Source code

001/*
002 *  $Rev: 1028 $:     Revision of last commit
003 *  $Author: tgutwin $:  Author of last commit
004 *  $Date: 2015-11-01 15:32:08 -0800 (Sun, 01 Nov 2015) $:    Date of last commit
005 *  $URL: svn://svn.webarts.bc.ca/open/trunk/projects/WebARTS/ca/bc/webarts/tools/sphinx/DialogDemo.java $
006 *
007 *  refactored by Tom Gutwin
008 *  Author Alexander Solovets
009 *
010 *  Copyright 1999-2015 Carnegie Mellon University.
011 *  Portions Copyright 2002-2008 Sun Microsystems, Inc.
012 *  Portions Copyright 2002-2008 Mitsubishi Electric Research Laboratories.
013 *  Portions Copyright 2013-2015 Alpha Cephei, Inc.
014 *
015 *  All Rights Reserved.  Use is subject to license terms.
016 *
017 *  Redistribution and use in source and binary forms, with or without
018 *  modification, are permitted provided that the following conditions
019 *  are met:
020 *
021 *  1. Redistributions of source code must retain the above copyright
022 *     notice, this list of conditions and the following disclaimer.
023 *
024 *  2. Redistributions in binary form must reproduce the above copyright
025 *     notice, this list of conditions and the following disclaimer in
026 *     the documentation and/or other materials provided with the
027 *     distribution.
028 *
029 *  3. Original authors' names are not deleted.
030 *
031 *  4. The authors' names are not used to endorse or promote products
032 *     derived from this software without specific prior written
033 *     permission.
034 *
035 *  This work was supported in part by funding from the Defense Advanced
036 *  Research Projects Agency and the National Science Foundation of the
037 *  United States of America, the CMU Sphinx Speech Consortium, and
038 *  Sun Microsystems, Inc.
039
040 *  CARNEGIE MELLON UNIVERSITY, SUN MICROSYSTEMS, INC., MITSUBISHI
041 *  ELECTRONIC RESEARCH LABORATORIES AND THE CONTRIBUTORS TO THIS WORK
042 *  DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ALL
043 *  IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL
044 *  CARNEGIE MELLON UNIVERSITY, SUN MICROSYSTEMS, INC., MITSUBISHI
045 *  ELECTRONIC RESEARCH LABORATORIES NOR THE CONTRIBUTORS BE LIABLE FOR
046 *  ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
047 *  WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
048 *  ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT
049 *  OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
050 *
051 */
052package ca.bc.webarts.tools.sphinx;
053
054import java.util.HashMap;
055import java.util.Map;
056
057import edu.cmu.sphinx.api.Configuration;
058import edu.cmu.sphinx.api.LiveSpeechRecognizer;
059import edu.cmu.sphinx.api.SpeechResult;
060import edu.cmu.sphinx.result.WordResult;
061
062
063public class DialogDemo
064{
065
066  private static final String ACOUSTIC_MODEL = "resource:/edu/cmu/sphinx/models/en-us/en-us";
067  //private static final String DICTIONARY_PATH = "resource:/edu/cmu/sphinx/models/en-us/cmudict-en-us.dict";
068  private static final String DICTIONARY_PATH = "resource:/ca/bc/webarts/tools/sphinx/cmudict-en-us_ca.dict";
069  private static final String GRAMMAR_PATH = "resource:/ca/bc/webarts/tools/sphinx/";
070  private static final String LANGUAGE_MODEL = "resource:/ca/bc/webarts/tools/sphinx/weather.lm";
071
072  /** The grammer file to use for digits recogognition. **/
073  private static String digitsGrammerFilename = "number.grxml"; //"number.grxml";  //"digits.grxml";
074
075  private static final Map<String, Integer> DIGITS = new HashMap <String, Integer>();
076  static
077  {
078    DIGITS.put("oh", 0);
079    DIGITS.put("zero", 0);
080    DIGITS.put("one", 1);
081    DIGITS.put("two", 2);
082    DIGITS.put("three", 3);
083    DIGITS.put("four", 4);
084    DIGITS.put("five", 5);
085    DIGITS.put("six", 6);
086    DIGITS.put("seven", 7);
087    DIGITS.put("eight", 8);
088    DIGITS.put("nine", 9);
089    DIGITS.put("ten", 10);
090    DIGITS.put("twenty", 20);
091    DIGITS.put("thirty", 30);
