skadio commited on
Commit
b7e6284
1 Parent(s): ed1d174

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +54 -0
README.md CHANGED
@@ -1,3 +1,57 @@
1
  ---
2
  license: apache-2.0
 
 
 
 
 
 
3
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
  license: apache-2.0
3
+ tags:
4
+ - optimized item selection
5
+ - recommender systems
6
+ - online experimentation
7
+ - multi-objective optimization
8
+ pretty_name: ISP
9
  ---
10
+
11
+ # Optimized Item Selection Datasets
12
+
13
+ We provide the datasets that are used to test the multi-level optimization framework ([CPAIOR'21](https://link.springer.com/chapter/10.1007/978-3-030-78230-6_27), [DSO@IJCAI'22](https://arxiv.org/abs/2112.03105)), for solving Item Selection Problem (ISP) to boost exploration in Recommender Systems.
14
+
15
+ ## Overview of Datasets
16
+ The datasets include:
17
+
18
+ * [**GoodReads datasets**](book_recommenders_data/) for book recommenders. Two datasets are randomly selected from the source data [GoodReads Book Reviews](https://dl.acm.org/doi/10.1145/3240323.3240369), a small version with 1000 items and a large version with 10,000 items. For book recommendations, there are 11 different genres (e.g., fiction, non-fiction, children), 231 different publishers (e.g. Vintage, Penguin Books, Mariner Books), and genre-publisher pairs. This leads to 574 and 1,322 unique book labels for the small and large datasets, respectively.
19
+
20
+ * [**MovieLens datasets**](movie_recommenders_data/) for movie recommenders. Two datasets are randomly selected from the source data [MovieLens Movie Ratings](https://dl.acm.org/doi/10.1145/2827872), a small version with 1000 items and a large version with 10,000 items. For movie recommendations, there are 19 different genres (e.g. action, comedy, drama, romance), 587 different producers, 34 different languages (e.g. English, French, Mandarin), and genre-language pairs. This leads to 473 and 1,011 unique movie labels for the small and large datasets, respectively.
21
+
22
+ Each dataset in GoodReads and MovieLens contains a `*_data.csv` file, which contain the text content (i.e., title + description) of the items, and a `*_label.csv`, which contains the labels (e.g., genre or language) and a binary 0/1 denoting whether an item exbihits a label.
23
+
24
+ Each column in the csv file is for an item, indexed by book/movie ID. The order of columns in data and label files are the same.
25
+
26
+ [Selective](https://github.com/fidelity/selective) implements the multi-objective optimization approach from ([CPAIOR'21](https://link.springer.com/chapter/10.1007/978-3-030-78230-6_27), [DSO@IJCAI'22](https://arxiv.org/abs/2112.03105)) as part of `TextBased Selection`.
27
+
28
+ By solving the ISP with Text-based Selection in Selective, we select a smaller subset of items with maximum diversity in the latent embedding space of items and maximum coverage of labels.
29
+
30
+ ## Usage Example
31
+ ```python
32
+ # Import Selective (for text-based selection) and TextWiser (for embedding space)
33
+ import pandas as pd
34
+ from feature.selector import Selective, SelectionMethod
35
+ from textwiser import TextWiser, Embedding, Transformation
36
+
37
+ # Load Text Contents
38
+ data = pd.read_csv("goodreads_1k_data.csv").astype(str)
39
+
40
+ # Load Labels
41
+ labels = pd.read_csv("goodreads_1k_label.csv")
42
+ labels.set_index('label', inplace=True)
43
+
44
+ # TextWiser featurization method to create text embeddings
45
+ textwiser = TextWiser(Embedding.TfIdf(), Transformation.NMF(n_components=20, random_state=1234))
46
+
47
+ # Text-based selection
48
+ selector = Selective(SelectionMethod.TextBased(num_features=30, featurization_method=textwiser))
49
+
50
+ # Result
51
+ subset = selector.fit_transform(data, labels)
52
+ print("Reduction:", list(subset.columns))
53
+ ```
54
+
55
+ ## Citation
56
+ If you use ISP in our research/applications, please cite as follows:
57
+