Artificial Intelligence with AWS: Amazon Transcribe and Amazon Comprehend
In the previous post we talked about the power of artificial intelligence applied to cloud and some examples of its integration into process improvements, data analysis and information extraction. But due to the large volume of data generated, undertaking these actions requires scalable solutions that organize and analyze this information. To do this, we tell you about two services with which you can take advantage of artificial intelligence with AWS: Amazon Transcribe y Amazon Understand.
If you found this post interesting, we invite you to Innovating with Apser and AWS
From audio to text without complications
Amazon Transcribe is an AI service based on the voice recognition that generates transcripts, documents or filters from audio materials such as calls, videos or conversations. This helps to identify topics or problems by detecting content, tags in text, among others.
What makes it a powerful tool? Transcribe uses powerful features. machine learning algorithm to generate easy-to-analyze transcripts. Thanks to its integration with call centers and other applications, it allows Identify key themes, create labels, identify the source/channel audio file and even add subtitles.
Based on natural language processing, Transcribe optimizes the accuracy of transcriptions by implementing vocabulary appropriate to each situation (product names, people in the conversation, technical terminology, etc.).
text analytics
Amazon Understand is another artificial intelligence solution with AWS that, once we have the transcriptions done, will help us analyze and discover valuable information. Like Transcribe, Identify patterns within text using NLP e indexes key phrases and sentiments to organize documents by topic or issue.
For example, Comprehend could analyze the text of a customer support call transcript to identify key phrases that suggest whether the customer had a positive or negative experience.
Among its main functions we can find APIs for entity recognition, that is, identifying people, places or locations and categorizing them from the text. Like the detailed opinion analysis, which provides sentiment information based on keywords. Let's look at some important functions:
sentiment analysis
Function with which we can define and categorize emotions within the text from a algorithm that will act as a “bag of words” model. This model will then go through each word in the text to identify sentiments based on values (neutral, positive, negative, and mixed). The sum of these values defines the overall feeling of the text.One advantage of Comprehend is that by integrating NLP and machine learning it goes beyond considering words in isolation, as it takes into account the structure and context of sentences to better understand the text.
Furthermore, the analysis goes much further since this solution integrates Three types of approaches: experience-based, rules-based and hybridThe first uses ML rules and opinion classification algorithms (neural networks) to train intelligence in identifying emotions, based on previous data and constant learning to be more precise.
Furthermore, the rules-based approach identifies, classifies and scores specific keywords from a predetermined lexicon, i.e. collections of positive and negative opinions with which the software searches for words that correspond to these categories to give a summary of the opinions.
Identity Extraction
This involves classifying text into entities such as people, quantities, dates, among others. To accomplish this task, Comprehend works by selecting and categorizing the entities and providing a confidence score. The most interesting thing is that the tool not only extracts quantities or numerical information, but also gives them a context or searches for keywords within the text that give meaning to those numbers.
Even in terms of security, Amazon Comprehend prioritizes the anonymization of names, adding values such as SKUs or reference numbers to be able to categorize them.
Artificial Intelligence Experts with AWS
From apser, our experts have worked on the deployment of both solutions for the Analysis of the audiovisual content catalogue from one of our clients. The objective was a categorization of themes and moods to create a more accessible library, both for users and for the company. Users now have the possibility to select the specific minute or second of the topic of their interest. In addition to having a more agile traceability of the topics.
Do you want to start innovating with AI? Don't hesitate to contact usWe will be happy to hear about your project.
