有个人长的像洋葱,走着走着就哭了…….

elastic7.4 中 关于相似度算法分值中bootst是如何计算的

Elasticsearch | 作者 cheerfulbug | 发布于2019年12月02日 | 阅读数:1197

在分析匹配的过程中发现除了 TF/IDF 分值外,还有一项分值 为4.4, 描述为boost的值,请问该值是如何计算取得的?
 

"_explanation": {
                    "value": 77.37436,
                    "description": "sum of:",
                    "details": [
                        {
                            "value": 12.705416,
                            "description": "weight(resourceDesc:5 in 172) [PerFieldSimilarity], result of:",
                            "details": [
                                {
                                    "value": 12.705416,
                                    "description": "score(freq=2.0), product of:",
                                    "details": [
                                        {
                                            "value": 4.4,
                                            "description": "boost",
                                            "details": []
                                        },
                                        {
                                            "value": 4.2525687,
                                            "description": "idf, computed as log(1 + (N - n + 0.5) / (n + 0.5)) from:",
                                            "details": [
                                                {
                                                    "value": 3,
                                                    "description": "n, number of documents containing term",
                                                    "details": []
                                                },
                                                {
                                                    "value": 245,
                                                    "description": "N, total number of documents with field",
                                                    "details": []
                                                }
                                            ]
                                        },
                                        {
                                            "value": 0.67902356,
                                            "description": "tf, computed as freq / (freq + k1 * (1 - b + b * dl / avgdl)) from:",
                                            "details": [
                                                {
                                                    "value": 2.0,
                                                    "description": "freq, occurrences of term within document",
                                                    "details": []
                                                },
                                                {
                                                    "value": 1.2,
                                                    "description": "k1, term saturation parameter",
                                                    "details": []
                                                },
                                                {
                                                    "value": 0.75,
                                                    "description": "b, length normalization parameter",
                                                    "details": []
                                                },
                                                {
                                                    "value": 536.0,
                                                    "description": "dl, length of field (approximate)",
                                                    "details": []
                                                },
                                                {
                                                    "value": 747.4367,
                                                    "description": "avgdl, average length of field",
                                                    "details": []
                                                }
                                            ]
                                        }
                                    ]
                                }
                            ]
                        },
 
QQ截图20191202192414.png
已邀请:

要回复问题请先登录注册