Example: You have chat data and you want to identify the most popular topics from all conversations, extract email addresses and phone numbers to determine if a topic was discussed in a Positive, Neutral or Negative manner.
Via our point and click Workflow Builder, you could add a Topic and Sentiment model and Parser objects.
Upload your texts with a unique row identifer for each sentence or conversation. AiModelBuilder schedules the job and passes the data to the Workflow. It creates various files with there own system generated keys, all related to your unique row identifier!
It zips up all files and sends you a direct download link to the results by email. Simply import them into your data store to support your offline analysis.
If you were using any other service, that would be a minimum of two API calls, each requiring custom coding to reach the different API services and custom models for each, not to mention the time invloved in building those models. And then your development team would need to code a system to parse the results and then create a data design to keep all those results related to support the data analysis.
AiModelBuilder does all of the above.
And when you need to add a different Model, Extractor or Parser, you can simply modify a Workflow or create a new one to solve another business need! No Coding, No API's!
Select from a variety of AI Models, Extractors or Regular Expression Parser objects. Optionally create your own Models or Extractors.
Sentiment Model: When you want to identify a sentence or conversation that has a Positive, Negative or Neutral sentiment, include the General Sentiment Model in a Workflow.
Topic Extractor: When you want to identify the theme or subject of a sentence include this Extractor in a Workflow. Helps you identify trending statements across a large data set.
Part of Speech Extractor: When you want to identify the nouns, pronouns, adjectives and other parts of speech for each token inlude this Extractor in a Workflow. As you can imagine, this type of extractor produces a ton of results. This extractor generates a one to many set of files. One includes system generated keys linked to your unique row identifer and sentence. The other is linked to the sentences with a list of the many tokens and Penn Treebank Part of Speech. When you include this Extractor and a Sentiment Model you can for example identify which products or services by name were mentioned in a Positive or Negative conversation.
Emotion Model: When you want to identify the mood of a sentence, include the Emotion Model. A variation on a Sentiment Model, this unique model identifies Love, Hate, Suprise and other Emotions in a sentence.
Big 5 Model: When you want to identify the personality of a person based on what they write about themselves, add this model to a Workflow. For example you could pass a cover letter from a resume to determine the writers personality. You could pass a LinkedIn profile summary to determine there temperament.
Urgency Model: Have chat data? Include the Urgency, Topic and Sentiment Model in a Workflow. May identify the top issues, sentiment and conversations that express a need to change sooner, rather than later. Find out if feature enhancements can wait or not.
When building a Custom Extractor, big tech companies will have you tag every occurence of text you wish to extract from your corpus when building an Extraction Model, also known as Named Entity Recognition. When you have 15,000 sentences to wade through, this could take quite some time, not to mention the staffing costs. We took a different approach. Imagine you have 20 texts you would like to extract. Simply search for the first occurence and tag it. Repeat for the other 19. You're done! AiModelBuilder automatically locates every occurence before it trains the model, saving you an immense amount of time, and money.
Each plan includes a Parser Library. Select form an assortment of common Regular Expression patterns. To protect customer privacy in processed results, select a Redact method. Set any Regular Expression to replace texts with either a Custom Label or a Token! AiModelBuilder returns the original sentence, the redacted version, and also identifies which items were replaced and the related Label or Token in the same result! Simply store the undredacted version in a secure datastore, and release the redacted version to your computer geeks or data scientists to support there data analysis.
Processing of your data generates various results. Each includes system generated unique identifiers that are linked to your row id. Some results will output two files, where one includes many items extracted from a sentence, and the other a list of the sentences (one to many relationship). Where a row of text has multiple sentences as determined by our sentence parser model, the system generates a key for each. All results from the different models you include in a Workflow include the same sentence identifier, allowing you to link all the results and potentially to other tables in your database via your unique key.
Select from an assortment of Regular Expresssions to parse texts. A Regular Expression identifies a pattern in texts. For example, you could extract emails, phone numbers, credit card numbers, social insurance numbers, file names, hashtags, postal/zip codes and so much more. Add them to a Workflow as a Task. When patterns are identified, AiModelBuilder extracts texts. No Coding, No APIs! Yes we took care of that. Regular Expressions tested, and ready to do parse, today!