2. Expert Finding in CrowdSearching
CrowdSearcher involves users belonging to a trusted
network, to answer complex queries.
Expertise Retrieval in the context of CrowdSearcher
What
When
Who: CrowdSearcher should automatically redirect the query to the set of
users that have a good knowledge of the questions’s topic.
Where: It also should do it in the best social network, where the crowd user
has more probability to answer.
2
3. Architecture
•Crawling social network Resources
through REST API
•Enriching Resources and Query
with Semantic Data (entities,
categorization) using a
combination of tools: AlchemyAPI,
TagMe, Open Calais…
•Storing and Indexing
•Matching : How ?
3
4. Future experiments
Matching method:
candidate model vs resource model
A model M(ca) is inferred for each Resources are analyzed and
candidate ca from his resources queried. The candidates
and profile. The model is then used associated with each resource are
to predict how likely a candidate then considered as possible
would produce a query q. experts.
Analysis of social graph:
Retrieve better results studying the network of contacts.
4
5. A Case study: Social Job Search
Automatically search best
candidates to a job offer.
Instances:
• Candidates are users that have
provided permissions to access to
their social network information
• Companies send job
offers, expecting a list of compatible
users as result
• Experts manually select a smaller
list of users from the set originally
returned by the system
5
6. Conclusions
Need for data:
We have developed a website in
order to obtain users with which
test our research.
It is possible to add
• Facebook
• LinkedIn
• Twitter
• Curriculum Vitae
http://expertfinding.altervista.org
eventually moving to
Dbgroup SeCo Server
6