实体返回的沃森NLU顺序

Question

我使用沃森NLU和一个自定义模型从简历中提取教育实体。

到目前为止，它运行良好，但如果有多个提到的教育，它没有返回的结果有序。

例如，如果简历中包含高中和大学，它将按错误的顺序返回数据，如下所示：

 {
    "type": "GraduationYear",
    "text": "2007",
    "disambiguation": {
        "subtype": [
        "Date"
        ]
    },
    "count": 1
},
{
    "type": "Institution",
    "text": "Rahman Colledge",
    "disambiguation": {
        "subtype": [
        "CollegeUniversity"
        ]
    },
    "count": 1
},
{
    "type": "GraduationYear",
    "text": "2017",
    "disambiguation": {
        "subtype": [
        "Date"
        ]
    },
    "count": 1
},

{
    "type": "Institution",
    "text": "MIS Highschool",
    "disambiguation": {
        "subtype": [
        "School"
        ]
    },
    "count": 1
},
{
    "type": "EducationLevel",
    "text": "Diploma",
    "disambiguation": {
        "subtype": [
        "Degree"
        ]
    },
    "count": 1
},
{
    "type": "EducationLevel",
    "text": "High School Certfication",
    "disambiguation": {
        "subtype": [
        "Degree"
        ]
    },
    "count": 1
}

我如何排序从模型本身发现的结果？

Answer 1

在NLU中有一个新的mentions参数，该参数在设置为true时返回文本中实体的位置。然后，可以使用每个实体的开始索引对它们进行排序。请参阅NLU引用。