Table 1 Feature functions description.
From: Sieve-based relation extraction of gene regulatory networks from biological literature
Name | Description | Options |
|---|---|---|
Target label distribution | Distribution of target labels. | -- |
Starts upper | Does a mention start with an upper case leter. | current, previous mention |
Starts upper twice | Do two consequent mentions start with an upper case letter. | current, previous mention |
Hearst co-occurence [58] | Does the text between the two mentions follow some predefined rules, e.g., mi such as mj. | -- |
Mention token distance | Distance between the two mentions in number of mentions. | -- |
Parse tree mention depth | Depth of the mention within the parse tree. | -- |
Parse tree parent value | Parse tree value of the mention on length l | l ∈ {1, 2, 3} |
Parse tree path | Path values between the two mentions in a parse tree, e.g., DT/NP/NNS/.../NP/NP/VBG. | up to three tokens from every mention |
BSubtilis | If the two mentions are known as B. subtilis, what is the probability of protein-protein interaction using STRING data [29], i.e., very low, low, medium, high, very high. | -- |
IsBSubtilis | Is the current mention known as B. subtilis gene. | -- |
IsBsubtilisPair | Which of the two consequent mentions is known as B. subtilis genes, i.e., left, right, both or none. | -- |