Basic HTTP server for Tagger, implementation of a Named Entity Recognizer that obtains state-of-the-art performance in NER on the 4 CoNLL datasets (English, Spanish, German and Dutch) without resorting to any language-specific knowledge or resources such as gazetteers.
The model is only loaded once when the server starts. Requests are then pretty quick, around ~150ms on my machine.
GET /?q=Barack Obama is president of the USA
Output:
{
"text": "Barack Obama is president of the USA",
"ranges": [
[ 0, 1, "PER" ],
[ 6, 6, "ORG" ]
],
"words": [
"Barack",
"Obama",
"is",
"president",
"of",
"the",
"USA"
]
}