|
Feature | Group | Description
|
|
ifQueryTerm | ★ | Whether the snippet term is a query term |
ifResulttitle | ★ | Whether the snippet term is a term in the result title |
ifInWiki | ★ | Whether the snippet term appears in the Wikipedia content of the query |
wikiCount | ★ | Frequency of the snippet term in the Wikipedia content of the query |
ifInBaidu | ★ | Whether the snippet term appears in the Baidu Baike content of the query |
baiduCount | ★ | Frequency of the snippet term in the Baidu Baike content of the query |
ifSearchRec | ★ | Whether the snippet term appears in the search recommendations of the query |
searchRecCount | ■ | Frequency of the snippet term in the search recommendations of the query |
queryTermJaccard | ■ | Jaccard distance between the snippet term and query |
queryTermEdit | ■ | Edit distance between the snippet term and query |
searchResultsOverlap | ■ | Number of shared results of the search result lists obtained by submitting the snippet term and query to commercial search engine |
wikiTfIdf | ■ | Tf-idf value of the snippet term in the Wikipedia corpus (Tf value is calculated as the frequency of the snippet term in the Wikipedia content of the query Wikipedia contents of all the queries used in our experiment are used to calculate the Idf value) |
baiduTfIdf | ■ | Tf-idf value of the snippet term in the Baidu Baike corpus. Similar to wikiTfIdf |
searchRecTfIdf | ■ | Tf-idf value of the snippet term in the search recommendation corpus. Similar to wikiTfIdf |
termTermW2V | ◆ | Cosine similarities between the snippet term vector and query term vectors (if the query is composed of n terms after segmentation, then we will get n cosine similarities) |
termTermProW2V | ◆ | Average, top 3 average, medium, maximum and minimum of termTermW2V
|
queryTermW2V | ◆ | The cosine similarity between the query vector and snippet term vector (if the query is composed of n terms after segmentation, we use the average vector of the n term vectors to be the query vector)
|
resultTitleTermW2V | ◆ | The cosine similarity between the title vector and snippet term vector (if the title is composed of n terms after segmentation, we use the average vector of the n term vectors to be the title vector)
|
searchRecW2V | ◆ | The cosine similarities between the snippet term and the search recommendation corpus. Similar to queryTermProW2V |
|