092    DIGITS.put("fourty", 40);
093    DIGITS.put("fifty", 50);
094    DIGITS.put("sixty", 60);
095    DIGITS.put("seventy", 70);
096    DIGITS.put("eighty", 80);
097    DIGITS.put("ninety", 90);
098    DIGITS.put("one hundred", 100);
099    DIGITS.put("two hundred", 200);
100    DIGITS.put("three hundred", 300);
101    DIGITS.put("four hundred", 400);
102    DIGITS.put("five hundred", 500);
103    DIGITS.put("six hundred", 600);
104    DIGITS.put("seven hundred", 700);
105    DIGITS.put("eight hundred", 800);
106    DIGITS.put("nine hundred", 900);
107  }
108
109
110  private static double parseNumber(String[] tokens)
111  {
112    StringBuilder sb = new StringBuilder();
113
114    for (int i = 1; i < tokens.length; ++i)
115    {
116      if (tokens[i].equals("point") || tokens[i].equals("decimal") || tokens[i].equals("dot") )
117      {
118        sb.append(".");
119      }
120      else
121      {
122        sb.append(DIGITS.get(tokens[i]));
123      }
124    }
125
126    return Double.parseDouble(sb.toString());
127  }
128
129
130  private static void recognizeDigits(LiveSpeechRecognizer recognizer)
131  {
132    System.out.println("Digits recognition (using GrXML)");
133    System.out.println("--------------------------------");
134    System.out.println("Example: one two three");
135    System.out.println("Say \"101\" to exit");
136    System.out.println("--------------------------------");
137
138    SpeechResult speechResult = null;
139    String utterance = null;
140
141    recognizer.startRecognition(true);
142    boolean notDone = true;
143    while (notDone)
144    {
145      speechResult = recognizer.getResult();
146      System.out.print("Got A Result... ");
147      if(speechResult!=null)
148      {
149        //System.out.print("      List of recognized words and their times: ");
150        for (WordResult r : speechResult.getWords())
151        {
152            System.out.print("   "+r+ ", ");
153        }
154
155        utterance = speechResult.getHypothesis();
156        if (utterance.equals("one zero one") || utterance.equals("one oh one"))
157        {
158          notDone = false;
159          System.out.print("Got A 101.................  EXIT Digits ");
160        }
161        else
162        {
163          System.out.print(utterance);
164        }
165      }
166      System.out.println();
167    }
168    recognizer.stopRecognition();
169  }
170
171
172  private static void recognizerBankAccount(LiveSpeechRecognizer recognizer)
173  {
174    System.out.println("This is bank account voice menu");
175    System.out.println("-------------------------------");
176    System.out.println("Example: balance");
177    System.out.println("Example: withdraw zero point five");
178    System.out.println("Example: deposit one two three");
179    System.out.println("Example: back");
180    System.out.println("-------------------------------");
181
182    double savings = .0;
183    recognizer.startRecognition(true);
184
185    while (true)
186    {
187      String utterance = recognizer.getResult().getHypothesis();
188      if (utterance.endsWith("back"))
189      {
190        break;
191      }
192      else if (utterance.startsWith("deposit"))
193      {
194        double deposit = parseNumber(utterance.split("\\s"));
195        savings += deposit;
196        System.out.format("Deposited: $%.2f\n", deposit);
197      }
198      else if (utterance.startsWith("withdraw"))
199      {
200        double withdraw = parseNumber(utterance.split("\\s"));
201        savings -= withdraw;
202        System.out.format("Withdrawn: $%.2f\n", withdraw);
203      }
204      else if (!utterance.endsWith("balance"))
205      {
206        System.out.println("Unrecognized command: " + utterance);
207      }
208
209      System.out.format("Your savings: $%.2f\n", savings);
210    }
211
212    recognizer.stopRecognition();
213  }
214
215
216  private static void recognizeWeather(LiveSpeechRecognizer recognizer)
