Init

.gitignore

@@ -0,0 +1,2 @@
+.DS_Store
+/node_modules

app.js

@@ -0,0 +1,19 @@
+const express = require('express')
+const app = express()
+
+app.get('/', (req, res) => {
+
+    const num = req.query.num;
+    const { spawn } = require('child_process');
+    const pyProg = spawn('python',['./mixer.py', num]);
+
+    pyProg.stdout.on('data', function(data) {
+
+        // console.log(data.toString());
+        res.write(data);
+        // res.end('end');
+
+    });
+})
+
+app.listen(4000, () => console.log('app listening on port 4000!'))

mixer.py

@@ -0,0 +1,72 @@
+import json
+import random
+import sys
+
+# Form a new dialog using paraphrases according to the given index (specify dialogs)
+def formDialog(index):
+  dialog = flexData[index]
+  newDialog = []
+  for sen in dialog[1]:
+    numPara = len(sen[2])
+    newSen = sen[4]
+    if numPara > 0:
+      ranSenInx = random.sample(xrange(0, numPara - 1), 1)[0]
+      newSen = sen[2][ranSenInx]
+    newDialog.append([sen[3], sen[0], sen[1], newSen])
+  newDialogs.append(newDialog)
+
+inputName = "AllFlexData_ProfileID.json"    # The input file name
+numWantDialog = (int)(sys.argv[1])         # How many new dialogs do you want to generate
+flexData = []
+# Read input json file
+with open(inputName) as json_file:
+    allData = json.load(json_file)
+    for ad in allData:    # for each dataset
+      for data in ad["data"]:    # for each dialog
+        flexData.append([data["dialogName"], []])
+        for sen in data["content"]:     # for each sentence in that dialog
+          paras = [sen["role"], sen["dialogAct"], [], 0, sen["oriSen"]]
+          sql = "N"
+          if "sql_annotation" in sen:
+            if len(sen["sql_annotation"]) > 0:
+              sql = "Y"
+          for para in sen["paraphrases"]:       # for each paraphrase of this sentence
+            paras[2].append(para)
+          paras[3] = sql
+          index = len(flexData) - 1
+          flexData[index][1].append(paras)
+
+numDialogs = len(flexData)
+divide = numWantDialog / numDialogs
+remainder = numWantDialog % numDialogs
+newDialogs = []
+
+# randomly select remainder number of dialogs
+remDialogs = random.sample(xrange(0, numDialogs), remainder)
+for index in remDialogs:
+  formDialog(index)
+
+# if # of wanted paraphrasing-dialogs > # of original dialogs, then every original dialog will be reformed accordingly
+for i in range(0, divide):
+  for j in range(0, numDialogs):
+    formDialog(j)
+
+# output new dialogs
+dic = []
+for i in range(0, len(newDialogs)):
+  newDialog = {}
+  dialogId = "dialog" + str(i)
+  newDialog["dialogId"] = dialogId
+  newDialog["dialog"] = []
+  for sen in newDialogs[i]:
+    newDialog["dialog"].append({"sql" : sen[0], "role" : sen[1], "dialogAct" : sen[2], "sen" : sen[3]})
+  dic.append(newDialog)
+
+# Write to output json file
+# file = open("NewDialogs.json", "w")
+# file.write(json.dumps(dic, sort_keys=True, indent=2))
+
+# Send data back to node
+dataToSendBack = json.dumps(dic, sort_keys=True, indent=2)
+print(dataToSendBack)
+sys.stdout.flush()