main_blocktest.go 6.8 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242
  1. package main
  2. import (
  3. "fmt"
  4. "jy/extract"
  5. "jy/mongodbutil"
  6. "jy/pretreated"
  7. ju "jy/util"
  8. "log"
  9. "os"
  10. qu "qfw/util"
  11. "regexp"
  12. "time"
  13. )
  14. var f *os.File
  15. var m = map[string]bool{}
  16. func main12() {
  17. //log.Println(pretreated.ProcTitle("以上公告内容如有变动将在相关网络媒体上另行通知凡购买本招标文件的单位必须就此采购项目的相关事宜详细咨询否则参与投标即被视为已经充分了解了招标方的需求中标后承担该文件范围内的所有要求投标前如对招标文件存有疑问请在投标截止日期前三个工作日以实名制书面文件向我公司询问否则视为接受已报名购买招标文件的投标商未递交投标文件或虽递交投标文件但未参加开标大会的投标商不得再参加该项目的采购活动"))
  18. //return
  19. //f, _ = os.OpenFile("./title.txt", os.O_RDWR|os.O_CREATE, 777)
  20. //all()
  21. one()
  22. }
  23. func all() {
  24. m := mongodbutil.MgoFactory(3, 3, 120, "192.168.3.207:27082", "extract_kf")
  25. sess := m.Get()
  26. defer m.Close(sess)
  27. it := sess.DB("extract_kf").C("bidding201901").Find(nil).Iter()
  28. pool := make(chan bool, 5)
  29. count := 0
  30. for temp := make(map[string]interface{}); it.Next(&temp); {
  31. pool <- true
  32. count++
  33. go func(d map[string]interface{}) {
  34. defer func() {
  35. <-pool
  36. }()
  37. com(d)
  38. }(temp)
  39. temp = make(map[string]interface{})
  40. if count%200 == 0 {
  41. log.Println(count)
  42. }
  43. }
  44. log.Println("over...")
  45. time.Sleep(time.Hour)
  46. }
  47. func one() {
  48. m := mongodbutil.MgoFactory(3, 3, 120, "192.168.3.207:27081", "qfw")
  49. d, _ := m.FindById("bidding", "5d41cddba5cb26b9b7ea6631", extract.Fields)
  50. com(*d)
  51. }
  52. func com(doc map[string]interface{}) {
  53. detail := GetDetail(doc)
  54. doc["detail"] = detail
  55. toptype := qu.ObjToString(doc["toptype"])
  56. subtype := qu.ObjToString(doc["subtype"])
  57. if qu.ObjToString(doc["type"]) == "bid" {
  58. toptype = "结果"
  59. }
  60. if toptype == "" {
  61. toptype = "*"
  62. }
  63. e := &extract.ExtractTask{
  64. TaskInfo: &extract.TaskInfo{
  65. Version: "V3.1.2",
  66. VersionId: "5cdd1c70e138234848c1d703",
  67. ProcessPool: make(chan bool, 1),
  68. },
  69. }
  70. e.Id = qu.ObjToString(ju.Config["udptaskid"])
  71. e.InitTaskInfo()
  72. //d.TaskInfo.FDB = db.MgoFactory(3, 5, 600, ext.TaskInfo.FromDbAddr, ext.TaskInfo.FromDB)
  73. //d.TaskInfo.TDB = db.MgoFactory(3, 5, 600, ext.TaskInfo.ToDbAddr, ext.TaskInfo.ToDB)
  74. e.InitRulePres()
  75. e.InitRuleBacks()
  76. e.InitRuleCore()
  77. e.InitBlockRule()
  78. e.InitTag()
  79. e.InitClearFn()
  80. if e.IsExtractCity { //版本上控制是否开始城市抽取
  81. //初始化城市DFA信息
  82. e.InitCityDFA()
  83. e.InitAreaCode()
  84. e.InitPostCode()
  85. }
  86. //质量审核
  87. e.InitAuditFields()
  88. e.InitAuditRule()
  89. e.InitAuditClass()
  90. e.InitAuditRecogField()
  91. //品牌抽取是否开启
  92. ju.IsBrandGoods, _ = ju.Config["brandgoods"].(bool)
  93. j := &ju.Job{
  94. SourceMid: qu.BsonIdToSId(doc["_id"]),
  95. Category: toptype,
  96. CategorySecond: subtype,
  97. Content: qu.ObjToString(doc["detail"]),
  98. SpiderCode: qu.ObjToString(doc["spidercode"]),
  99. //Domain: qu.ObjToString(doc["domain"]),
  100. //Href: qu.ObjToString(doc["href"]),
  101. Title: qu.ObjToString(doc["title"]),
  102. Data: &doc,
  103. City: qu.ObjToString(doc["city"]),
  104. Province: qu.ObjToString(doc["area"]),
  105. Result: map[string][]*ju.ExtField{},
  106. BuyerAddr: qu.ObjToString(doc["buyeraddr"]),
  107. RuleBlock: e.RuleBlock,
  108. }
  109. e.TaskInfo.ProcessPool <- true
  110. pretreated.AnalyStart(j)
  111. e.ExtractProcess(j, nil)
  112. log.Println("=============块信息================")
  113. for _, v := range j.Block {
  114. log.Println("----", v.Title, v.Titles, "----")
  115. if v.ColonKV != nil {
  116. for kk, vv := range v.ColonKV.KvTags {
  117. for _, vvv := range vv {
