skadio commited on
Commit
ddbd344
1 Parent(s): cefe732

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +29 -14
README.md CHANGED
@@ -1,6 +1,10 @@
1
  # Optimized Item Selection Datasets
2
 
3
- 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. The datasets are extracted and processed from original public sources that will be introduced in the following, and are only used in our research study in ISP. Please kindly refrain from distributing the datasets.
 
 
 
 
4
 
5
  ## Overview of Datasets
6
  The datasets include:
@@ -9,13 +13,12 @@ The datasets include:
9
 
10
  * [**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.
11
 
12
- 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.
 
 
13
 
14
  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.
15
 
16
- [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`.
17
-
18
- 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.
19
 
20
  ## Quick Start
21
  To run the example, install required packages by `pip install selective datasets`
@@ -26,7 +29,6 @@ from datasets import load_dataset
26
  from textwiser import TextWiser, Embedding, Transformation
27
  from feature.selector import Selective, SelectionMethod
28
 
29
-
30
  # Load Text Contents
31
  data = load_dataset('skadio/optimized_item_selection', data_files='book_recommenders_data/goodreads_1k_data.csv', split='train')
32
  data = data.to_pandas()
@@ -39,9 +41,10 @@ labels.set_index('label', inplace=True)
39
  # TextWiser featurization method to create text embeddings
40
  textwiser = TextWiser(Embedding.TfIdf(), Transformation.NMF(n_components=20, random_state=1234))
41
 
42
- # Text-based selection with the default selection method - Multi-Level Optimization that maximizes
43
- # coverage and diversity within an upper bound on subset size [CPAIOR'21, DSO@IJCAI'22], by choosing
44
- # the default configuration, i.e. optimization_method="exact" and cost_metric ="diverse".
 
45
  selector = Selective(SelectionMethod.TextBased(num_features=30, featurization_method=textwiser))
46
 
47
  # Result
@@ -49,8 +52,11 @@ subset = selector.fit_transform(data, labels)
49
  print("Reduction:", list(subset.columns))
50
  ```
51
  ## Advanced Usages
52
- Text-based Selection provides access to multiple selection methods. The following configurations are
53
- available to apply these methods:
 
 
 
54
 
55
  - (Default) Solve for Problem *P_max_cover@t* in **CPAIOR'21** - Selecting a subset of items that
56
  maximizes coverage of labels and maximizes the diversity in latent embedding space within an upper
@@ -81,8 +87,11 @@ selector = Selective(SelectionMethod.TextBased(num_features=30,
81
  optimization_method='exact',
82
  cost_metric='unicost'))
83
  ```
 
 
 
84
  - Selecting a subset by performing random selection. If num_features is not set, subset size is defined
85
- by solving #2.
86
  ```python
87
  selector = Selective(SelectionMethod.TextBased(num_features=None, optimization_method='random'))
88
  ```
@@ -91,9 +100,12 @@ selector = Selective(SelectionMethod.TextBased(num_features=None, optimization_m
91
  selector = Selective(SelectionMethod.TextBased(num_features=30,
92
  optimization_method='random'))
93
  ```
 
 
 
94
  - Selecting a subset by adding an item each time using a greedy heuristic in selection with a given
95
  cost_metric, i.e. `diverse` by default or `unicost`. If num_features is not set, subset size is defined
96
- by solving #2.
97
  ```python
98
  selector = Selective(SelectionMethod.TextBased(num_features=None,
99
  optimization_method='greedy',
@@ -106,8 +118,11 @@ selector = Selective(SelectionMethod.TextBased(num_features=30,
106
  optimization_method='greedy',
107
  cost_metric='unicost'))
108
  ```
 
 
 
109
  - Selecting a subset by clustering items into a number of clusters and the items close to the centroids
110
- are selected. If num_features is not set, subset size is defined by solving #2. `cost_metric` argument
111
  is not used in this method.
112
  ```python
113
  selector = Selective(SelectionMethod.TextBased(num_features=None, optimization_method='kmeans'))
 
