of medical documents is described in Deliverable 4.4.
  • A comprehensive search interface is developed in the ezDL framework. Support for query specification and personalization in a search system is described in Deliverable 3.3, while support for collaborative search capabilities is described in Deliverable 3.4. The definitive description is in Deliverable 3.6. Adaptations of the interface for the visually impaired are in Deliverable 3.5.

Machine Translation

  • Techniques for machine translation focussed on the medical domain are being developed (Deliverable 4.2). They are evaluated in Deliverable 4.3
  • Translation support allowing querying of documents in English with queries in other languages is described in Deliverable 3.1.

Domain Analyses

Analyses of the use of search for medical information for the following end users were done:

Further analysis of the results for the medical professional and general public groups led to a specification based on their requirements (Deliverable 8.2).

Evaluation Methodology

  • An integrated evaluation strategy for all Khresmoi components was developed in order to allow an overview of the interaction of all componenents in the full prototype. The meta-analysis of the first round of evaluations based on this strategy is described in Deliverable 7.2.
  • The CLEF eHealth retrieval task (Task 3) is based on Khresmoi data.
  • The results of the first round of user-centred evaluation of the prototypes are in the following deliverables:

Contact Information

Information on how to contact the coordinators of the Khresmoi project.