Google officially releases “ Dataset Search '' that allows searching for data sets for machine learning from the Internet
The key to building algorithms in machine learning is the "data set". More data and time is required to improve the accuracy of the algorithm, but collecting and searching for a sufficiently large data set is a particularly difficult point in machine learning. You can search for such a data set onlineDataset Search"Has been released by Google.
Discovering millions of datasets on the web
When you access Dataset Search, it looks like this.
To search for a dataset, just enter the keyword of the dataset you want to search in the input field. If you enter the word "Temperture" and search,
The left column shows the search results, and the right column shows detailed information of the selected search results. For example, from the search results, “ Cretaceous stable isotopic record and sea surface temperture estimation for ODP Site 207-1259 (ODP(The record of stable isotopes in the Cretaceous period and estimation of sea surface temperature at points 207 to 1259) was selected, and in the right column, the page link and the update date of the dataset, the provider, the creator, the license, the download format, and the data The target period and target area of the set are displayed. Click on the page link …
I was able to jump to the dataset distribution page.
In addition, Dataset Search can search datasets in Japanese, but search results sometimes displayed simple databases and portal sites instead of datasets. For example, when searching for "cancer" below, the links to databases and portal sites were the top search results.
According to Google researcher Natasha Neu, Dataset Search officially completed its beta test, which lasted almost a year, on January 23, 2020. As a feedback from the beta test, the official version of Dataset Search has added the ability to filter search results by dataset update date, download format, license, and free availability. Also, if the dataset is geographic, the map will be displayed in conjunction with Google Maps.
According to Neu, Dataset Search has already been released as an official version instead of a beta version, but the system is still improving, and feedback is received even at the time of article creation.