Content based Recommendation from Explicit Ratings
|Thesis title in Czech:||Content based Recommendation from Explicit Ratings|
|Thesis title in English:||Content based Recommendation from Explicit Ratings|
|Key words:||doporučovací systémy, obsah, explicitní hodnocení|
|English key words:||recommender systems, content, explicit rating|
|Academic year of topic announcement:||2013/2014|
|Type of assignment:||diploma thesis|
|Department:||Department of Software Engineering (32-KSI)|
|Supervisor:||prof. RNDr. Peter Vojtáš, DrSc.|
|Author:||hidden - assigned and confirmed by the Study Dept.|
|Date of registration:||03.06.2014|
|Date of assignment:||03.06.2014|
|Confirmed by Study dept. on:||24.06.2014|
|Date and time of defence:||16.06.2016 10:30|
|Date of electronic submission:||13.05.2016|
|Date of submission of printed version:||13.05.2016|
|Date of proceeded defence:||16.06.2016|
|Reviewers:||Mgr. Ladislav Peška, Ph.D.|
|Goal of this thesis is to produce a research report in the area of recommendation. Research should comprise of several problems, models, methods, prototypes, data and metrics in the area of content based recommendation from explicit ratings. Contribution is expected in comparison of obtained results with respect to baseline technology and results in literature. Emphasis is neither on newest technology nor computing efficiency.
Emphasis is on principles, methods and metrics reflecting quality of recommendation, especially with respect to order sensitive metrics and fine tuning of various alternatives. Evaluation of statistical significance of obtained results is welcome.
Another axis, along which methods should be tested, is various types of data, ranging from dominantly numeric, via mixed numeric-nominal to dominantly nominal.
One of output of work is a prototypical implementation, which enables comparison in a single system environment (focused not on scalability, rather on quality of recommendation).
Last deliverable is formulation of main problems for a promising project (usable e.g. as formulation of a diploma thesis of type Implementation, Research problem and/or Analysis and design of a solution).
Topic is very broad and deep, one can expect that focus will be specified and driven to a part of problem (mainly in implementation, solely simulating some algorithms might suffice).
|To create a representative list of references is one of thesis tasks. For start, one can use publications of the project Web Semantization, related diploma and PhD thesis etc. e.g.
- Topics covered in the lecture NDBI021 Querying with preferences
- Start with Alan Eckhardt: Similarity of users' (content-based) preference models for Collaborative filtering in few ratings scenario. Expert Syst. Appl. 39(14): 11511-11516 (2012) and follow works quoting recommendation on chosen data benchmarks
- Publicly available recommender tools
- Publicly available benchmark data
|Preliminary scope of work in English|
|Our main motivation is the problem with information overload and the task is automated personalized support of a user for filtering big amount of data (already annotated, e.g. linked data and/or annotated without permission of owner).
Typical use-cases are retail, search, filtering news, decision support in private and/or professional situations and can cover several domains. Of course we are not going to create a new Google. We aim to offer an added value for a small community (e.g. friends on a social network, which are keen to supply some annotations) and only for a small portion of the Web visited be members of this community and pages similar to those.
In this work we focus on personalization, user preferences and recommendation. Understanding data (extraction, annotation, organization and intelligence) are part of another thesis and here we assume this is solved.