trec

  • คำอธิบาย :

ชุดข้อมูลการจำแนกประเภทคำถามของ Text REtrieval Conference (TREC) ประกอบด้วยคำถาม 5500 ข้อในชุดฝึกอบรม และอีก 500 ชุดสำหรับชุดทดสอบ ชุดข้อมูลมี 6 ป้ายกำกับ 47 ป้ายกำกับระดับ 2 ความยาวเฉลี่ยของแต่ละประโยคคือ 10 ขนาดคำศัพท์ 8700 ข้อมูลถูกรวบรวมจากสี่แหล่ง: คำถามภาษาอังกฤษ 4,500 ข้อที่เผยแพร่โดย USC (Hovy et al., 2001) คำถามที่สร้างขึ้นเองประมาณ 500 ข้อสำหรับชั้นเรียนที่หายากไม่กี่ประเภท 894 TREC 8 และ ข้อสอบ TREC 9 และชุดข้อสอบ TREC 10 จำนวน 500 ข้อ

แยก ตัวอย่าง
'test' 500
'train' 5,452
  • โครงสร้างคุณลักษณะ :
FeaturesDict({
    'label-coarse': ClassLabel(shape=(), dtype=int64, num_classes=6),
    'label-fine': ClassLabel(shape=(), dtype=int64, num_classes=47),
    'text': Text(shape=(), dtype=string),
})
  • เอกสารคุณสมบัติ :
คุณสมบัติ ระดับ รูปร่าง Dประเภท คำอธิบาย
คุณสมบัติDict
ฉลากหยาบ ป้ายกำกับคลาส int64
ฉลากดี ป้ายกำกับคลาส int64
ข้อความ ข้อความ สตริง
  • การอ้างอิง :
@inproceedings{li-roth-2002-learning,
    title = "Learning Question Classifiers",
    author = "Li, Xin  and
      Roth, Dan",
    booktitle = "{COLING} 2002: The 19th International Conference on Computational Linguistics",
    year = "2002",
    url = "https://www.aclweb.org/anthology/C02-1150",
}
@inproceedings{hovy-etal-2001-toward,
    title = "Toward Semantics-Based Answer Pinpointing",
    author = "Hovy, Eduard  and
      Gerber, Laurie  and
      Hermjakob, Ulf  and
      Lin, Chin-Yew  and
      Ravichandran, Deepak",
    booktitle = "Proceedings of the First International Conference on Human Language Technology Research",
    year = "2001",
    url = "https://www.aclweb.org/anthology/H01-1069",
}