  118. log.Println("ColonKV", kk, "---", vvv.Key, "---", vvv.Value, "---", vvv.Weight)
  119. }
  120. }
  121. }
  122. if v.SpaceKV != nil {
  123. for kk, vv := range v.SpaceKV.KvTags {
  124. for _, vvv := range vv {
  125. log.Println("SpaceKV", kk, "---", vvv.Key, "---", vvv.Value, "---", vvv.Weight)
  126. }
  127. }
  128. }
  129. if v.TableKV != nil {
  130. for kk, vv := range v.TableKV.KvTags {
  131. for _, vvv := range vv {
  132. log.Println("TableKV", kk, "---", vvv.Key, "---", vvv.Value, "---", vvv.Weight)
  133. }
  134. }
  135. }
  136. //log.Println("Classify", v.Classify)
  137. //log.Println("Tag", v.Tag)
  138. }
  139. log.Println("=============抽取结果================")
  140. set := (e.ResultArr[0][1]["$set"]).(map[string]interface{})
  141. for k, v := range set {
  142. if k == "budget" || k == "bidamount" || k == "winner" || k == "amount" || k == "projectname" || k == "projectcode" || k == "buyer" || k == "buyerperson" || k == "buyertel" || k == "agency" {
  143. log.Println(k, "---", v)
  144. }
  145. }
  146. log.Println("=============抽取结果 result================")
  147. for k, v := range set["result"].(map[string][]*ju.ExtField) {
  148. break
  149. if k != "budget" {
  150. continue
  151. }
  152. for _, vv := range v {
  153. log.Println(k, fmt.Sprintf("%+v", vv))
  154. for kkk, vvv := range vv.ScoreItem {
  155. log.Println("--", kkk, k, fmt.Sprintf("%+v", vvv))
  156. }
  157. log.Println("\n")
  158. }
  159. }
  160. log.Println("=============中标候选人================")
  161. for _, v := range j.Winnerorder {
  162. log.Println(v)
  163. }
  164. log.Println("=============分包================")
  165. for k, v := range j.BlockPackage {
  166. log.Println(k, v)
  167. }
  168. log.Println("=============正文================")
  169. //log.Println(j.Content)
  170. return
  171. for _, v := range j.Block {
  172. if v.ColonKV != nil && v.ColonKV.KvTags != nil {
  173. for kk, vv := range v.ColonKV.KvTags {
  174. for _, vvv := range vv {
  175. log.Println(kk, vvv.Weight, vvv.Value)
  176. }
  177. }
  178. }
  179. if v.TableKV != nil && v.TableKV.KvTags != nil {
  180. for kk, vv := range v.TableKV.KvTags {
  181. for _, vvv := range vv {
  182. log.Println(kk, vvv.Weight, vvv.Value)
  183. }
  184. }
  185. }
  186. if v.SpaceKV != nil && v.SpaceKV.KvTags != nil {
  187. for kk, vv := range v.SpaceKV.KvTags {
  188. for _, vvv := range vv {
  189. log.Println(kk, vvv.Weight, vvv.Value)
  190. }
  191. }
  192. }
  193. }
  194. log.Println(len(j.Block))
  195. return
  196. for _, v := range j.Block {
  197. if m[v.Title] || v.Title == "" {
  198. continue
  199. }
  200. if !regexp.MustCompile("或|和|以?及|与|、|或").MatchString(v.Title) {
  201. //continue
  202. }
  203. m[v.Title] = true
  204. f.WriteString(j.SourceMid + "-----" + v.Title + "---" + fmt.Sprint(v.Titles) + "\n")
  205. continue
  206. for _, kv := range v.ColonKV.Kvs {
  207. log.Println("\n")
  208. log.Println(kv.Key, "---", kv.Value)
  209. log.Println(kv.Line)
  210. log.Println("=======================")
  211. }
  212. }
  213. }
  214. func GetDetail(doc map[string]interface{}) (detail string) {
  215. detail = ""
  216. d1, _ := doc["detail"].(string)
  217. d2, _ := doc["contenthtml"].(string)
  218. if len(d1) >= len(d2) || d2 == "" {
  219. detail = d1
  220. } else {
  221. detail = d2
  222. }
  223. detail = ju.CutLableStr(detail)
  224. detail = ju.NewCut().ClearHtml(detail)
  225. tabs, ration := pretreated.ComputeConRatio(detail, 1)
  226. if len(tabs) > 0 {
  227. newcon, newtabs, newration := pretreated.FindBigText(detail, ration, tabs)
  228. //log.Println(newcon, newtabs, newration)
  229. if newcon != "" && newration == 0 {
  230. detail = newcon
  231. tabs = newtabs
  232. ration = newration
  233. }
  234. }
  235. return detail
  236. }