Repository for the Web Search (Computer Science Msc) course, offered at FCT-NOVA.
19 September – Tutorial 1: Test-bed setup
24 September – Tutorial 2: Low-level search spaces
The first goal of this tutorial is to create low-level search spaces:
- Features: BoW, HoC, HoG, GIST
The second goal is to implement a search functionality for any search spaces:
- Given an image query or a text query and a search space return the k nearest documents.
- Identify the correct distance for each space.
- Normalize data.
1 October – Tutorial 3: High level search spaces
- Information categorization: image classification
- High-level image embeddings.
- Named entities recognition
- Parse the queries file and generate the search results
- Experimental evaluation setup.
All checkpoints will be graded as follows:
- Implementation correctness (30%).
- Results analysis (30%): Discuss the performance of each algorihm and their robustness to different data. The performance should be tied to clear understanding of the algorithm properties and parameters. Make sure you justify your decisions wisely.
- Critical discussion (40%): Critical comparison of different algorithms for the same task. Identify the advantages of each method. Clearly identify bottlenecks and suggest possible improvements.
29 October – Tutorial 4: Graph method - Label Propagation
5 November – Tutorial 5: Multi-Feature Label Propagation
Checkpoint 1 discussions.
Checkpoint 1 discussions.
19 November – Tutorial 5: Cross-modal retrieval
Checkpoint 2 discussions.
Course website: http://ctp.di.fct.unl.pt/~jmag/ws/index.html
If you have any question or if you find any issue, please contact: Prof. João Magalhães - Email: jm.magalhaes@fct.unl.pt or David Semedo - Email: df.semedo@campus.fct.unl.pt