Generating AI Summaries for Files

Trackops offers the ability to generate AI summaries for files, which allows you to quickly get an overall analysis of a file based on certain parameters you define.  Once generated, summaries can be used to create case updates which can be used in reports or for client updates. 

Note: Currently, Trackops only supports the ability to summarize videos.

Generating a File Summary

To generate a file summary, navigate to the Files tab on a case, then click the name of the file to view the details.  If the file is eligible for summarization (see File Summarization Requirements below), you can generate a new summary by clicking New > New AI Summary.

• Access Group - Users will only be able to see summaries in access groups their user role is a part of.
• Title - The title of the summary.  Use something clear and easy to pick from a list
• Date - The date the file was created, or the start time of the video.
• Type - The type of media this file represents (i.e. surveillance video).
• Description - A detailed description of the file and what is contained inside it. (See Best Practices for AI Summary Generation below)
• Summary Objective - A description of what you want to have generated.  (See Best Practices for AI Summary Generation below)

The system will process the summary and it will be listed under the AI Summaries section at the bottom of the file overview.

The time it takes to summarize a file is largely dependent on several criteria:

• The size and length of the file.  Longer and more complex files require more time to analyze.
• The type of information to analyze as defined by the summary objective.  Pinpointing specific items or activities to analyze or extract can be quicker and more accurate than letting AI determine its own criteria for analysis.

Once the summary has processed, you'll be able to click on the summary title to read the output.  If it's close to how you'd like it, you can click the Create Case Update button to create a new case update from this summary and make any necessary changes.

If the summary does not look right or needs refinement, scroll to the bottom of the summary and click the Re-Generate Summary link to refine your requirements and try again.  Re-generating a summary will note erase the previous summary, but instead will create a new summary.

File Summarization Requirements

To generate the best summarization, you should adhere to the following requirements when creating files intended to be summarized in Trackops.

Best Practices for AI Summary Generation

Best Practices For Summarizing Videos

• Keep your videos as short as possible:  AI does a much better job of summarizing shorter videos (clips). This is especially true when dealing with timestamps.  AI does not analyze videos frame by frame, so longer videos experience more "drift" and AI generated timestamps become inaccurate.  Shorter videos also allow AI to focus on one set of activities or setting.
• Don't use merged videos: If you've generated a merge file from a number of clips, consider summarizing the individual clips instead of the merge file.  AI has a hard time understanding when scenes jump around or are hard to follow.
• Use good video descriptions: When entering a description of the video, clearly describe the subject of the video and the circumstances surrounding why you are capturing the video.  AI does a much better job of summarizing videos when it understands why it is watching the video (just like humans).  For example, if you recorded a video of an injured claimant, describe the injury, limitations, and what you are watching for.  If you want to guide AI in its determination, you might also let it know if you think the video contains fraud.
• Clearly state the summary objective: When entering a summary objective, define what you want AI to output (i.e. Generate a report..., Summarize activities related to...,  Generate one paragraph about..., etc...).  In addition, if you want to have AI focus on certain aspects of the video, indicate what types of information you want to analyze.  For example, if your Video Description outlined an injury and limitations of a claimant's right ankle, you might tell AI to pay special attention to the right ankle, highlight any signs of limping or discomfort, or if there appears to be any activities that violate medical restrictions.