Grouper Application Training and Leasehold Analysis Guide

Participants and Internal Roles: - Kyle: Providing the training session. - Olivia: Land Admin Intern in her fifth day. She is an OU student commuting from Norman and is currently mentored by Mindy. - Brad: Present for the training session. - Tamara and Jacob: Members of teams that Olivia will rotate through during her internship. - Karen: Assisted Olivia in her transition into the team.
Meeting Environment: - There was an overlapping "trade meeting" involving the whole land team and land admin which was noted as being very crowded, leading some attendees to sit against the walls. - The training session on the Grouper application was prioritized as being high importance for the attending group.

Definition and Scope: - Grouper is a software tool utilized for leasehold analysis. It currently contains data for over $19,000$ oil and gas leases specifically from the Utica and Appalachia regions. - The application processes uploaded documents to extract text and analyze specific lease provisions and header information.
Core Capabilities: - Text Extraction: The tool pulls verbatim text from PDF documents. - Provision Analysis: It analyzes results to determine the existence and meaning of specific lease provisions. - Future Chatbox Feature: A natural language processing chat element is in development. Once active, users will be able to ask the tool natural language questions regarding leasehold details. - Reasoning Log: In addition to extracting text, the tool provides a "reasoning" column that details the logic used by the machine to reach its conclusion (e.g., why a provision is marked as allowed or committed).

Primary Screens: - The search screen is the primary interface for users. Other screens in the application are designated for backend processing tasks.
Search Features: - Wide Open Search: By default, the tool displays all $19,000$ records until filters are applied. - Text Search: Allows for keyword searching across the entire database of leases. A provided PDF guide outlines specific characters and syntax used to build advanced text searches. - Filter Field: Used to narrow results based on specific data points or criteria. - Select Field: Determines which columns (provision categories) are returned in the results at the bottom of the screen. - Save and Share: Users can save their specific search queries and share direct links to those searches with other team members.
Data Columns Explained: - Exists: Indicates whether the lease contains language pertaining to a specific topic or provision. - Values: Identifies specific numeric or day-based restrictions (e.g., acreage limitations or day thresholds for extensions). - Text: Displays the verbatim text extracted from the PDF file.

Building a Query: - Start typing the name of a provision in the select box. The tool will auto-populate options. - Separate multiple selected fields with a comma. - To pull a whole category of objects, use the text to the left of the backslash in the header options (e.g., using the umbrella category rather than individual sub-checkboxes).
Filtering by Lease List: - For specific operations, such as working on a unit (e.g., the Campbell unit), users can obtain a list of lease numbers from systems like LM10 and paste that list into the text search box to isolate those results.
Sorting and Navigation: - Sort order can be changed by clicking on the column headers. - Row selection: When reviewing text, users must select the specific row to see the corresponding PDF document in the viewer above.
Document View Options: - Card View: A graphic text layout that allows users to see the full extent of the extracted text without horizontal scrolling. - Search Within Document: An icon in the top right allows users to search for specific terms (e.g., "pool") within a single document after it has been selected from the wider list.

Current Limitations Regarding Amendments: - The tool currently analyzes the "four corners" of the specific PDF document uploaded for a given row. - Handling Silence and Subsequent Amendments: A participant (Robert) noted that an old lease (e.g., from $1910$ ) might be silent on pooling, but have a subsequent pooling amendment from $2004$ . Currently, the tool may show the original lease as having no rights; users must still check the physical lease file or QLS for supplemental documents. Work is ongoing to incorporate secondary documents into primary lease rows.
AI Refinement Process: - The development team undergoes cycles of validating results and providing feedback to the AI developers. - Instruction modifications (prompt engineering): If the tool extracts incorrect information, the developer changes the instructions for the AI and reprocesses the provision across the entire dataset.
Instruction for Downloading Documents: - When downloading, ensure the "expand field" checkbox is checked to prevent receiving a condensed or truncated version of the extracted data.

Comparison to QLS (Quality Land System): - Question: How is this different from a standard QQM report from QLS? - Response: QLS only captures data for specific, predetermined questions. Grouper is dynamic; it can extract information on topics that were not previously categorized or thought of during the initial QLS setup.
Interface Feedback: - The speaker acknowledged the current UI looks "old and archaic" (comparing it to the game Oregon Trail). Improvements to the user interface and the chatbot functionality are being developed in parallel with current use.
Feedback Loop: - There is currently no automated "flagging" system within the tool (unlike other platforms like ThoughtTrace). Users are encouraged to share "Safe Search" links with the administration team to highlight errors or suggest improvements.
Future Applications: - Discussions are underway to apply this technology to the Division Order (DO) process to help analyze instruments for TDX coding and deck setup.