217  {
218    System.out.println("Try some forecast. End with \"the end\"");
219    System.out.println("-------------------------------------");
220    System.out.println("Example: mostly dry some fog patches tonight");
221    System.out.println("Example: sunny spells on wednesday");
222    System.out.println("-------------------------------------");
223
224    recognizer.startRecognition(true);
225    while (true)
226    {
227      String utterance = recognizer.getResult().getHypothesis();
228      if (utterance.equals("the end"))
229      {
230        break;
231      }
232      else
233      {
234        System.out.println(utterance);
235      }
236    }
237    recognizer.stopRecognition();
238  }
239
240
241  public static void main(String[] args) throws Exception
242  {
243    Configuration configuration = new Configuration();
244    configuration.setAcousticModelPath(ACOUSTIC_MODEL);
245    configuration.setDictionaryPath(DICTIONARY_PATH);
246    configuration.setGrammarPath(GRAMMAR_PATH);
247
248    configuration.setUseGrammar(true);// use the fixed grammer instead of language model
249    configuration.setGrammarName("dialog");
250    LiveSpeechRecognizer jsgfRecognizer = new LiveSpeechRecognizer(configuration);
251
252    configuration.setUseGrammar(true); // use the fixed grammer instead of language model
253    configuration.setGrammarName(digitsGrammerFilename);
254    LiveSpeechRecognizer grxmlDigitsRecognizer = new LiveSpeechRecognizer(configuration);
255
256    configuration.setUseGrammar(false);
257    configuration.setLanguageModelPath(LANGUAGE_MODEL);
258    LiveSpeechRecognizer lmWeatherRecognizer = new LiveSpeechRecognizer(configuration);
259
260    boolean notDone = true;
261    SpeechResult speechResult = null;
262    String utterance = null;
263    jsgfRecognizer.startRecognition(true);
264    while (notDone)
265    {
266      System.out.println("Choose menu item:");
267      System.out.println("Example: go to the bank account");
268      System.out.println("Example: exit the program");
269      System.out.println("Example: weather forecast");
270      System.out.println("Example: digits\n");
271
272      // this blocks until a result is returned
273      speechResult = jsgfRecognizer.getResult();
274      System.out.print("Got A Result... ");
275      //System.out.print("      List of recognized words and their times: ");
276      for (WordResult r : speechResult.getWords())
277      {
278          System.out.print("   "+r+ ", ");
279      }
280      System.out.println();
281
282      if(speechResult!=null)
283      {
284        utterance = speechResult.getHypothesis();
285        if (utterance!=null )
286        {
287          /*System.out.println("\nBest 3 hypothesis:");
288          for (String s : speechResult.getNbest(3))
289            System.out.println("   "+s);
290          */
291
292          if (utterance.startsWith("exit"))
293          {
294            notDone = false;
295          }
296
297          else if (utterance.equals("digits"))
298          {
299            jsgfRecognizer.stopRecognition();
300            recognizeDigits(grxmlDigitsRecognizer);
301            jsgfRecognizer.startRecognition(true);
302          }
303
304          else if (utterance.equals("bank account"))
305          {
306            jsgfRecognizer.stopRecognition();
307            recognizerBankAccount(jsgfRecognizer);
308            jsgfRecognizer.startRecognition(true);
309          }
310
311          else if (utterance.endsWith("weather forecast"))
312          {
313            jsgfRecognizer.stopRecognition();
314            recognizeWeather(lmWeatherRecognizer);
315            jsgfRecognizer.startRecognition(true);
316          }
317
318          else
319            System.out.println("Utterance did not match ("+utterance+")");
320        }
321        else
322          System.out.println("Utterance was null");
323      }
324    }
325
326    jsgfRecognizer.stopRecognition();
327  }
328
329}