1
  # Optimized Item Selection Datasets
2
 
3
+ 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.
4
+
5
+ The the multi-objective optimization framework is implemented in [Selective](https://github.com/fidelity/selective) as part of `TextBased Selection`. 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.
6
+
7
+ The datasets are extracted and processed from their original public sources for research purposes as detailed below.
8
 
9
  ## Overview of Datasets
10
  The datasets include:
 
13
 
14
  * [**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.
15
 
16
+ Each dataset in GoodReads and MovieLens contains:
17
+ * `*_data.csv` that contains the text content (i.e., title + description) of the items, and
18
+ * `*_label.csv` that contains the labels (e.g., genre or language) and a binary 0/1 value denoting whether an item exbihits a label.
19
 
20
  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.
21
 
 
 
 
22
 
23
  ## Quick Start
24
  To run the example, install required packages by `pip install selective datasets`
 
29
  from textwiser import TextWiser, Embedding, Transformation
30
  from feature.selector import Selective, SelectionMethod
31
 
 
32
  # Load Text Contents
33
  data = load_dataset('skadio/optimized_item_selection', data_files='book_recommenders_data/goodreads_1k_data.csv', split='train')
34
  data = data.to_pandas()
 
41
  # TextWiser featurization method to create text embeddings
42
  textwiser = TextWiser(Embedding.TfIdf(), Transformation.NMF(n_components=20, random_state=1234))
43
 
44
+ # Text-based selection with the default configuration
45
+ # The default configuration is optimization_method="exact" and cost_metric ="diverse"
46
+ # By default, multi-level optimization maximizes coverage and diversity as described in (CPAIOR'21, DSO@IJCAI'22)
47
+ # within an upper bound on subset size given as num_features
48
  selector = Selective(SelectionMethod.TextBased(num_features=30, featurization_method=textwiser))
49
 
50
  # Result
 
52
  print("Reduction:", list(subset.columns))
53
  ```
54
  ## Advanced Usages
55
+ Text-based Selection provides access to multiple selection methods.
56
+
57
+ At a high-level, the configurations can be divided into exact, randomized, greedy or cluster-based optimization.
58
+
59
+ ### Exact
60
 
61
  - (Default) Solve for Problem *P_max_cover@t* in **CPAIOR'21** - Selecting a subset of items that
62
  maximizes coverage of labels and maximizes the diversity in latent embedding space within an upper
 
87
  optimization_method='exact',
88
  cost_metric='unicost'))
89
  ```
90
+
91
+ ### Randomized
92
+
93
  - Selecting a subset by performing random selection. If num_features is not set, subset size is defined
94
+ by solving *P_unicost*.
95
  ```python
96
  selector = Selective(SelectionMethod.TextBased(num_features=None, optimization_method='random'))
97
  ```
 
100
  selector = Selective(SelectionMethod.TextBased(num_features=30,
101
  optimization_method='random'))
102
  ```
103
+
104
+ ### Greedy
105
+
106
  - Selecting a subset by adding an item each time using a greedy heuristic in selection with a given
107
  cost_metric, i.e. `diverse` by default or `unicost`. If num_features is not set, subset size is defined
108
+ by solving *P_unicost*.
109
  ```python
110
  selector = Selective(SelectionMethod.TextBased(num_features=None,
111
  optimization_method='greedy',
 
118
  optimization_method='greedy',
119
  cost_metric='unicost'))
120
  ```
121
+
122
+ ### Clustering
123
+
124
  - Selecting a subset by clustering items into a number of clusters and the items close to the centroids
125
+ are selected. If num_features is not set, subset size is defined by solving *P_unicost*. `cost_metric` argument
126
  is not used in this method.
127
  ```python
128
  selector = Selective(SelectionMethod.TextBased(num_features=None, optimization_method='kmeans'))