Speech API/SDK


Mihup’s Speech API and SDK enables developers to convert audio and spoken sentences to text by an easy to use API. The API recognizes Hindi, Bengali and Indian English languages. Optimized to your use case for user base. You can transcribe the text of users dictating to an application’s microphone, enable command-and-control, information discovery through voice, or transcribe audio files. Entire operation can run with or without internet.

Major Indian languages

The system is available for Hindi, Bengali and Indian English at the moment. Support for Other Major Indian languages will be available soon.

Works with any device or application

Speech API can be integrated with any device or application like- phones, pcs, tablets, and IOT devices (e.g- set top box, car, smart TVs, remote, etc.)

Features

Automatic Speech Recognition powered by Machine learning (Speech to text)

• Automatic Speech Recognition (ASR) to power your applications like voice search or speech transcription.
• Vocabulary

Recognizes over 3 Indian languages and variants for Entertainment, connected car, contact centre and e-commerce.
It can be customized for any domain by providing set of most frequent words that are likely to be spoken.

• Wake up word to start the system. Like- ok, mihup.
• Vocabulary
• Inappropriate Content Filtering: Filter inappropriate content in text results for some languages.
• Real-time, Offline or Pre-recorded Audio Support
• Positive and Negative content: Filter inappropriate content ( positive and negative word) in text results.

NLP API/SDK


Mihup's NLP API consists of two parts; Query Parsing and Answer Retrieval which can work independently (specific solutions) and together (complete solution).

Major Indian languages

The system is available for Hindi, Bengali and Indian English at the moment. Support for Other Major Indian languages will be available soon.

Works with any device or application

NLP API can be integrated with any device or application like- phones, pcs, tablets with a working internet connection. The Query Parsing framework works offline when a static model is being used.

Features

Query Parsing:

1. Given a text query, the parsed result will identify the important named entities within the text query and also intent of the query.
2. New domains, entities and intents can be added with training data to improve the capabilities of the system and scale it.

Answer Retrieval:

1. Custom knowledge graph can be built using our API and your data in the specified format.
2. Knowledge graph can be queried with the help of properties of nodes and connections between the nodes thus returning an appropriate answer based on the query.

The complete solution involves first parsing the query and extracted the required information to retrieve an answer from the knowledge graph.

Why Only Mihup?


We understand your language and voice makes it possible to clearly express your needs and wishes. Moreover, being connected is not enough so you can access Mihup's capabilities even if you are offline.