Speech-to-Intent engine fuses Natural Language Understanding with Speech-to-Text, outperforming Google Dialogflow, Amazon Lex, and IBM Watson.
Rhino Speech-to-Intent infers user intents from utterances, allowing users to interact with applications via voice. Rhino Speech-to-Intent understands complex voice commands, such as "find the maintenance checklist for Boeing 707" or "call 987 655 4433".
1o = pvrhino.create(2 access_key,3 context_path)4
5while not o.process(audio()):6 pass7
8inference = o.get_inference()1let o = new Rhino(2 accessKey,3 contextPath);4
5while (!o.process(audio())) { }6
7let inference = o.getInference();1RhinoManagerCallback2callback =3new RhinoManagerCallback() {4 @Override5 public void invoke(6 RhinoInference inference) {7 // Inference callback8 }9}10
11RhinoManager o =12new RhinoManager.Builder()13 .setAccessKey(accessKey)14 .setContextPath(contextPath)15 .build(16 appContext,17 callback);18
19o.start()1let o = RhinoManager(2 accessKey: accessKey,3 contextPath: contextPath,4 onInferenceCallback:5 { inference in6 // Inference callback7 }8 );9
10try o.start()1const {2 inference,3 contextInfo,4 isLoaded,5 isListening,6 error,7 init,8 process,9 release,10} = useRhino();11
12await init(13 accessKey,14 context,15 model16);17
18await process();19
20useEffect(() => {21 if (inference !== null) {22 // Handle inference23 }24}, [inference]);1RhinoManager o =2 await RhinoManager.create(3 accessKey,4 contextPath,5 (inference) => {6 // Inference callback7 });8
9await o.process()1let o =2 await RhinoManager.create(3 accessKey,4 contextPath,5 (inference) => {6 // Inference callback7 });8
9await o.process()1RhinoManager o =2 RhinoManager.Create(3 accessKey,4 contextPath,5 (inference) => {6 // Inference callback7 });8
9o.Start();1Rhino o = Rhino.Create(2 accessKey,3 contextPath);4
5while (o.Process(AudioFrame())) {6
7}8
9Inference inference =10 o.GetInference();1Rhino o = new Rhino.Builder()2 .setAccessKey(accessKey)3 .setContextPath(contextPath)4 .build();5
6while (!o.process(audioFrame())) {7
8}9
10RhinoInference inference =11 o.getInference();1pv_rhino_init(2 access_key,3 model_path,4 context_path,5 sensitivity,6 require_endpoint,7 &rhino);8
9while (true) {10 pv_rhino_process(11 rhino,12 audio_frame(),13 &is_finalized);14
15 if (is_finalized) {16 pv_rhino_get_intent(17 rhino,18 &intent,19 &num_slots,20 &slots,21 &values);22 }23}Traditional voice systems use a two-step process: Speech-to-Text (convert voice to text) and Natural Language Understanding (analyze text for intent).
Rhino Speech-to-Intent uses an innovative single-step approach: Direct Speech-to-Intent - fusing speech recognition and intent detection into one optimized process. This unified approach eliminates the accuracy loss and latency issues common in conventional systems.
Natural language understanding deals with meaning, i.e., comprehending users’ intent. Researchers initially started with understanding user intents from the text. While spoken language understanding is a more specific term to refer to understanding user intent from speech, many people, including the industry and researchers, still use natural language understanding for both text and speech data. This is mainly due to the conventional approach of running speech-to-text and natural language understanding engines subsequently.
Intent Detection is a subtask of natural language processing and a critical component of any task-oriented system. Natural language understanding solutions match users' utterances with one of the predefined classes by understanding the user’s goal (i.e., intention). After matching utterances with intents, the software can initiate a task to achieve users’ goals. For example, users with the intention to turn the lights off may say: “Turn the lights off.”, “Switch off the lights.”, “Can you please turn the lights off?”. Intent detection captures the users’ goal: “change the state of the lights from on to off” despite the different ways to communicate it.
Rhino Speech-to-Intent is a more accurate, resource-efficient, and faster alternative to Amazon Lex, Google DialogFlow, or other NLU engines for use-case-specific intent detection. Picovoice’s Free Trial allows enterprises to evaluate Rhino Speech-to-Intent and compare it with the alternatives. However, if you’re still not sure how to overcome the limitations of Amazon Lex, Google DialogFlow, and other NLU engines with Rhino Speech-to-Intent or need help with migration, leverage Picovoice’s Consulting Services!
Rhino Speech-to-Intent -as the name suggests, converts speech into intent directly without relying on text, eliminating the need for text representation. Rhino Speech-to-Intent uses the modern end-to-end approach to infer intents and intent details directly from spoken commands. This enables developers to train jointly optimized automatic speech recognition (ASR) and natural language understanding (NLU) engines tailored to their specific domain, achieving higher accuracy.
Rhino Speech-to-Intent excels in use-case-specific applications, such as voice-enabled coffee machines or surgical robots, which involve a limited number of commands, offering high accuracy with minimal resources. In contrast, open-domain applications like voice-enabled ChatGPT handle a wide range of topics and variations. Thus, we recommend Cheetah Streaming Speech-to-Text and picoLLM for such applications.
Intents, expressions, and slots are commonly used in conversational AI and across various engines such as Amazon Lex, IBM Watson, Google Dialogflow, or Rasa NLU. They’re used to build voice assistants or bots. You can check out the Rhino Speech-to-Intent Syntax Cheat Sheet to start building or the Picovoice Glossary to learn the terminology.
Rhino Speech-to-Intent processes voice data locally on the device.
Picovoice docs, blog, Medium posts, and GitHub are great resources to learn about voice AI, Picovoice technology, and how to start building with voice commands. Enterprise customers get dedicated support specific to their applications from Picovoice Product & Engineering teams. While Picovoice customers reach out to their contacts, prospects can also purchase Enterprise Support before committing to any paid plan.
Rhino Speech-to-Intent supports English, French, German, Italian, Japanese, Korean, Chinese (Mandarin), Portuguese, and Spanish.
Reach out to Picovoice Consulting team to get a custom language model trained for your use case.